The world's first role-based AI creative studio — where specialized AI agents collaborate to produce professional short-form video content through an intelligent project graph system.
TakeOne treats post-production as an operating system, not a collection of point tools. Four specialized agents (Scriptwriter, Storyboard Artist, Cinematographer, Reel Editor) coordinate through a shared project graph with cross-scene continuity tracking and 230+ configurable parameters.
This is the product design & specification layer for TakeOne — everything a team needs to pick up the project and start building.
takeone/
├── docs/
│ ├── PRD.md # Full Product Requirements Document (~190KB)
│ └── tech-architecture.pdf # System architecture diagrams
│
├── design/
│ ├── cinematic-dark/ # 19 screens — dark cinematic theme
│ └── ivory-light/ # 19 screens — light ivory theme
│
├── CONTRIBUTING.md
└── README.md
The PRD covers 12 sections: problem statement & market sizing, three user personas, full feature specs for each agent, technical architecture (6-layer prompt assembly engine, project graph schema, credit system), UX workflows, risk mitigation, go-to-market strategy, pricing tiers (Free / Starter $9 / Pro $39 / Studio $99 / Enterprise), success metrics, and a phased roadmap.
The tech architecture PDF contains system diagrams for the agent orchestration layer, data flow, and infrastructure topology.
38 interactive HTML prototypes covering every core screen across two visual themes. Each screen has a .html file (live Tailwind CSS prototype you can open in any browser) and a .png screenshot.
Screen inventory:
| # | Screen | Description |
|---|---|---|
| 01 | Homepage | Landing page with value proposition and CTA |
| 02 | Dashboard | Project overview, recent work, quick actions |
| 03 | New Project | Project creation wizard with brand kit integration |
| 04 | Scriptwriter Agent | AI script generation with tone, structure, and hook controls |
| 05 | Storyboard Agent | Visual scene planning — AI generation, manual uploads, visual intelligence |
| 06 | Reel Editor Agent | Timeline editing, media selection, transitions |
| 07 | Brand Kit | Logo, colors, fonts, voice guidelines management |
| 08 | Cinematographer Agent | Camera language, shot composition, visual style controls |
| 09 | Project Export | Render settings, format selection, delivery |
| 10 | Collaboration | Team access, permissions, project sharing |
| 11 | Credits & Billing | Usage tracking, plan management, payment |
| 12 | Profile | Account settings and preferences |
| 13 | Support Center | Help docs, ticket submission |
| 14 | Security Settings | Password, 2FA, session management |
| 15 | Pricing | Plan comparison and subscription management |
Content creators produce 10–30 short-form videos per week to stay visible. A single 30-second professional reel takes 4–6 hours across 5–7 different tools. Existing AI tools generate assets but don't orchestrate workflows — they're primitives, not professionals.
TakeOne closes this gap with agent specialization (each agent has deep domain knowledge in its craft) and cross-agent orchestration (the project graph maintains context from script through final export). The result: professional-quality content in under 15 minutes.
Target: $500K ARR Year 1, 25,000 active creators.
See CONTRIBUTING.md for a detailed guide on how to start building from this spec.
Quick orientation:
- Read
docs/PRD.md— Sections 1–3 for context, Section 5 for architecture, Section 4 for feature specs - Open any
design/*.htmlfile in a browser to explore the UI - Review
docs/tech-architecture.pdffor system diagrams - Check the open decisions in PRD Section 12
- No backend code yet — this repo is the design/spec layer. The PRD Section 5 (Technical Architecture) defines the stack and schema.
- Binary files with legacy branding —
tech-architecture.pdfand all.pngscreenshots still show "Crushed Studios" (the original working title). The HTML prototypes and PRD markdown are fully rebranded to TakeOne. - No CI/CD, tests, or infra config — these are defined in the PRD roadmap (Phase 1–3) but not yet implemented.
MIT
Rishi Bhattacharjee