What's growing today?
Pick a model above and start typing. Attach an image for vision-capable models.
🧪 Experimental — video generation support and speed vary a lot by provider/model. If it fails, that's usually the provider, not Meadows.
Build a real pipeline: your voice → STT → LLM → TTS — pick a different model for each stage, even across different providers. For the genuine low-latency realtime protocol with one integrated model instead, see 🎧 Realtime.
The genuine low-latency Realtime API — a live WebSocket to the model (server voice-activity detection handles turn-taking for you), not a chained approximation.
Account
Sign in with your real Pollinations account (OAuth 2.1 + PKCE). The resulting key lives server-side behind an httpOnly session cookie — this page's JavaScript never sees it, and there's nothing to paste.
Active endpoint
Meadows talks to Pollinations by default. Switch to any OpenAI-compatible endpoint here — OpenRouter, OpenAI itself, a local gateway — if you'd rather bring your own.