Popcorn Ready is an AI-native video editor that turns clips and your brief into an editable timeline quickly.

• Upload media and descriptions
• Generate a rough cut automatically
• Revise with chat
• Preview in-browser and export MP4

🌐 Website: https://popcornready.ai

What it is

Popcorn Ready is a content production workflow:

1. Upload your video and image assets with short descriptions.
2. Give the app a creative goal, target length, and style.
3. Generate a first-pass timeline and a critic response.
4. Refine the cut with conversational prompts.
5. Preview in Remotion and export to MP4.

The editor never directly manipulates raw footage. It plans and patches a structured timeline model (src/lib/types.ts) and then renders it deterministically.

• Timeline changes are generated as validated patches.
• Asset metadata and generated media are organized per project store.
• Rendering is reproducible across preview and export.

Features

• Structured pipeline
  • planEdit, selectClips, critique, and revise run through stable JSON contracts.
• AI review loop
  • Timeline suggestions are scored and patched before final playback.
• Interactive revision
  • Ask for changes in plain language and apply targeted patch updates.
• Generative fallback assets
  • Missing visuals can be auto-generated from provider integrations.
• Export-ready output
  • In-browser preview and MP4 export are part of the same timeline pipeline.

Who it is for

• Marketers and creators iterating quickly on short-form campaigns.
• Product teams creating consistent brand motion content.
• Creators wanting a fast first draft before manual finishing.

Typical flow

1. Upload media assets.
2. Add short descriptions for each asset.
3. Use Generate missing asset for any gaps.
4. Set length/aspect/style, then click Generate rough cut.
5. Inspect the plan, timeline, and critic scores.
6. Revise with commands like:
   • make it punchier
   • shorten to 15s
   • add captions
   • use less talking head
7. Export MP4.

Setup

cp .env.local.example .env.local   # add provider keys as needed
npm install
npm run dev

Open http://localhost:3000

The home page (/) is the marketing landing page: it explains the product, lets you create a 30-second video from a single prompt (with template chips to start from), lists hosted pricing, and links to GitHub for self-hosting. Submitting the prompt opens the studio (/studio) and one-shots the video: it plans beats, generates a visual for each beat, and cuts a timeline — no uploads required.

By default the one-shot generates a real video clip per beat (OpenAI Sora when OPENAI_API_KEY is set, Gemini Veo when GEMINI_API_KEY is set), so you get an actual moving 30-second video. This is expensive, so it is gated by a kill switch: set ONESHOT_VIDEO=off to fall back to fast still-frame generation under load. With no video-capable key it uses still images (real with OPENAI_API_KEY, placeholder frames without) so the flow always completes. You can also go straight to /studio to bring your own clips with the full editor below.

1. Upload a handful of video or image assets. Add a short description for each — in this MVP the AI reasons over the filename + your description + duration (real transcription/vision analysis is the documented next step, not in this slice).
2. If the library is missing a visual, use Generate missing asset to create an image or short video asset. OpenAI is live when OPENAI_API_KEY is set; Gemini video generation is live when GEMINI_API_KEY is set. ElevenLabs audio generation is live when ELEVENLABS_API_KEY is set.
3. Write a creative goal, set length/aspect/style, and Generate rough cut.
4. Inspect the plan, timeline, and critic scores; preview plays in the browser.
5. Revise (chat): "make it punchier", "shorten to 15s", "add captions", "use less talking head". Each message is turned into timeline patches.
6. Export MP4: renders the real cut of your clips via Remotion. The first export downloads a headless browser, so it takes a bit longer.

Scope / limitations (deliberate, for the MVP)

• Clip understanding is description-based — no FFmpeg proxies, Whisper transcription, vision tagging, or embeddings yet (those are the "real analysis" extension from the architecture doc).
• Single project, file-based store (no Postgres/pgvector, no auth, no queue).
• Critic runs one pass on generate; the full critique→re-render loop and multiple rough-cut variants are future work.
• Gemini image generation and NanoBanano provider adapters are placeholders in this pass.
• Generated audio is saved as an asset but is not yet mixed into exported timelines. Audio clips are excluded from visual clip selection prompts.
• MP4 export requires the dev server running (Remotion fetches the uploaded clips over http://localhost:3000).

Productionization docs

• docs/productionization-scope.md
• docs/railway-deployment.md
• docs/streaming-generation-plan.md

Railway configuration is in railway.toml; healthcheck is /api/v1/health.

Deploy to Railway

Railway deployment notes live in docs/railway-deployment.md. The repo includes a railway.toml that uses Railpack, runs npm run build, starts the service with npm run start, and healthchecks /api/v1/health.

Set the provider keys from .env.local.example as Railway service variables. For a hosted demo, be aware that the current MVP stores project state and media on the local filesystem; see the Railway deployment doc for the persistence limitations and production storage recommendations.

Project layout

src/
  app/                Next.js App Router (landing + studio + API routes)
    page.tsx          marketing landing page (/)
    studio/page.tsx   the editor (/studio); ?goal=&length=&autostart=1 one-shots
    api/{project,upload,generate,oneshot,generate-assets,revise,export}/route.ts
  components/         Editor (UI) + Preview (Remotion Player)
  lib/
    agent/            planEdit / selectClips / critique / revise + JSON schemas
    anthropic.ts      Claude client + structured JSON call helper
    timeline.ts       Patch engine + prompt formatting
    types.ts          Timeline / Plan / Patch / Clip types
    store.ts          JSON-file project store
    generative/       Provider abstraction + OpenAI and Gemini adapters
  remotion/           VideoComposition + registered root for render/preview