Turn a rough video brief into a continuity-aware shot map and generator-ready prompt packet.
Live demo: https://shotmap-agent.vercel.app
- Converts one creative brief into a timed five-to-seven-shot sequence with camera and transition direction.
- Maintains a continuity ledger across subject, wardrobe, props, location, lighting, geography, and screen direction.
- Produces independently usable per-shot prompts, negative guidance, a ranked risk pass, and an assembly checklist.
- Keeps every provider key in sessionStorage and sends it only to the same-origin BYOK route for each request.
bun install
bun devVisit http://localhost:3000, choose a supported model, and paste your own provider key in the session panel.
+-------------------------------+
| BYOK agent app |
+---------------+---------------+
|
v
+-------------------------------+
| Shotmap Agent |
+---------------+---------------+
|
v
+-------------------------------+
| User-owned runtime / config |
+-------------------------------+
- The workbench packages the brief, channel, aspect ratio, duration, visual direction, continuity anchors, and constraints into one planning request.
- The frozen same-origin route reads the visitor key from x-provider-key, constructs the selected provider client, and streams the response without persistence or logging.
- Shotmap runs brief audit, continuity extraction, shot architecture, prompt synthesis, and risk review before returning the production packet.
- No author-owned inference key is included or required.
- Visitor keys live only in browser sessionStorage and are redacted from runtime errors.
- The starter registry includes OpenAI GPT-5.5 through the Responses API, Claude Sonnet 4.6, and Gemini 3.5 Flash.
- The scaffold resolves AI SDK 6.0.208 and @ai-sdk/openai 3.0.74.
bun install
bun run lint
bun run typecheck
bun run build
bun run test- Next.js 16
- AI SDK 6
- OpenAI Responses API
- Tailwind CSS 4
- shadcn/ui
- Playwright
- BYOK runtime
Read CONTRIBUTING.md before opening an issue or pull request. This project follows the Code of Conduct.
- Fork it for product launches, social campaigns, narrative experiments, or any workflow that needs repeatable generative-video shots.
- Reference the release from the dvnc.dev open-source catalog and focused AI-video creator communities.
MIT - see LICENSE. Built by Omid Saffari.
