Speech to Text

# 1. Upload the bytes — the ordinary way from a form or a file picker.
curl -s https://zyraxonai.lovable.app/api/public/stt \
  -F "file=@recording.mp3" \
  -F "language=bn" | jq -r '.text'

# 2. Base64 in JSON — for a client that only speaks JSON.
curl -s https://zyraxonai.lovable.app/api/public/stt \
  -H "Content-Type: application/json" \
  -d '{"audio":"'"$(base64 -w0 recording.mp3)"'","language":"bn"}' | jq -r '.text'

# 3. A URL the server fetches for you — no upload, no base64 inflation.
curl -s https://zyraxonai.lovable.app/api/public/stt \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/recording.mp3","language":"bn"}' | jq -r '.text'

{
  "success": true,
  "text": "আজ আকাশ পরিষ্কার এবং সূর্য উজ্জ্বল।",
  "language": "bn",
  "locale": "bn-BD",
  "confidence": 0.94,
  "engine": "chromium",
  "container": "mp3",
  "duration": 8.1,
  "elapsed": 1.96,
  "truncated": false
}

# .m4a -> .wav, keeping the sample rate. One command, no quality loss.
ffmpeg -i memo.m4a -ac 1 -ar 16000 memo.wav
curl -s https://zyraxonai.lovable.app/api/public/stt -F "file=@memo.wav" -F "language=en" | jq -r '.text'

# Or straight to MP3, which is smaller to upload.
ffmpeg -i memo.m4a -ac 1 -ar 16000 -b:a 64k memo.mp3

# Live, from any audio source. One second of 16 kHz mono PCM is 32000 bytes.
ffmpeg -i recording.mp3 -ar 16000 -ac 1 -f s16le - \
| curl -sN -X POST "https://zyraxonai.lovable.app/api/public/stt?stream=1&language=bn" \
    -H "Content-Type: audio/l16; rate=16000" \
    -H "Accept: text/event-stream" \
    --data-binary @-

# event: ready     {"rate":16000,"language":"bn-BD"}
# event: partial   {"text":"আজ আকাশ"}
# event: partial   {"text":"আজ আকাশ পরিষ্কার"}
# event: segment   {"text":"আজ আকাশ পরিষ্কার এবং সূর্য উজ্জ্বল।","seconds":4.2}
# event: done      {"text":"...","segments":2,"seconds":8.1}

// Live transcription from a browser microphone.
const STT = "https://zyraxonai.lovable.app/api/public/stt";

async function start(language: string, onText: (text: string, done: boolean) => void) {
  const stream = await navigator.mediaDevices.getUserMedia({ audio: true });

  // Ask the browser for the rate the endpoint wants, so no resampling is needed.
  const ctx = new AudioContext({ sampleRate: 16000 });
  const source = ctx.createMediaStreamSource(stream);

  // An AudioWorklet hands us raw frames; a ScriptProcessor would work too but
  // is deprecated. Either way the goal is the same: 16-bit mono samples.
  const worklet = await ctx.audioWorklet.addModule("/pcm-worklet.js");
  const node = new AudioWorkletNode(ctx, "pcm-worklet");
  source.connect(node);

  const res = await fetch(`${STT}?stream=1&language=${language}`, {
    method: "POST",
    headers: {
      "Content-Type": "audio/l16; rate=16000",
      Accept: "text/event-stream",
    },
    body: new ReadableStream({
      start(controller) {
        node.port.onmessage = (e) => controller.enqueue(new Uint8Array(e.data));
      },
      cancel() {
        node.disconnect();
        ctx.close();
      },
    }),
    duplex: "half",
  } as RequestInit & { duplex: "half" });

  // Parse the SSE frames: a blank line ends one event.
  const reader = res.body!.getReader();
  const decoder = new TextDecoder();
  let buffer = "";
  for (;;) {
    const { value, done } = await reader.read();
    if (done) break;
    buffer += decoder.decode(value, { stream: true });
    const frames = buffer.split("\n\n");
    buffer = frames.pop() ?? "";
    for (const frame of frames) {
      const line = frame.split("\n").find((l) => l.startsWith("data:"));
      if (!line) continue;
      const ev = JSON.parse(line.slice(5).trim());
      if (ev.type === "partial") onText(ev.text, false);
      if (ev.type === "segment") onText(ev.text, true);
      if (ev.type === "done") onText(ev.text, true);
    }
  }
}

# Bengali, with the region
curl -s https://zyraxonai.lovable.app/api/public/stt -F "file=@recording.mp3" -F "language=bn-BD" | jq -r '.text'

# Hindi
curl -s https://zyraxonai.lovable.app/api/public/stt -F "file=@recording.mp3" -F "language=hi" | jq -r '.text'

# The live mode takes it the same way, in the query string.
curl -sN -X POST "https://zyraxonai.lovable.app/api/public/stt?stream=1&language=bn" --data-binary @audio.pcm

curl -s https://zyraxonai.lovable.app/api/public/stt | jq '{containers, limits, longAudio}'