Say what you want.
Not which model does it.
Pick an input and an output. Free Models Hub finds a free model that can do the job, chains several together when no single one can, and shows you exactly how it ran.
Text to Image — we'll pick the models and fall back if one is out of quota.
Utilities
Seven routes across text, image and audio.
Text to Text
Ask, rewrite, summarise or extract. Routed to the fastest healthy free LLM.
Generate →
Image to Text
Describe an image, read the text inside it, or pull out structured detail.
Describe →
Audio to Text
Transcribe speech from a recording or a file, up to two minutes.
Transcribe →
- 2 models
Text to Image
Generate an image. Your idea is sharpened into a full prompt first.
Generate →
Text to Audio
Read text aloud. Falls back to your device voice if hosted TTS is spent.
Speak →
- 3 models
Audio to Image
Describe a picture out loud and get it drawn. Three models, one click.
Visualise →
- 2 models
Image to Audio
Hear an image described aloud. Useful as an accessibility aid.
Narrate →
How it works
Free tiers are fragmented, capped and change monthly. The router treats that as a routing problem rather than something you should have to learn.
Composed routing
No model turns speech into a picture. So audio is transcribed, an LLM rewrites the spoken intent into a proper image prompt, and a diffusion model draws it — one click, three models.
Quota-aware fallback
Every model's remaining free budget is checked before the call. If a step fails mid-pipeline, only that step retries on another provider — finished work is never paid for twice.
X-Ray mode
Open the trace on any result to see each provider, model, token count, latency and fallback. The routing is the product, so none of it is hidden.