LLM Proxy

Appearance

Create speech ​

Text-to-speech. Returns raw audio bytes (non-streaming) or a byte stream (stream: true).

POST/v1/audio/speech

The endpoint is excluded from the non-streaming keepalive because the body is binary.

Request
bash
curl http://localhost:8080/v1/audio/speech \
  -H "Authorization: Bearer $LLM_PROXY_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "tts-model", "input": "Hello from the proxy!", "voice": "alloy"}' \
  --output speech.mp3
Request — OpenAI SDK
python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/v1", api_key="sk-your-proxy-key")
audio = client.audio.speech.create(
    model="tts-model",
    voice="alloy",
    input="Hello from the proxy!",
)
audio.stream_to_file("speech.mp3")
Response
http
HTTP/1.1 200 OK
Content-Type: audio/mpeg

<binary audio bytes — save or play directly>

Request fields ​

model, input (≤ 4096 chars), voice, instructions, response_format (mp3 default; opus, aac, flac, wav, pcm), speed (0.25–4.0), stream_format (sse/audio), stream.

Response ​

  • Non-streaming: raw audio bytes with the upstream content type (fallback audio/<format>; mp3 → audio/mpeg).
  • stream: true: byte stream; Gemini TTS advertises PCM/WAV.

Provider support ​

openai, the openai-compatible family (except qwen/qwen-intl, which have no upstream speech endpoint), and gemini (native TTS) — see the capability matrix.

Request
bash
curl http://localhost:8080/v1/audio/speech \
  -H "Authorization: Bearer $LLM_PROXY_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "tts-model", "input": "Hello from the proxy!", "voice": "alloy"}' \
  --output speech.mp3
Request — OpenAI SDK
python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/v1", api_key="sk-your-proxy-key")
audio = client.audio.speech.create(
    model="tts-model",
    voice="alloy",
    input="Hello from the proxy!",
)
audio.stream_to_file("speech.mp3")
Response
http
HTTP/1.1 200 OK
Content-Type: audio/mpeg

<binary audio bytes — save or play directly>

MIT licensed

LLM Proxy