Streaming, multi-turn. Great for coding, writing, research, and planning.
Free voices via StreamElements (Brian, Amy, Joanna, Matthew…). Set TTS_PROVIDER=edge for neural voices.
Free via Pollinations.ai. Tip: ask Kimi K3 in Chat to write a detailed prompt for you.
Requires REPLICATE_API_TOKEN on the server (paid per run) or a self-hosted GPU model. K3 is a language model — it can't render video itself.
Generates a portrait, then animates the mouth in sync with TTS audio — fully free. (True talking-head video needs a GPU model like Wav2Lip — see README.)
Kimi K3 writes a complete single-file website and your server hosts it instantly at /sites/<slug>/.