MERaLiON API

# Speech-to-text: use the dedicated /audio/transcription endpoint.
# It auto-chunks long audio and can return word-level timestamps and
# speaker diarization. (Do NOT use /v1/chat/completions for transcription.)
curl -X POST http://meralion.org:8010/audio/transcription \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "audio_url": "data:audio/wav;base64,<BASE64_AUDIO>",
    "return_diarization": true
  }'

import base64, pathlib, requests

audio_b64 = base64.b64encode(pathlib.Path("speech.wav").read_bytes()).decode()

# Speech-to-text via the dedicated ASR endpoint — handles long audio and,
# with return_diarization, adds word-level timestamps + speaker labels.
resp = requests.post(
    "http://meralion.org:8010/audio/transcription",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "audio_url": f"data:audio/wav;base64,{audio_b64}",
        "return_diarization": True,   # speaker labels + timestamps
    },
)
message = resp.json()["choices"][0]["message"]
print(message["content"])             # full transcript
print(message.get("segments"))        # [{speaker, start, end, text}, ...]

import fs from "node:fs";

const base64 = fs.readFileSync("speech.wav").toString("base64");

// Speech-to-text via the dedicated ASR endpoint (timestamps + diarization).
const resp = await fetch("http://meralion.org:8010/audio/transcription", {
  method: "POST",
  headers: {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    audio_url: `data:audio/wav;base64,${base64}`,
    return_diarization: true,   // speaker labels + timestamps
  }),
});
const { message } = (await resp.json()).choices[0];
console.log(message.content);    // full transcript
console.log(message.segments);   // speaker-labelled segments