sorker/ai-shotlive

AI ShotLive Director – One‑stop AI short‑drama / motion‑comic production platform

What it is

• A web (and desktop‑via Electron) application that lets creators turn a novel or a story outline into a full‑length short video or comic. The workflow is fully industrialised: Novel → Script → Storyboard → Assets → Keyframes → Final video.
• It does not host any AI models itself; instead it calls external text, image, video and audio APIs (OpenAI, Anthropic, Google Gemini, AntSK, etc.) through a thin adapter layer.

Core ideas

• Keyframe‑driven generation – the system first creates a start‑frame and an end‑frame for each shot, then interpolates the middle frames with a video model. This keeps visual consistency while reducing the amount of expensive video generation.
• Asset constraints – character “make‑up” images and scene concept art are used as references for every generated frame, ensuring the same character looks the same across shots.
• Multi‑model support – you can mix‑and‑match providers for text, image, video and audio, swapping models by simply changing the API key in the Model Config page.

Main modules

Phase	Functionality
01 – Novel & Script	Upload a .txt novel, auto‑parse chapters, pick chapters to generate episode scripts; or paste a story outline directly. The AI produces scene breakdowns, dialogue and visual prompts.
02 – Assets & Casting	Generate character reference images (with wardrobe variations) and scene concept art; upload custom props.
03 – Director Workbench	Grid‑based storyboard editor; edit first/last keyframes; 9‑grid preview for each shot; context‑aware generation that re‑uses the current character and scene images; supports both image‑to‑video (I2V) and keyframe‑interpolation modes.
04 – Export & Editing	Timeline preview, multi‑track video editor (video, audio, subtitles, images), AI‑generated subtitles and TTS audio, then export as high‑resolution frames or MP4.
Account & Data	Per‑user isolation, password change, full project/asset backup‑and‑restore via a ZIP that contains the MySQL/SQLite data and the data/ media folder.

Technology stack

• Frontend: React 19 + Vite 6, Tailwind CSS 4, Radix UI, Framer Motion.
• Backend: Express.js, JWT auth, MySQL 8 (or SQLite for local mode).
• Desktop client: Electron 33 with better‑sqlite3; builds for macOS (Intel & Apple‑Silicon) and Windows.
• AI integration: Adapter layer (services/adapters) that normalises calls to any supported provider for chat, image, video and audio generation.
• Deployment: Docker compose (with optional external MySQL) or simple npm run dev for local development.

Typical workflow for a creator

1. Log in (default admin/admin123 or register).
2. Configure models – add API keys for the text, image, video and audio services you want to use.
3. Create a project → upload a novel or paste a story.
4. Generate a script – AI rewrites selected chapters into a screenplay with scene descriptions.
5. Produce assets – generate character look‑books and scene concept art; upload any custom props.
6. Storyboard – edit shots, set first/last keyframes, choose a 9‑grid view, and let the video model render the clip.
7. Edit & polish – use the built‑in multi‑track editor to add AI‑generated subtitles, TTS voice‑overs, or extra media.
8. Export – download the final video or the individual high‑resolution frames.

Getting started

# clone and install
git clone https://github.com/sorker/ai-shotlive.git
cd ai-shotlive
npm install
cp .env.example .env   # edit DB and JWT settings

# local SQLite mode (no external DB needed)
npm run build:local
npm run start:local   # http://localhost:3001

For production you can use the Docker files (docker-compose.yaml or docker-compose.mysql.yaml). The Electron client can be built with npm run build:electron and distributed as a .dmg or .exe.

Who might use it

• Independent creators who want to prototype animated shorts or motion comics without learning video‑editing software.
• Small studios looking for a rapid “script‑to‑video” pipeline that can be customised with their own model providers.
• Educators demonstrating AI‑augmented storytelling or keyframe‑driven video generation.

License

• Creative Commons BY‑NC‑SA 4.0 – free for personal and non‑commercial use, with the requirement to share derivative works under the same license.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.