Composable AI media pipelines.
Website • Documentation • Discord
Synthome is the simplest way to build multi-model AI media workflows — all in TypeScript.
AI video and image models behave differently across providers. Different inputs, different outputs, different response formats, different async handling. Building even a simple pipeline becomes a mess of glue code, retries, polling, and media stitching.
Synthome standardizes all of this.
You write a clean pipeline → Synthome handles:
- Model invocation (Fal, Replicate, ElevenLabs, Hume, etc.)
- Async job execution & retries
- Media storage
- Input/output normalization
- Multi-step orchestration
- JSON-defined pipelines (AI-generated or manual)
It’s like OpenRouter, but for AI media pipelines — and fully composable.
- Create your API key at synthome.dev
- Install the SDK
npm install @synthome/sdk- Read the docs https://synthome.dev/docs
Synthome orchestrates multi-step pipelines across providers. Instead of manually chaining API calls, handling webhooks, and stitching media together, you compose operations declaratively and Synthome handles the execution, polling, and media processing.
Yes. You provide your own Replicate, Fal, ElevenLabs, etc. keys. Synthome orchestrates the pipeline but doesn't proxy your provider costs.
We're still working on our contribution guide. Reach out to us on Discord if you'd like to contribute.