Transcribe7 API

Speech-to-text for your apps: real-time (WebSocket) and async (files), with speaker diarization, language identification, translation (one-way and two-way) and context. The API follows the widely used STT REST + WebSocket conventions, so existing clients usually only need a new base URL and key.

REST base URLhttps://api.transcribe7.com/v1
WebSocket URLwss://api.transcribe7.com/transcribe-websocket
API keyst7_live_… (servers) Β· t7_temp_… (temporary, for browsers / apps)
πŸ€– Using an AI coding assistant? Give it this link and let it do the integration: https://api.transcribe7.com/llms-full.txt β€” the complete reference in one Markdown file (also: llms.txt Β· instructions for AI assistants).

Introduction

Authentication

Authorization: Bearer t7_live_…

Keep t7_live_ keys on your servers. Browsers and mobile apps use a temporary key created by your backend.

Errors & limits

REST errors:

{"status_code":400,"error_type":"invalid_request","message":"…","validation_errors":[{"error_type":"missing","location":"body.model","message":"Field required"}],
 "request_id":"…","more_info":"https://api.transcribe7.com/docs/#errors"}

WebSocket errors arrive as a message, then the connection closes:

{"error_code":401,"error_type":"unauthenticated","error_message":"Incorrect API key provided.","request_id":"…"}
Statuserror_type
400invalid_request, invalid_cursor, and model / audio specific types
401unauthenticated
402account_balance_exhausted β€” no credit left (top up at platform.transcribe7.com) Β· account_budget_exhausted β€” monthly allowance used up
404file_not_found, transcription_not_found
409transcription_invalid_state (e.g. transcript requested before completion)
413file_too_large (max 1 GB), max_duration_reached (WebSocket)
429limit_exceeded β€” request rate, concurrent streams or pending transcriptions (see Retry-After)
503service_unavailable β€” retry shortly

Limits are per account: concurrent real-time streams, pending transcriptions, stored files (200 / 20 GB), optional monthly audio allowance. Usage is paid from prepaid credit: a stream or transcription needs credit to start, and a real-time stream ends with account_balance_exhausted when the credit it may use runs out (running streams share your balance; each open file transcription holds 1 hour of credit until its real length is known).

Real-time WebSocket API

WSwss://api.transcribe7.com/transcribe-websocket

Send the key with the connection β€” in the Authorization: Bearer header (servers) or, in browsers, in the protocols list: new WebSocket(url, ["t7-api-key", temporaryKey]). (An api_key field in the start message is also accepted.) A wrong key doesn't fail the handshake: an error message arrives, then the connection closes.

  1. Send the start message (JSON text).
  2. Stream audio as binary messages, in real time (e.g. 100 ms chunks).
  3. Optional control messages: {"type":"finalize"} (finalize pending tokens now), {"type":"keepalive"} (during pauses, at least every 20 s).
  4. End of audio: send an empty binary message. Remaining tokens arrive, then {"finished":true}, and the server closes the connection.
// Node.js (npm i ws)
import WebSocket from "ws";
import fs from "fs";

const ws = new WebSocket("wss://api.transcribe7.com/transcribe-websocket", {
  headers: { Authorization: `Bearer ${process.env.T7_API_KEY}` },
});
ws.on("open", () => {
  ws.send(JSON.stringify({
    model: "stt-rt-v6",
    audio_format: "pcm_s16le", sample_rate: 16000, num_channels: 1,
    language_hints: ["sl", "en"],
    enable_speaker_diarization: true,
    enable_language_identification: true,
  }));
  const pcm = fs.readFileSync("audio.raw");            // 16 kHz mono s16le
  let i = 0;
  const t = setInterval(() => {
    if (i >= pcm.length) { clearInterval(t); ws.send(Buffer.alloc(0)); return; }   // empty = end of audio
    ws.send(pcm.subarray(i, i += 3200));                // 100 ms
  }, 100);
});
ws.on("message", (data) => {
  const res = JSON.parse(data.toString());
  if (res.error_code) return console.error(res.error_type, res.error_message);
  for (const t of res.tokens ?? []) if (t.is_final) process.stdout.write(t.text);
  if (res.finished) console.log("\n[done]");
});

Start message

FieldDescription
modelReal-time model, e.g. stt-rt-v6 β€” see GET /v1/models.
audio_formatauto (aac, aiff, amr, asf, flac, mp3, ogg, wav, webm) or raw: pcm_s16le, pcm_f32le, mulaw, alaw…
sample_rate, num_channelsRequired for raw formats.
language_hintsExpected languages, e.g. ["sl","en"]. language_hints_strict to rely on them more.
enable_language_identificationAdds language to each token.
enable_speaker_diarizationAdds speaker to each token.
enable_endpoint_detectionFinalizes when the speaker stops and sends an <end> token. Tuning: endpoint_sensitivity, max_endpoint_delay_ms, endpoint_latency_adjustment_level.
translation{"type":"one_way","target_language":"en"} or {"type":"two_way","language_a":"sl","language_b":"en"}
context{"general":[{"key":"Domain","value":"medicine"}],"text":"…","terms":["Transcribe7"],"translation_terms":[{"source":"…","target":"…"}]}
client_reference_idYour own tracking id (max 256).

Responses

{"tokens":[{"text":"How","start_ms":600,"end_ms":760,"confidence":0.97,"is_final":true,
            "speaker":"1","language":"en","translation_status":"original"}],
 "final_audio_proc_ms":760,"total_audio_proc_ms":1250}

{"tokens":[],"finished":true}

is_final:false tokens are provisional and are re-sent until they become final; final tokens are sent once. With translation, translated tokens have translation_status:"translation". Special tokens: <end> (endpoint), <fin> (after finalize).

Temporary keys

POST/v1/auth/temporary-api-key β€” call from your backend, hand the key to the client.

curl https://api.transcribe7.com/v1/auth/temporary-api-key \
  -H "Authorization: Bearer $T7_API_KEY" -H "Content-Type: application/json" \
  -d '{"usage_type":"transcribe_websocket","expires_in_seconds":60,"single_use":true,"max_session_duration_seconds":3600}'

β†’ 201 {"api_key":"t7_temp_…","expires_at":"2026-11-13T10:01:00.000Z"}
// Browser
const ws = new WebSocket("wss://api.transcribe7.com/transcribe-websocket", ["t7-api-key", tempKey]);
ws.onopen = () => ws.send(JSON.stringify({ model: "stt-rt-v6", audio_format: "auto" }));

expires_in_seconds (1–3600) limits when new streams can be opened; single_use allows one stream; max_session_duration_seconds caps each stream; client_reference_id is bound to the key.

Files

POST/v1/filesMultipart: file (+ optional client_reference_id). Max 1 GB. β†’ 201 {"id","filename","size","created_at","client_reference_id"}
GET/v1/files?limit=&cursor={"files":[…],"next_page_cursor":…} β€” newest first, limit 1–1000.
GET/v1/files/count{"playground":0,"public_api":N,"total":N}
GET/v1/files/{id}File metadata.
DELETE/v1/files/{id}204
curl https://api.transcribe7.com/v1/files -H "Authorization: Bearer $T7_API_KEY" -F file=@meeting.m4a

Transcriptions

POST/v1/transcriptions β†’ 201 transcription object

curl https://api.transcribe7.com/v1/transcriptions \
  -H "Authorization: Bearer $T7_API_KEY" -H "Content-Type: application/json" \
  -d '{"model":"stt-async-v6","file_id":"84c32fc6-…","language_hints":["sl"],
       "enable_speaker_diarization":true,"enable_language_identification":true,
       "translation":{"type":"one_way","target_language":"en"},
       "webhook_url":"https://example.com/hooks/t7","client_reference_id":"order-42"}'

Use either file_id (a file you uploaded) or audio_url (a public URL). Other fields: language_hints_strict, context (as in real-time), webhook_auth_header_name / webhook_auth_header_value.

{"id":"548d023b-…","status":"queued","created_at":"…","model":"stt-async-v6","audio_url":null,"file_id":"84c32fc6-…",
 "filename":"meeting.m4a","language_hints":["sl"],"enable_speaker_diarization":true,"enable_language_identification":true,
 "audio_duration_ms":null,"error_type":null,"error_message":null,"webhook_url":"https://example.com/hooks/t7",
 "webhook_auth_header_name":null,"webhook_auth_header_value":null,"webhook_status_code":null,"client_reference_id":"order-42"}

Status: queued β†’ processing β†’ completed | error.

GET/v1/transcriptions?limit=&cursor={"transcriptions":[…],"next_page_cursor":…}
GET/v1/transcriptions/countCounts.
GET/v1/transcriptions/{id}Status and details (poll every few seconds, or use a webhook).
DELETE/v1/transcriptions/{id}204 (409 while still processing).

Transcript

GET/v1/transcriptions/{id}/transcript (when completed)

{"id":"548d023b-…","text":"Hello everyone…",
 "tokens":[{"text":"Hel","start_ms":0,"end_ms":180,"confidence":0.98,"speaker":"1","language":"en","translation_status":"original"}, …]}

Webhooks

With webhook_url we POST {"id":"…","status":"completed"} (or "error") when the transcription finishes, with your webhook_auth_header_name: value if set (3 attempts). The URL must be publicly reachable. The delivery result is in webhook_status_code. You can add your own query parameters to the URL.

Models

GET/v1/models β€” available models (real-time and async), their languages and translation support.

Usage

GET/v1/usage?month=2026-11 (default: this month, UTC)

{"month":"2026-11","realtime_seconds":5400,"async_seconds":7260,"total_seconds":12660,"monthly_cap_seconds":0,
 "daily":[{"date":"2026-11-13","realtime_seconds":5400,"async_seconds":7260,"realtime_sessions":3,"transcriptions":2}]}

Data retention

Uploaded files are deleted 24 hours after upload, and transcriptions (with their transcripts) 24 hours after they finish. Download what you need within that time, or delete earlier yourself. Real-time audio and text are not stored.

Questions or higher limits: support@transcribe7.com