# TongFlow

> An open-source multi-modal generative AI workflow studio. Every AI model is a node on an infinite canvas — wire modalities together, combine the results.

TongFlow lets you build generative AI workflows visually. Each AI model is wrapped as a modality-transform node: text→image, image→video, audio→text, and more. Nodes connect on an infinite canvas using three operations: Add (bring in materials), Transform (run a model), Combine (merge outputs). Available as a desktop app, self-hosted, or via cloud.

## Quick facts

- License: AGPL-3.0 (free for personal, research, and open-source use)
- Version: 0.0.1 — released June 2026
- Local-first: all workflows and files stored in local SQLite, no account required
- GPU inference: Modal cloud (free H100 tier available)
- LLM keys: bring your own (OpenRouter, Gemini, OpenAI, DeepSeek)
- Platforms: macOS (arm64 / x64), Windows, self-hosted (Node.js 20+)

## Models integrated

- Image generation: Z-Image, FLUX.2 Klein 9B
- Video generation: LTX-2, SeedVR2, Wan-Animate (character swap / motion transfer)
- Music generation: ACE-Step
- Speech: Qwen3-TTS, Qwen3-ASR, Whisper, InfiniteTalk (lip sync)
- Text / LLM: Qwen3, Gemini, OpenAI, OpenRouter-Free
- Utilities: FFmpeg, PaddleOCR, Docling, Crawl4AI

## Self-hosting (one command)

```
git clone https://github.com/tong-io/tongflow
cd tongflow && pnpm install && pnpm dev
```

Requires: Node.js 20+, a Modal token (free tier works), and one LLM API key.

## Links

- Homepage: https://tongflow.com
- GitHub (source): https://github.com/tong-io/tongflow
- Cloud: https://app.tongflow.com
- Discord community: https://discord.gg/K7V8az94Zf
- Business contact: business@tongflow.com

## Compared to alternatives

- vs ComfyUI: TongFlow covers all 7 modalities (not image-only); Combine nodes (lip sync, image fusion, motion transfer) are built-in
- vs n8n: TongFlow is purpose-built for AI model chaining, not general API orchestration
- vs Zapier/Make: no-code but AI-native; local-first with no data leaving your machine by default

## Plugin SDK

Developers can add custom AI backends. Define a slot in the ABI, annotate a Python method with `@node_slot`, publish as a package. Any backend works: Modal, Replicate, local GPU, or plain API.

SDK: https://pypi.org/project/tongflow/