Timeless Tales
Timeless Tales
Timeless Tales
Designing the end-to-end architecture for an AI creative studio. From category strategy to character consistency systems to the pipeline that serves a manga artist, a professional filmmaker, and a brand founder from the same platform.
Designing the end-to-end architecture for an AI creative studio. From category strategy to character consistency systems to the pipeline that serves a manga artist, a professional filmmaker, and a brand founder from the same platform.
Designing the end-to-end architecture for an AI creative studio. From category strategy to character consistency systems to the pipeline that serves a manga artist, a professional filmmaker, and a brand founder from the same platform.
AI Creative Platform
Product Strategy
Product Architecture
Product Architecture
Conversational AI
AI Transparency
AI Transparency
Anchor-Based Consistency
Conversational AI
AI Creative Platform
Product Strategy
Product Architecture
Conversational AI
AI Transparency
Anchor-Based Consistency
Role
Lead Product Designer
Platform
Web Application
Type
Sole Designer, End-to-End
Scope
Full Architecture, Research, Information Architecture, Design Systems, Handoff
Role
Role
Lead Product Designer
Lead Product Designer
Platform
Platform
Web Application
Web Application
Type
Type
Sole Designer, End-to-End
Sole Designer, End-to-End
Scope
Scope
Full Architecture, Research, Information Architecture, Design Systems, Handoff
Full Architecture, Research, Information Architecture, Design Systems, Handoff
Overview
Overview
A product that stopped people before they started
A product that stopped people before they started
A product that stopped people before they started
The CEO came with a product already live. The brief was to take it from what it was - a fragmented AI generation tool with duplicate navigation, no production pipeline, and no coherent architecture - and turn it into a real product that a storyteller could actually work inside.
What I found in the competitive landscape was a repeated pattern: many AI generation tools place the hardest task first, asking users to start with a prompt before they have a clear mental model of what the system needs. For an experienced creative director, that can be a workable starting point. For a drama writer with a 12-episode concept and no production background, it can become a barrier before creation even begins.
I was the only designer. Four months. Daily sessions directly with the CEO and CTO. No PM. No brief document. Just a vision, a shared screen, and sessions where I designed, questioned, pushed back on my own decisions, and rebuilt when the reasoning demanded it.
The CEO came with a product already live. The brief was to take it from what it was - a fragmented AI generation tool with duplicate navigation, no production pipeline, and no coherent architecture - and turn it into a real product that a storyteller could actually work inside.
What I found in the competitive landscape was a repeated pattern: many AI generation tools place the hardest task first, asking users to start with a prompt before they have a clear mental model of what the system needs. For an experienced creative director, that can be a workable starting point. For a drama writer with a 12-episode concept and no production background, it can become a barrier before creation even begins.
I was the only designer. Four months. Daily sessions directly with the CEO and CTO. No PM. No brief document. Just a vision, a shared screen, and sessions where I designed, questioned, pushed back on my own decisions, and rebuilt when the reasoning demanded it.
No Memory
The AI does not know who your characters are. Every generation starts from zero. The same face looks different in every scene. The lead character in scene 3 is a different person from scene 1. The setting changes colour between shots. The atmosphere that defines the story's emotional register disappears the moment the user generates something new.
The AI does not know who your characters are. Every generation starts from zero. The same face looks different in every scene. The lead character in scene 3 is a different person from scene 1. The setting changes colour between shots. The atmosphere that defines the story's emotional register disappears the moment the user generates something new.
Result: Character drift. Stories fall apart visually before they are finished.
Single Output
Generate one clip. Download it. That is where the product ends. No pipeline connecting outputs into a coherent story. No assembly environment. No continuity between shots. No way to build what you generated into something larger. The user produces assets with no destination.
Generate one clip. Download it. That is where the product ends. No pipeline connecting outputs into a coherent story. No assembly environment. No continuity between shots. No way to build what you generated into something larger. The user produces assets with no destination.
Result: A folder of disconnected clips that never become a film.
Intimidating UX
Technical tools can alienate everyday creators when the starting point assumes production or prompting knowledge. A prompt box with no structure can become a wall for someone who simply has a story in their head, while simplified tools can feel too shallow for professionals who want direct control over camera, lens, lighting, and other production parameters. The design challenge was serving both levels of expertise without forcing either user into the other's workflow.
Technical tools can alienate everyday creators when the starting point assumes production or prompting knowledge. A prompt box with no structure can become a wall for someone who simply has a story in their head, while simplified tools can feel too shallow for professionals who want direct control over camera, lens, lighting, and other production parameters. The design challenge was serving both levels of expertise without forcing either user into the other's workflow.
Result: Power or simplicity. Never both. Half the potential audience is always in the wrong tool.
The CEO came with a product already live. The brief was to take it from what it was - a fragmented AI generation tool with duplicate navigation, no production pipeline, and no coherent architecture - and turn it into a real product that a storyteller could actually work inside.
What I found in the competitive landscape was a repeated pattern: many AI generation tools place the hardest task first, asking users to start with a prompt before they have a clear mental model of what the system needs. For an experienced creative director, that can be a workable starting point. For a drama writer with a 12-episode concept and no production background, it can become a barrier before creation even begins.
I was the only designer. Four months. Daily sessions directly with the CEO and CTO. No PM. No brief document. Just a vision, a shared screen, and sessions where I designed, questioned, pushed back on my own decisions, and rebuilt when the reasoning demanded it.
AI Generation Tools: The Three Failures
Across the platforms I reviewed, three recurring product problems shaped the redesign.
AI Generation Tools: The Three Failures
Across the platforms I reviewed, three recurring product problems shaped the redesign.
AI Generation Tools: The Three Failures
Across the platforms I reviewed, three recurring product problems shaped the redesign.
The Problem
A product duplicating itself and serving nobody specifically
A product duplicating itself and serving nobody specifically
A product duplicating itself and serving nobody specifically
The live product had three separate navigation items - Story to Video, Text to Video, Image to Video - all routing to the same feature. Users were making a choice that did not exist. That was the first symptom. The deeper problem was structural.
The live product had three separate navigation items - Story to Video, Text to Video, Image to Video - all routing to the same feature. Users were making a choice that did not exist. That was the first symptom. The deeper problem was structural.
Problem
Problem
What it caused
What it caused
Three navigation items, one destination
Three navigation items, one destination
Users could not understand what was different. Trust dropped before anything generated.
Users could not understand what was different. Trust dropped before anything generated.
No production continuity between outputs
No production continuity between outputs
No production continuity between outputs
Every output was disconnected from every other. A folder of clips. Never a film.
Every output was disconnected from every other. A folder of clips. Never a film.
No character memory
No character memory
No character memory
The same character looked different in every generation. Stories fell apart visually.
The same character looked different in every generation. Stories fell apart visually.
Prompt box as the only entry point
Prompt box as the only entry point
Prompt box as the only entry point
A writer with a full story in their head had no way to bring it in.
A writer with a full story in their head had no way to bring it in.
Two completely different users, one interface
Two completely different users, one interface
Two completely different users, one interface
Too technical for everyday creators. Too shallow for professionals. Built for nobody.
Too technical for everyday creators. Too shallow for professionals. Built for nobody.
The live product had three separate navigation items - Story to Video, Text to Video, Image to Video - all routing to the same feature. Users were making a choice that did not exist. That was the first symptom. The deeper problem was structural.
My Role
My Role
Sole designer. Live product. Two direct stakeholders.
Sole designer. Live product. Two direct stakeholders.
Sole designer. Live product. Two direct stakeholders.
No design system. No prior art. No PM. The product was already live in production - not a concept to audit, but a running platform to rebuild architecturally without interrupting what was already there.
I worked directly with the CEO on product vision and with the CTO on technical feasibility. Decisions went from design straight to the people who could reverse them. Two major decisions were reversed based on those reviews: the Foundation library structure and the canvas-first approach. Both reversals made the product better. The work was better for the friction.
No design system. No prior art. No PM. The product was already live in production - not a concept to audit, but a running platform to rebuild architecturally without interrupting what was already there.
I worked directly with the CEO on product vision and with the CTO on technical feasibility. Decisions went from design straight to the people who could reverse them. Two major decisions were reversed based on those reviews: the Foundation library structure and the canvas-first approach. Both reversals made the product better. The work was better for the friction.
"No PM. No design system. No prior art. Just a CEO, a CTO, and a product that needed to be rebuilt while it was running."
"No PM. No design system. No prior art. Just a CEO, a CTO, and a product that needed to be rebuilt while it was running."
"No PM. No design system. No prior art. Just a CEO, a CTO, and a product that needed to be rebuilt while it was running."
I was the sole designer on this product from V1 audit through complete V2 redesign. No team. Research, information architecture, every UX flow, UI decisions, component direction, and developer handoff were mine.
The CEO's brief was direct: no loopholes, no dead ends, no dropped use cases.
Process
What I Actually Did
What I Actually Did
What I Actually Did
Weeks 1-2
Weeks 1-2
Weeks 1-2
Mapped every existing screen and navigation path. Found 3 redundant entry points routing to the same feature. Ran a heuristic audit against 6 competitor products using the same non-technical user filter.
Mapped every existing screen and navigation path. Found 3 redundant entry points routing to the same feature. Ran a heuristic audit against 6 competitor products using the same non-technical user filter.
Mapped every existing screen and navigation path. Found 3 redundant entry points routing to the same feature. Ran a heuristic audit against 6 competitor products using the same non-technical user filter.
Weeks 3-4
Weeks 3-4
Weeks 3-4
Presented category pushback to CEO. Proposed expanding from AI video generator to every story format. CEO aligned on the spot. Wrote the competitive analysis table - Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, Synthesia.
Presented category pushback to CEO. Proposed expanding from AI video generator to every story format. CEO aligned on the spot. Wrote the competitive analysis table - Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, Synthesia.
Presented category pushback to CEO. Proposed expanding from AI video generator to every story format. CEO aligned on the spot. Wrote the competitive analysis table - Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, Synthesia.
Weeks 5-6
Weeks 5-6
Weeks 5-6
Designed Foundation restructure. CEO question mid-review: "Then if there is Foundation why is there Projects too? Isn't it the duplication?" Reversed the library structure on the spot. Foundation moved inside the project.
Designed Foundation restructure. CEO question mid-review: "Then if there is Foundation why is there Projects too? Isn't it the duplication?" Reversed the library structure on the spot. Foundation moved inside the project.
Designed Foundation restructure. CEO question mid-review: "Then if there is Foundation why is there Projects too? Isn't it the duplication?" Reversed the library structure on the spot. Foundation moved inside the project.
Weeks 7-8
Weeks 7-8
Weeks 7-8
Designed three related but distinct systems: Visual Anchor, Character Profiling, and Voice Anchor. Building on Foundation as the shared project context. Diagnosed that text descriptions are interpretations rather than fixed references, so reducing character drift required both persistent story context and a stronger identity-locking system.
Weeks 9-10
Weeks 9-10
Weeks 9-10
Designed Creator's Sketchbook and brand profile form. Identified the anonymous creator gap through Inspiration page research. Reduced brand form from 20 fields to 5 after auditing which fields the AI actually needed.
Designed Creator's Sketchbook and brand profile form. Identified the anonymous creator gap through Inspiration page research. Reduced brand form from 20 fields to 5 after auditing which fields the AI actually needed.
Designed Creator's Sketchbook and brand profile form. Identified the anonymous creator gap through Inspiration page research. Reduced brand form from 20 fields to 5 after auditing which fields the AI actually needed.
Weeks 11-12
Weeks 11-12
Weeks 11-12
Designed V1 interaction patterns and extended into Studio Mode. V1 shipped the spatial canvas as the environment. Agreed with CEO to defer advanced gesture layer until users had the mental model.
Designed V1 interaction patterns and extended into Studio Mode. V1 shipped the spatial canvas as the environment. Agreed with CEO to defer advanced gesture layer until users had the mental model.
Designed V1 interaction patterns and extended into Studio Mode. V1 shipped the spatial canvas as the environment. Agreed with CEO to defer advanced gesture layer until users had the mental model.
Weeks 13-16
Weeks 13-16
Weeks 13-16
Full design system, component library, and handoff documentation. Semantic colour system - orange for Projects, pink for Brand, green for complete, red for danger. 8 structured handoff documents covering every module, each with wireframes, states, and acceptance criteria.
Full design system, component library, and handoff documentation. Semantic colour system - orange for Projects, pink for Brand, green for complete, red for danger. 8 structured handoff documents covering every module, each with wireframes, states, and acceptance criteria.
Full design system, component library, and handoff documentation. Semantic colour system - orange for Projects, pink for Brand, green for complete, red for danger. 8 structured handoff documents covering every module, each with wireframes, states, and acceptance criteria.
Inspiration page wireframe
Video generation flow
Cast and character flow
Portrait creation flow (Cast flow 2)
Foundation / Knowledge Base architecture
Research and Discovery
Research and Discovery
Four users who could not exist in the same product - until they had to
Four users who could not exist in the same product - until they had to
Four users who could not exist in the same product - until they had to
The user base for Timeless Tales spans a wider range than almost any product I have designed for. Unlike most platforms, I could not choose one audience and ignore the others. All four had to be served from the same interface without any of them feeling like they were in the wrong tool.
On one end: the everyday storyteller. A drama writer with a 12-episode concept. A first-time creator with a manga idea sketched in a notebook. No technical knowledge. No film vocabulary. Just a story and the determination to produce it.
On the other end: the professional filmmaker. Someone who knows what a RED V-Raptor anamorphic shot looks like, what f/2.0 means for depth of field, what colour science choices separate ARRI from Sony. Someone who will not tolerate a prompt box.
The user base for Timeless Tales spans a wider range than almost any product I have designed for. Unlike most platforms, I could not choose one audience and ignore the others. All four had to be served from the same interface without any of them feeling like they were in the wrong tool.
On one end: the everyday storyteller. A drama writer with a 12-episode concept. A first-time creator with a manga idea sketched in a notebook. No technical knowledge. No film vocabulary. Just a story and the determination to produce it.
On the other end: the professional filmmaker. Someone who knows what a RED V-Raptor anamorphic shot looks like, what f/2.0 means for depth of field, what colour science choices separate ARRI from Sony. Someone who will not tolerate a prompt box.
Core finding: "These two users think in completely different languages. The everyday creator thinks in feeling - warm, intimate, the kind of light you get in the hour before sunset. The professional thinks in parameters - 85mm, f/1.8, warm tungsten grade, shallow depth of field. They are describing the same shot. They just do not know it."
Core finding: "These two users think in completely different languages. The everyday creator thinks in feeling - warm, intimate, the kind of light you get in the hour before sunset. The professional thinks in parameters - 85mm, f/1.8, warm tungsten grade, shallow depth of field. They are describing the same shot. They just do not know it."
Core finding: "These two users think in completely different languages. The everyday creator thinks in feeling - warm, intimate, the kind of light you get in the hour before sunset. The professional thinks in parameters - 85mm, f/1.8, warm tungsten grade, shallow depth of field. They are describing the same shot. They just do not know it."
The user base for Timeless Tales spans a wider range than almost any product I have designed for. Unlike most platforms, I could not choose one audience and ignore the others. All four had to be served from the same interface without any of them feeling like they were in the wrong tool.
On one end: the everyday storyteller. A drama writer with a 12-episode concept. A first-time creator with a manga idea sketched in a notebook. No technical knowledge. No film vocabulary. Just a story and the determination to produce it.
On the other end: the professional filmmaker. Someone who knows what a RED V-Raptor anamorphic shot looks like, what f/2.0 means for depth of field, what colour science choices separate ARRI from Sony. Someone who will not tolerate a prompt box.
User
User
Everyday storyteller
Everyday storyteller
Professional filmmaker
Professional filmmaker
Brand founder and marketer
Brand founder and marketer
Anonymous creator
Anonymous creator
What they bring
What they bring
A story, characters, a world. No production experience.
A story, characters, a world. No production experience.
Deep craft knowledge, strong opinions, high standards.
Deep craft knowledge, strong opinions, high standards.
A brand, a product, a customer problem.
A brand, a product, a customer problem.
Creative work they cannot attach their real name to.
Work they cannot attach their real name to.
Creative work they cannot attach their real name to.
What the design had to do
What the design had to do
Must produce professional output without requiring any technical knowledge.
Professional output without requiring technical knowledge.
Must produce professional output without requiring any technical knowledge.
Must offer real technical camera control without making the other users feel it.
Real technical camera control without the other users feeling it.
Must offer real technical camera control without making the other users feel it.
Must produce conversion-ready content from simple inputs.
Conversion-ready content from simple inputs.
Must produce conversion-ready content from simple inputs.
Must give them a public creative identity without requiring personal exposure.
A public creative identity without requiring personal exposure.
Must give them a public creative identity without requiring personal exposure.
In the competitive set I reviewed, Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, and Synthesia were primarily organised around a dominant output type or creation workflow. I did not find the same combination of project-level story context, character profiling, visual anchoring, and cross-format assembly that I was designing for Timeless Tales. That gap helped define the product architecture: the opportunity was not simply another generation tool, but a system that could carry the same story context across multiple creative outputs.
In the competitive set I reviewed, Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, and Synthesia were primarily organised around a dominant output type or creation workflow. I did not find the same combination of project-level story context, character profiling, visual anchoring, and cross-format assembly that I was designing for Timeless Tales. That gap helped define the product architecture: the opportunity was not simply another generation tool, but a system that could carry the same story context across multiple creative outputs.
Competitor
Competitor
Runway ML
Runway ML
Pika Labs
Pika Labs
Leonardo AI
Leonardo AI
ImagineArt
ImagineArt
Midjourney
Midjourney
HeyGen
HeyGen
Synthesia
Synthesia
Timeless Tales
Timeless
Tales
Type
Type
AI Video
AI Video
AI Video
AI Video
AI Image
AI Image
AI Image
AI Image
AI Image
AI Image
Avatar Video
Avatar Video
Enterprise Video
Enterprise Video
AI Creative Platform
AI Creative Platform
Where They're Strong
Where They're Strong
Professional quality, fine motion control
Professional quality, fine motion control
Fast turnaround, accessible to casual creators
Fast turnaround, accessible to casual creators
Detailed style control, consistent visual language within a session
Detailed style control, consistent visual language within a session
Speed to first image, growing model library
Speed to first image, growing model library
Highest aesthetic quality, strong community prompt library
Highest aesthetic quality, strong community prompt library
Instant presenter videos, 80+ languages
Instant presenter videos, 80+ languages
Enterprise polish, compliance-readiness
Enterprise polish, compliance-readiness
Cross-format character consistency - visual, voice, and story identity locked across every output
Cross-format character consistency - visual, voice, and story identity locked across every output
What They Trade Away
What They Trade Away
Story context disappears at render - every clip starts fresh
Story context disappears at render - every clip starts fresh
Format lock-in - no path to brand assets or publishing workflow
Format lock-in - no path to brand assets or publishing workflow
Still images stay still - no voice, no motion, no cross-format continuity
Still images stay still - no voice, no motion, no cross-format continuity
Image-only - every character looks different every time you generate
Image-only - every character looks different every time you generate
No memory, no project continuity, no character consistency
No memory, no project continuity, no character consistency
Presenter is the product - characters exist only per video, not across a narrative
Presenter is the product - characters exist only per video, not across a narrative
No creative storytelling, no character identity persistence
No creative storytelling, no character identity persistence
Newer product, smaller community, fewer third-party integrations
Newer product, smaller community, fewer third-party integrations
In the competitive set I reviewed, Runway, Pika, Leonardo, ImagineArt, Midjourney, HeyGen, and Synthesia were primarily organised around a dominant output type or creation workflow. I did not find the same combination of project-level story context, character profiling, visual anchoring, and cross-format assembly that I was designing for Timeless Tales. That gap helped define the product architecture: the opportunity was not simply another generation tool, but a system that could carry the same story context across multiple creative outputs.
Within this competitive set, Timeless Tales was positioned differently: story context comes before generation, character identity persists through shared references, and outputs are designed to belong to a larger narrative workflow rather than exist as isolated assets.
Within this competitive set, Timeless Tales was positioned differently: story context comes before generation, character identity persists through shared references, and outputs are designed to belong to a larger narrative workflow rather than exist as isolated assets.
Within this competitive set, Timeless Tales was positioned differently: story context comes before generation, character identity persists through shared references, and outputs are designed to belong to a larger narrative workflow rather than exist as isolated assets.
What They Trade Away
Story context disappears at render - every clip starts fresh
Format lock-in - no path to brand assets or publishing workflow
Still images stay still - no voice, no motion, no cross-format continuity
Image-only - every character looks different every time you generate
No memory, no project continuity, no character consistency
Presenter is the product - characters exist only per video, not across a narrative
No creative storytelling, no character identity persistence
Newer product, smaller community, fewer third-party integrations
Competitor
Runway ML
Pika Labs
Leonardo AI
ImagineArt
Midjourney
HeyGen
Synthesia
Timeless Tales
Type
AI Video
AI Video
AI Image
AI Image
AI Image
Avatar Video Generation
Enterprise Video
AI Creative Platform
Where They're Strong
Professional video quality, fine motion control, Gen-3 model
Fast turnaround on short clips, accessible to casual creators
Detailed style control, consistent visual language within a session, asset generation pipeline
Speed to first image, lower cost per generation, growing model library
Highest artistic aesthetic quality, strong community prompt library, style versatility
Instant presenter videos, 80+ languages, fast dubbing workflow
Enterprise polish, brand safety, compliance-readiness, structured template library
Cross-format character consistency - visual, voice, and story identity stay locked across every output
Key Design Decisions
Key Design Decisions
The decisions that changed the product
The decisions that changed the product
The decisions that changed the product
Decision 01 — Most Defining
Decision 01 — Most Defining
Pushing back on the category: a story is not only a video
Pushing back on the category: a story is not only a video
Pushing back on the category: a story is not only a video
Most products I reviewed were built around a dominant output format. Following the same pattern would have limited Timeless Tales to another generation tool.
Most products I reviewed were built around a dominant output format. Following the same pattern would have limited Timeless Tales to another generation tool.
Most products I reviewed were built around a dominant output format. Following the same pattern would have limited Timeless Tales to another generation tool.
In the competitive set I reviewed, most products were organised around a dominant output format or creation workflow. If Timeless Tales followed the same framing, it would compete primarily as another AI video or generation tool rather than as a broader storytelling system.
I pushed back on that framing. A manga artist tells a story through panels. A novelist uses chapters. A drama writer may work through a screenplay. An educator may use illustrated explanations. A brand may tell a story through campaign assets. Video was one storytelling format, not the definition of storytelling itself.
I proposed expanding the product across cinematic film, manga, anime, novels, storyboards, educational content, image series, and brand campaigns. These were not separate products. They became tale types inside one shared architecture, using the same project context and production structure while producing different forms of output.
This shifted Timeless Tales from a video-generator framing toward a cross-format storytelling platform. Within the competitive set I reviewed, I did not find the same combination of shared story context, character systems, and multiple assembled output formats.
In the competitive set I reviewed, most products were organised around a dominant output format or creation workflow. If Timeless Tales followed the same framing, it would compete primarily as another AI video or generation tool rather than as a broader storytelling system.
I pushed back on that framing. A manga artist tells a story through panels. A novelist uses chapters. A drama writer may work through a screenplay. An educator may use illustrated explanations. A brand may tell a story through campaign assets. Video was one storytelling format, not the definition of storytelling itself.
I proposed expanding the product across cinematic film, manga, anime, novels, storyboards, educational content, image series, and brand campaigns. These were not separate products. They became tale types inside one shared architecture, using the same project context and production structure while producing different forms of output.
This shifted Timeless Tales from a video-generator framing toward a cross-format storytelling platform. Within the competitive set I reviewed, I did not find the same combination of shared story context, character systems, and multiple assembled output formats.
In the competitive set I reviewed, most products were organised around a dominant output format or creation workflow. If Timeless Tales followed the same framing, it would compete primarily as another AI video or generation tool rather than as a broader storytelling system.
I pushed back on that framing. A manga artist tells a story through panels. A novelist uses chapters. A drama writer may work through a screenplay. An educator may use illustrated explanations. A brand may tell a story through campaign assets. Video was one storytelling format, not the definition of storytelling itself.
I proposed expanding the product across cinematic film, manga, anime, novels, storyboards, educational content, image series, and brand campaigns. These were not separate products. They became tale types inside one shared architecture, using the same project context and production structure while producing different forms of output.
This shifted Timeless Tales from a video-generator framing toward a cross-format storytelling platform. Within the competitive set I reviewed, I did not find the same combination of shared story context, character systems, and multiple assembled output formats.
“The way of conveying stories is not only video. That observation changed how the entire product was structured.”
“The way of conveying stories is not only video. That observation changed how the entire product was structured.”
“The way of conveying stories is not only video. That observation changed how the entire product was structured.”
Decision 02 - Most Foundational
Decision 02 - Most Foundational
Foundation: giving the AI a world to generate from, not just a prompt
Foundation: giving the AI a world to generate from, not just a prompt
Foundation: giving the AI a world to generate from, not just a prompt
Without persistent project context, each generation risks starting from a different interpretation of the same story.
Without persistent project context, each generation risks starting from a different interpretation of the same story.
Many generation workflows treat each new output as a fresh request unless project context or references are deliberately carried forward. That makes continuity difficult across scenes, characters, locations, and atmosphere, especially in longer-form storytelling.
I designed Foundation as the knowledge base beneath every generation in a project. Before generation, the user defines the characters, world, storyline, geographic location, era, cultural context, visual atmosphere, and base narrative. Instead of repeatedly restating that information, every generation begins from the same project context.
Foundation became the first layer of the consistency architecture. It reduces drift by keeping every generation grounded in the same story world, but context alone cannot lock a specific face or visual identity. Visual Anchor provides the approved visual reference, while Character Profiling provides structured identity context. Together, Foundation and the character-locking systems were designed to reduce character drift across repeated generations.
Foundation is therefore both the creative brief of the project and the context layer beneath the consistency system. Visual Anchor, Character Profiling, and Voice Anchor build on that shared foundation rather than asking the user to recreate the same context every time.
Many generation workflows treat each new output as a fresh request unless project context or references are deliberately carried forward. That makes continuity difficult across scenes, characters, locations, and atmosphere, especially in longer-form storytelling.
I designed Foundation as the knowledge base beneath every generation in a project. Before generation, the user defines the characters, world, storyline, geographic location, era, cultural context, visual atmosphere, and base narrative. Instead of repeatedly restating that information, every generation begins from the same project context.
Foundation became the first layer of the consistency architecture. It reduces drift by keeping every generation grounded in the same story world, but context alone cannot lock a specific face or visual identity. Visual Anchor provides the approved visual reference, while Character Profiling provides structured identity context. Together, Foundation and the character-locking systems were designed to reduce character drift across repeated generations.
Foundation is therefore both the creative brief of the project and the context layer beneath the consistency system. Visual Anchor, Character Profiling, and Voice Anchor build on that shared foundation rather than asking the user to recreate the same context every time.
Many generation workflows treat each new output as a fresh request unless project context or references are deliberately carried forward. That makes continuity difficult across scenes, characters, locations, and atmosphere, especially in longer-form storytelling.
I designed Foundation as the knowledge base beneath every generation in a project. Before generation, the user defines the characters, world, storyline, geographic location, era, cultural context, visual atmosphere, and base narrative. Instead of repeatedly restating that information, every generation begins from the same project context.
Foundation became the first layer of the consistency architecture. It reduces drift by keeping every generation grounded in the same story world, but context alone cannot lock a specific face or visual identity. Visual Anchor provides the approved visual reference, while Character Profiling provides structured identity context. Together, Foundation and the character-locking systems were designed to reduce character drift across repeated generations.
Foundation is therefore both the creative brief of the project and the context layer beneath the consistency system. Visual Anchor, Character Profiling, and Voice Anchor build on that shared foundation rather than asking the user to recreate the same context every time.
“Context produces consistency. Every other decision in this product exists because Foundation was designed to be read first.”
“Context produces consistency. Every other decision in this product exists because Foundation was designed to be read first.”
"Context produces consistency. Every other decision in this product exists because Foundation was designed to be read first."
Decision 03 - Most Technically Consequential
Decision 03 - Most Technically Consequential
Decision 03 - Most Technically Consequential
Visual Anchor: the AI generates options, the user selects the truth
Visual Anchor: the AI generates options, the user selects the truth
Visual Anchor: the AI generates options, the user selects the truth
In real life, seven people can share one face description. A text requirement fulfils multiple interpretations. The system needed to generate options and let the user identify which one is their character.
In real life, seven people can share one face description. A text requirement fulfils multiple interpretations. The system needed to generate options and let the user identify which one is their character.
In real life, seven people can share one face description. A text requirement fulfils multiple interpretations. The system needed to generate options and let the user identify which one is their character.
The natural approach to character consistency is better prompting. Describe the character more precisely. I challenged this assumption before it became a design, because it is wrong.
A text description of a face matches thousands of different faces that all technically satisfy it. "26 year old woman, warm olive skin, dark wavy hair, sharp expressive eyes" is accurate for many different people. Text cannot lock a specific identity. Text is interpretation. An anchor is a reference.
The Visual Anchor system works in three stages. The AI generates four portrait variations per run from the character's structured profile - all technically matching the description, each a different person. The user reviews them and selects the one that matches their vision of the character. That approved image becomes the locked visual anchor - the reference the system uses for every future generation involving this character: teasers, trailers, posters, episode shots, character introductions. The AI references the approved face. It does not reinterpret the description.
The consistency check runs after batch generation. Every output is compared against the approved anchor. Outputs that drift are flagged with a specific callout - what changed, not just that something changed. The user regenerates only the flagged outputs. Everything consistent is untouched.
The natural approach to character consistency is better prompting. Describe the character more precisely. I challenged this assumption before it became a design, because it is wrong.
A text description of a face matches thousands of different faces that all technically satisfy it. "26 year old woman, warm olive skin, dark wavy hair, sharp expressive eyes" is accurate for many different people. Text cannot lock a specific identity. Text is interpretation. An anchor is a reference.
The Visual Anchor system works in three stages. The AI generates four portrait variations per run from the character's structured profile - all technically matching the description, each a different person. The user reviews them and selects the one that matches their vision of the character. That approved image becomes the locked visual anchor - the reference the system uses for every future generation involving this character: teasers, trailers, posters, episode shots, character introductions. The AI references the approved face. It does not reinterpret the description.
The consistency check runs after batch generation. Every output is compared against the approved anchor. Outputs that drift are flagged with a specific callout - what changed, not just that something changed. The user regenerates only the flagged outputs. Everything consistent is untouched.
The natural approach to character consistency is better prompting. Describe the character more precisely. I challenged this assumption before it became a design, because it is wrong.
A text description of a face matches thousands of different faces that all technically satisfy it. "26 year old woman, warm olive skin, dark wavy hair, sharp expressive eyes" is accurate for many different people. Text cannot lock a specific identity. Text is interpretation. An anchor is a reference.
The Visual Anchor system works in three stages. The AI generates four portrait variations per run from the character's structured profile - all technically matching the description, each a different person. The user reviews them and selects the one that matches their vision of the character. That approved image becomes the locked visual anchor - the reference the system uses for every future generation involving this character: teasers, trailers, posters, episode shots, character introductions. The AI references the approved face. It does not reinterpret the description.
The consistency check runs after batch generation. Every output is compared against the approved anchor. Outputs that drift are flagged with a specific callout - what changed, not just that something changed. The user regenerates only the flagged outputs. Everything consistent is untouched.
"The AI references an approved face. It does not re-interpret a text description. That is the difference between a constraint and a suggestion."
"The AI references an approved face. It does not re-interpret a text description. That is the difference between a constraint and a suggestion."
"The AI references an approved face. It does not reinterpret a text description. That is the difference between a constraint and a suggestion."
Decision 04 - The Language Gap
Decision 04 - The Language Gap
Character Profiling: users communicate in story facts, not prompting language
Knowing your character and knowing how to prompt an AI are completely different skills. Every generation tool I reviewed required users to translate character knowledge into prompt language.
Knowing your character and knowing how to prompt an AI are completely different skills. Every generation tool I reviewed required users to translate character knowledge into prompt language.
Knowing your character and knowing how to prompt an AI are completely different skills. Every generation tool I reviewed required users to translate character knowledge into prompt language.
Across the generation tools I reviewed, character definition happened almost entirely through free-form text prompts, which made prompting skill a requirement for creating a character.
The real problem is that free-form prompts force the user into a language that is not natural to them. "Sharp expressive eyes, olive skin, Sardinian heritage, military bearing, warm but authoritative" is how an AI needs to be prompted. It is not how a person thinks about a character they created. A person thinks: she is a Carabinieri marshal from Sardinia, she investigates art theft, her ancestors were weavers, she is hardened by tactical training but warm with the people she protects. These are not prompts. They are story facts.
The Character Profiling system bridges this gap with structured identity fields that mirror how a person already knows their character, using observational and relational terms instead of prompting terms:
These are not prompting fields. They are identity fields. The user communicates in what they already know, and the AI translates that into the context it needs, so the user never has to construct a prompt themselves.
Across the generation tools I reviewed, character definition happened almost entirely through free-form text prompts, which made prompting skill a requirement for creating a character.
The real problem is that free-form prompts force the user into a language that is not natural to them. "Sharp expressive eyes, olive skin, Sardinian heritage, military bearing, warm but authoritative" is how an AI needs to be prompted. It is not how a person thinks about a character they created. A person thinks: she is a Carabinieri marshal from Sardinia, she investigates art theft, her ancestors were weavers, she is hardened by tactical training but warm with the people she protects. These are not prompts. They are story facts.
The Character Profiling system bridges this gap with structured identity fields that mirror how a person already knows their character, using observational and relational terms instead of prompting terms:
Across the generation tools I reviewed, character definition happened almost entirely through free-form text prompts, which made prompting skill a requirement for creating a character.
The real problem is that free-form prompts force the user into a language that is not natural to them. "Sharp expressive eyes, olive skin, Sardinian heritage, military bearing, warm but authoritative" is how an AI needs to be prompted. It is not how a person thinks about a character they created. A person thinks: she is a Carabinieri marshal from Sardinia, she investigates art theft, her ancestors were weavers, she is hardened by tactical training but warm with the people she protects. These are not prompts. They are story facts.
The Character Profiling system bridges this gap with structured identity fields that mirror how a person already knows their character, using observational and relational terms instead of prompting terms:
Name, Age, Heritage:
Name, Age, Heritage:
Who the character is in the world
Who the character is in the world
Who the character is in the world
Physical Description:
Physical Description:
Physical Description:
What they look like, using observational rather than prompting terms
What they look like, using observational rather than prompting terms
What they look like, using observational rather than prompting terms
Occupation:
Occupation:
Occupation:
What they do and how that shapes them
What they do and how that shapes them
What they do and how that shapes them
Personality Axis
Personality Axis
Personality Axis
Two poles that define behavioural range, such as resilient and sweet or introverted and spontaneous
Two poles that define behavioural range, such as resilient and sweet or introverted and spontaneous
Two poles that define behavioural range, such as resilient and sweet or introverted and spontaneous
Style Paradigm
Style Paradigm
Style Paradigm
How they present themselves in different contexts, such as uniform versus off-duty
How they present themselves in different contexts, such as uniform versus off-duty
How they present themselves in different contexts, such as uniform versus off-duty
Story Context
Story Context
Where they come from, what shaped them, and what drives them
Where they come from, what shaped them, and what drives them
Where they come from, what shaped them, and what drives them
These are not prompting fields. They are identity fields. The user communicates in what they already know, and the AI translates that into the context it needs, so the user never has to construct a prompt themselves.
These are not prompting fields. They are identity fields. The user communicates in what they already know, and the AI translates that into the context it needs, so the user never has to construct a prompt themselves.
"The user thinks in story facts. The AI works in context. The profiling system is the translation layer between them."
"The user thinks in story facts. The AI works in context. The profiling system is the translation layer between them."
"The user thinks in story facts. The AI works in context. The profiling system is the translation layer between them."
Decision 05 - The Anonymous Creator Gap
Decision 05 - The Anonymous Creator Gap
Decision 05 - The Anonymous Creator Gap
Anonymous creators deserve an audience without requiring a face
Some creators want an audience for their work without making their real identity part of the product.
Some creators want an audience for their work without making their real identity part of the product.
Some creators want an audience for their work without making their real identity part of the product.
This need surfaced during a discussion about the Inspiration page. Manga artists, novelists, and drama writers often build a public body of work while keeping their personal identity private, and I did not find a clear path for them in the competitive products I reviewed.
I designed Creator's Sketchbook as both a content category and a Foundation feature. Creators define an AI persona of themselves, with its own voice, aesthetic, delivery style, and personality, and that persona generates process and behind-the-scenes content. The creator's real identity stays private, while the AI persona becomes the bridge between them and the audience that wants to know the person behind the work.
The work and the AI persona carry the creator's public identity: some of the most personal content on the platform can exist without a real face ever being shown. It was a deliberate architectural decision for creators who want connection and visibility without personal exposure.
This need surfaced during a discussion about the Inspiration page. Manga artists, novelists, and drama writers often build a public body of work while keeping their personal identity private, and I did not find a clear path for them in the competitive products I reviewed.
I designed Creator's Sketchbook as both a content category and a Foundation feature. Creators define an AI persona of themselves, with its own voice, aesthetic, delivery style, and personality, and that persona generates process and behind-the-scenes content. The creator's real identity stays private, while the AI persona becomes the bridge between them and the audience that wants to know the person behind the work.
The work and the AI persona carry the creator's public identity: some of the most personal content on the platform can exist without a real face ever being shown. It was a deliberate architectural decision for creators who want connection and visibility without personal exposure.
This need surfaced during a discussion about the Inspiration page. Manga artists, novelists, and drama writers often build a public body of work while keeping their personal identity private, and I did not find a clear path for them in the competitive products I reviewed.
I designed Creator's Sketchbook as both a content category and a Foundation feature. Creators define an AI persona of themselves, with its own voice, aesthetic, delivery style, and personality, and that persona generates process and behind-the-scenes content. The creator's real identity stays private, while the AI persona becomes the bridge between them and the audience that wants to know the person behind the work.
The work and the AI persona carry the creator's public identity: some of the most personal content on the platform can exist without a real face ever being shown. It was a deliberate architectural decision for creators who want connection and visibility without personal exposure.
"The opportunity was not to force anonymous creators to become public personalities. It was to give their work a public identity without requiring personal exposure."
"The opportunity was not to force anonymous creators to become public personalities. It was to give their work a public identity without requiring personal exposure."
"The opportunity was not to force anonymous creators to become public personalities. It was to give their work a public identity without requiring personal exposure."
Decision 06 - The Spatial Canvas as the Foundation
Decision 06 - The Spatial Canvas as the Foundation
Studio Mode: the canvas is not a feature. It is the environment.
Studio Mode: the canvas is not a feature. It is the environment.
Studio Mode: the canvas is not a feature. It is the environment.
Every module lives inside the spatial canvas. Not accessed through it. Inside it. The container is the product.
Every module lives inside the spatial canvas. Not accessed through it. Inside it. The container is the product.
Every module lives inside the spatial canvas. Not accessed through it. Inside it. The container is the product.
Studio Mode in V1 is the spatial canvas - a director-on-a-set interaction model where characters, scenes, script sections, and generated assets all exist simultaneously as draggable objects. The creator drags a character card onto a scene. Releases a storyboard frame into the assembly timeline. Pans between sections of the project. The interaction is spatial and continuous. The way Figma feels to a designer or Final Cut feels to an editor.
The distinction that matters: the canvas is not a separate module users navigate to. It is the environment that every other module lives inside. Characters live on it. Scenes live on it. Script sections live on it. Generated assets live on it. The canvas is always present underneath whatever the user is working on.
Studio Mode in V1 is the spatial canvas - a director-on-a-set interaction model where characters, scenes, script sections, and generated assets all exist simultaneously as draggable objects. The creator drags a character card onto a scene. Releases a storyboard frame into the assembly timeline. Pans between sections of the project. The interaction is spatial and continuous. The way Figma feels to a designer or Final Cut feels to an editor.
The distinction that matters: the canvas is not a separate module users navigate to. It is the environment that every other module lives inside. Characters live on it. Scenes live on it. Script sections live on it. Generated assets live on it. The canvas is always present underneath whatever the user is working on.
Studio Mode in V1 is the spatial canvas - a director-on-a-set interaction model where characters, scenes, script sections, and generated assets all exist simultaneously as draggable objects. The creator drags a character card onto a scene. Releases a storyboard frame into the assembly timeline. Pans between sections of the project. The interaction is spatial and continuous. The way Figma feels to a designer or Final Cut feels to an editor.
The distinction that matters: the canvas is not a separate module users navigate to. It is the environment that every other module lives inside. Characters live on it. Scenes live on it. Script sections live on it. Generated assets live on it. The canvas is always present underneath whatever the user is working on.
"The canvas is not a feature you enter. It is the environment you never leave."
"The canvas is not a feature you enter. It is the environment you never leave."
"The canvas is not a feature you enter. It is the environment you never leave."
Decision 07 - The Two-Language Form
Decision 07 - The Two-Language Form
One form. Two users. Neither knows the other exists.
One form. Two users. Neither knows the other exists.
One form. Two users. Neither knows the other exists.
The professional filmmaker and the everyday creator describe the same shot in completely different languages. The design had to serve both without asking either to learn the other's language.
The professional filmmaker and the everyday creator describe the same shot in completely different languages. The design had to serve both without asking either to learn the other's language.
The professional filmmaker and the everyday creator describe the same shot in completely different languages. The design had to serve both without asking either to learn the other's language.
The professional says 85mm, f/1.8, warm tungsten grade, shallow depth of field. The everyday creator says warm and intimate, the kind of light you get in the hour before sunset, the feeling of a conversation that has been avoided for too long. Both descriptions produce the same shot. Neither user knows the other's language.
This principle applies across the platform, not only in ad creation. Every module is designed with an everyday creator layer and a professional depth layer. The everyday creator gets guided defaults and structured fields. The professional gets technical parameters and override capability. Both get the same output quality. Neither knows how the other described what they wanted.
In the UGC ad creation form specifically: the form has three parts. General Ad Details - campaign basics everyone fills. Ad Script and Proof of Quality - the content layer everyone fills. Video Direction - the technical layer. The everyday creator fills parts 1 and 2. The AI fills Video Direction with sensible defaults. The professional fills all three with camera body, lens type, aperture, colour science, aspect ratio. The AI fills whatever is left empty from context. Same output. Neither user compromised.
The professional says 85mm, f/1.8, warm tungsten grade, shallow depth of field. The everyday creator says warm and intimate, the kind of light you get in the hour before sunset, the feeling of a conversation that has been avoided for too long. Both descriptions produce the same shot. Neither user knows the other's language.
This principle applies across the platform, not only in ad creation. Every module is designed with an everyday creator layer and a professional depth layer. The everyday creator gets guided defaults and structured fields. The professional gets technical parameters and override capability. Both get the same output quality. Neither knows how the other described what they wanted.
In the UGC ad creation form specifically: the form has three parts. General Ad Details - campaign basics everyone fills. Ad Script and Proof of Quality - the content layer everyone fills. Video Direction - the technical layer. The everyday creator fills parts 1 and 2. The AI fills Video Direction with sensible defaults. The professional fills all three with camera body, lens type, aperture, colour science, aspect ratio. The AI fills whatever is left empty from context. Same output. Neither user compromised.
The professional says 85mm, f/1.8, warm tungsten grade, shallow depth of field. The everyday creator says warm and intimate, the kind of light you get in the hour before sunset, the feeling of a conversation that has been avoided for too long. Both descriptions produce the same shot. Neither user knows the other's language.
This principle applies across the platform, not only in ad creation. Every module is designed with an everyday creator layer and a professional depth layer. The everyday creator gets guided defaults and structured fields. The professional gets technical parameters and override capability. Both get the same output quality. Neither knows how the other described what they wanted.
In the UGC ad creation form specifically: the form has three parts. General Ad Details - campaign basics everyone fills. Ad Script and Proof of Quality - the content layer everyone fills. Video Direction - the technical layer. The everyday creator fills parts 1 and 2. The AI fills Video Direction with sensible defaults. The professional fills all three with camera body, lens type, aperture, colour science, aspect ratio. The AI fills whatever is left empty from context. Same output. Neither user compromised.
"The everyday creator fills warm evening light. The professional fills 85mm, f/1.8, tungsten. They get the same output. Neither user knows the other exists."
"The everyday creator fills warm evening light. The professional fills 85mm, f/1.8, tungsten. They get the same output. Neither user knows the other exists."
"The everyday creator fills warm evening light. The professional fills 85mm, f/1.8, tungsten. They get the same output. Neither user knows the other exists."
The Design Principles I Applied
The Design Principles I Applied
Every principle backed by a specific decision
Every principle backed by a specific decision
Every principle backed by a specific decision
Category definition constrains the entire product architecture
→ Applied in Decision 01
The moment the product stopped assuming stories are videos, every module, every tale type, every assembled output changed. Positioning precedes architecture.
Context as a precondition for consistency
→ Applied in Decision 02
Foundation defines the world before generation begins. Context must exist before consistency can be enforced. Visual Anchor, Voice Anchor, and Character Profiling all operate on this premise.
Recognition rather than recall
→ Applied in Decision 04
Character Profiling translates story facts to AI context. Users work with structured fields they recognise rather than free-form prompts they must recall. The AI does not have to interpret.
Reference over interpretation
→ Applied in Decision 03
The AI generates options. The user identifies the truth. That approved reference - not the text description - becomes the constraint for every future generation.
Transparent complexity layering
→ Applied in Decision 07
The everyday creator and the professional filmmaker use the same form and get the same output. Complexity is invisible to the user who does not need it and accessible to the one who does.
Identity without exposure
→ Applied in Decision 05
Creator's Sketchbook gives creators a public creative identity without requiring their real face or personal identity to become part of the experience.
Environment-first architecture
→ Applied in Decision 06
The canvas is not a feature the user enters and exits. It is the persistent environment every other module lives inside. The container is the product.
Category definition constrains the entire product architecture
→ Applied in Decision 01
The moment the product stopped assuming stories are videos, every module, every tale type, every assembled output changed. Positioning precedes architecture.
Context as a precondition for consistency
→ Applied in Decision 02
Foundation defines the world before generation begins. Context must exist before consistency can be enforced. Visual Anchor, Voice Anchor, and Character Profiling all operate on this premise.
Reference over interpretation
→ Applied in Decision 03
The AI generates options. The user identifies the truth. That approved reference - not the text description - becomes the constraint for every future generation.
Recognition rather than recall
→ Applied in Decision 04
Character Profiling translates story facts to AI context. Users work with structured fields they recognise rather than free-form prompts they must recall. The AI does not have to interpret.
Identity without exposure
→ Applied in Decision 05
Creator's Sketchbook gives creators a public creative identity without requiring their real face or personal identity to become part of the experience.
Environment-first architecture
→ Applied in Decision 06
The canvas is not a feature the user enters and exits. It is the persistent environment every other module lives inside. The container is the product.
Transparent complexity layering
→ Applied in Decision 07
The everyday creator and the professional filmmaker use the same form and get the same output. Complexity is invisible to the user who does not need it and accessible to the one who does.
Context as a precondition for consistency
→ Applied in Decision 02
Foundation defines the world before generation begins. Context must exist before consistency can be enforced. Visual Anchor, Voice Anchor, and Character Profiling all operate on this premise.
Category definition constrains the entire product architecture
→ Applied in Decision 01
The moment the product stopped assuming stories are videos, every module, every tale type, every assembled output changed. Positioning precedes architecture.
Reference over interpretation
→ Applied in Decision 03
The AI generates options. The user identifies the truth. That approved reference - not the text description - becomes the constraint for every future generation.
Recognition rather than recall
→ Applied in Decision 04
Character Profiling translates story facts to AI context. Users work with structured fields they recognise rather than free-form prompts they must recall. The AI does not have to interpret.
Transparent complexity layering
→ Applied in Decision 07
The everyday creator and the professional filmmaker use the same form and get the same output. Complexity is invisible to the user who does not need it and accessible to the one who does.
Identity without exposure
→ Applied in Decision 05
Creator's Sketchbook gives identity without requiring a face. Anonymous creation is a legitimate creative need. Every other platform required exposure as the price of connection.
Environment-first architecture
→ Applied in Decision 06
The canvas is not a feature the user enters and exits. It is the persistent environment every other module lives inside. The container is the product.
Outcome
Outcome
What the redesign delivered
What the redesign delivered
What the redesign delivered
Earlier Search Console Growth Checkpoint
Later Search Console Growth Checkpoint
Organic discovery continued growing after the redesign. The earliest Search Console baseline showed 80 impressions and 16 clicks. A later portfolio snapshot reached 40,200 impressions and 313 clicks, and the latest three-month company report recorded 45,300 impressions and 389 organic clicks. This growth was a shared product, content, SEO, engineering, and design outcome rather than a design-only result.
Organic discovery increased by 50,150% in impressions and 1,856% in clicks. The new Tale and Brand architecture created distinct indexable landing pages for each use case. Previously there was one page for everything. The restructure gave search engines separate pages to crawl and surface for queries related to AI storytelling, manga generation, brand campaign creation, and anonymous creative tools - queries the old structure could not serve because it had no dedicated pages for them.
Organic discovery continued growing after the redesign. The earliest Search Console baseline showed 80 impressions and 16 clicks. A later portfolio snapshot reached 40,200 impressions and 313 clicks, and the latest three-month company report recorded 45,300 impressions and 389 organic clicks. This growth was a shared product, content, SEO, engineering, and design outcome rather than a design-only result.
The nature of discovery also changed as search coverage expanded. The latest company report shows Timeless Tales ranking across more than 1,000 distinct queries, supported by a growing library of comparison and use-case pages. As visibility expanded into a much broader and more competitive query set, CTR and average position became less useful as standalone measures of success. I treat the increase in impressions, clicks, indexed pages, and query coverage as evidence of broader organic discovery, not as a direct measure of product adoption.
The nature of discovery also changed as search coverage expanded. The latest company report shows Timeless Tales ranking across more than 1,000 distinct queries, supported by a growing library of comparison and use-case pages. As visibility expanded into a much broader and more competitive query set, CTR and average position became less useful as standalone measures of success. I treat the increase in impressions, clicks, indexed pages, and query coverage as evidence of broader organic discovery, not as a direct measure of product adoption.
Metric
Clicks
Impressions
CTR
Avg. Position
BASELINE SNAPSHOT
16
80
20%
8.1
LATEST VERIFIED
389
45,300
0.9%
20.9
INTERPRETATION
Broader organic discovery
Broader organic discovery
Broader query mix
More competitive query coverage
The nature of discovery also changed as search coverage expanded. The latest company report shows Timeless Tales ranking across more than 1,000 distinct queries, supported by a growing library of comparison and use-case pages. As visibility expanded into a much broader and more competitive query set, CTR and average position became less useful as standalone measures of success. I treat the increase in impressions, clicks, indexed pages, and query coverage as evidence of broader organic discovery, not as a direct measure of product adoption.
2
Projects and brand; 2 Independent sections designed
Projects and brand; 2 Independent sections designed
6
Writing formats in one module
Writing formats in one module
7
Genuinely distinct decisions documented
Genuinely distinct decisions documented
Reflection
Reflection
What this project taught me
What this project taught me
What this project taught me
The most important thing Timeless Tales taught me is the difference between a decision and an angle on the same decision. It is easy to document ten decisions for a complex product. It is much harder to identify which ones are actually solving different problems and which ones are the same principle described from a different direction. The discipline of a case study is not documenting everything that happened. It is identifying what was genuinely distinct.
Every significant decision in this product had a moment where someone questioned whether it was right. The Foundation restructure came from a CEO question that exposed a structural problem I had not yet named. The Visual Anchor came from diagnosing the actual root cause - not accepting the industry standard workaround of better prompting. The Character Profiling system came from realising that users think in story facts, not prompts. The reasoning that survived the pushback was the right reasoning.
The most important thing Timeless Tales taught me is the difference between a decision and an angle on the same decision. It is easy to document ten decisions for a complex product. It is much harder to identify which ones are actually solving different problems and which ones are the same principle described from a different direction. The discipline of a case study is not documenting everything that happened. It is identifying what was genuinely distinct.
Every significant decision in this product had a moment where someone questioned whether it was right. The Foundation restructure came from a CEO question that exposed a structural problem I had not yet named. The Visual Anchor came from diagnosing the actual root cause - not accepting the industry standard workaround of better prompting. The Character Profiling system came from realising that users think in story facts, not prompts. The reasoning that survived the pushback was the right reasoning.
Timeless Tales taught me that the most consequential design decisions are not the ones that add something. They are the ones that correctly name the problem no one had named yet. The category was not AI video. The constraint was not better prompting. The audience was not whoever could write a prompt. Every decision in this product that actually changed something came from questioning an assumption the entire industry had accepted without examining. That is where the real work is. Not in the execution of known solutions. In the diagnosis that makes a different solution necessary.
Timeless Tales taught me that the most consequential design decisions are not the ones that add something. They are the ones that correctly name the problem no one had named yet. The category was not AI video. The constraint was not better prompting. The audience was not whoever could write a prompt. Every decision in this product that actually changed something came from questioning an assumption the entire industry had accepted without examining. That is where the real work is. Not in the execution of known solutions. In the diagnosis that makes a different solution necessary.
Timeless Tales taught me that the most consequential design decisions are not the ones that add something. They are the ones that correctly name the problem no one had named yet. The category was not AI video. The constraint was not better prompting. The audience was not whoever could write a prompt. Every decision in this product that actually changed something came from questioning an assumption the entire industry had accepted without examining. That is where the real work is. Not in the execution of known solutions. In the diagnosis that makes a different solution necessary.
The most important thing Timeless Tales taught me is the difference between a decision and an angle on the same decision. It is easy to document ten decisions for a complex product. It is much harder to identify which ones are actually solving different problems and which ones are the same principle described from a different direction. The discipline of a case study is not documenting everything that happened. It is identifying what was genuinely distinct.
Every significant decision in this product had a moment where someone questioned whether it was right. The Foundation restructure came from a CEO question that exposed a structural problem I had not yet named. The Visual Anchor came from diagnosing the actual root cause - not accepting the industry standard workaround of better prompting. The Character Profiling system came from realising that users think in story facts, not prompts. The reasoning that survived the pushback was the right reasoning.
Still images stay still - no voice, no motion, no cross-format continuity