HBAI-Ltd/Toonflow-app

Toonflow – AI‑powered short‑film creation platform

What it is – Toonflow is an open‑source desktop‑and‑server application that lets creators write scripts, manage assets, and generate images and videos with AI, all on a single “infinite canvas”. It is built with TypeScript, Vue 3, Bun, and Electron‑like Electrobun for the desktop client.

Key capabilities

Feature	What it does
Local‑first deployment	Runs on Windows, macOS, Android (ARM64) and on Linux servers or Docker containers. All project files and assets stay on the user’s device or self‑hosted server.
Infinite canvas	A visual workspace where scripts, characters, scenes and generated video clips are arranged together.
MCP (Model‑Connector‑Plugin)	A generic bridge that lets the app call external AI services (image, video, LLM) or a locally‑run ComfyUI pipeline.
Plugin marketplace	Users can install additional nodes, tools or whole workflows from an online store (api.toonflow.net/console/plugIn).
Open agents	Prompt templates, tool definitions and A2A (agent‑to‑agent) flows can be customised, enabling bespoke AI assistants inside a project.
Model‑agnostic	Any third‑party API that follows the OpenAI‑style request format can be configured, and local models can be hooked in.
21 UI languages	Full interface localisation (Chinese, English, Japanese, Russian, Vietnamese, Thai, Korean, Hindi, Indonesian, Malay, Filipino, Bengali, Urdu, Tamil, Telugu, Marathi, Punjabi, Arabic, Persian, Turkish, etc.).

How you use it

1. Install – download a native client (Windows / macOS / Android) or spin up the Docker/Server version following the step‑by‑step instructions in the README.
2. Create a project – write a script, add characters and scenes, then drag them onto the canvas.
3. Generate assets – configure a video model (e.g., Seedance 2.0), an image model (GPT Image 2) and an LLM (Claude Opus 4.6) via the MCP; the app sends prompts and receives generated media.
4. Edit & preview – use the built‑in 3‑D director studio to preview shots, adjust timing, and stitch clips together.
5. Export – render the final short‑film or export individual assets.

Installation flavours

• Desktop client – bundles Bun and a WebView2/WebKit runtime; Android version includes a local Bun backend.
• Docker – docker compose up -d --build builds an image with Bun 1.3.14 and FFmpeg; data lives in a Docker volume.
• Bare‑metal server – install Git, FFmpeg, Bun 1.3.14, then run bun install && bun run build:server && bun run start:server.
• Cloud image – pre‑configured images are offered on the “智星云” platform.

Community & ecosystem

• Plugin market – extend the canvas with new nodes, AI tools, or whole workflow packs.
• TF‑Router – an open‑source model‑proxy service (HBAI‑Ltd/TF‑Router) that can be self‑hosted to route API calls securely.
• Support channels – Discord server, WeChat group, and a public issue tracker.

License – MIT (with a note that versions before v1.0.8 were AGPL‑3.0). The project is actively maintained (130 days of development, >800 commits, 20 releases at the time of writing).

Bottom line – Toonflow is a genuine, AI‑driven content‑creation tool aimed at short‑film and animation makers. It combines script writing, asset management, and AI‑generated media in a single, locally‑hostable application, making it relevant to anyone interested in generative AI for visual storytelling.