Key Info
LiteLLM has added streaming support for text-to-speech audio on the /v1/audio/speech endpoint, so applications can start playing audio before the provider finishes sending the full clip.
Highlights
- Send POST /v1/audio/speech with "stream": true, or use stream=True in the SDK.
- LiteLLM forwards each chunk as the provider produces it; a long ElevenLabs clip begins playing after 0.72 seconds.
- When audio ends, LiteLLM writes the spend row and releases the key's max_parallel_requests slot.
- The feature was merged via a public PR.