API DocsTranscription

Transcription

Transcribe audio files, optionally with speaker labels and meeting notes. Poll status and export documents.

Upload an audio file, get a job id, and fetch the transcript when it is ready. Large files run asynchronously: the upload answers 202 immediately.

Two modes: transcript returns only the text. meeting also creates a title and summaries, like the meeting view in the app.

POST creates the job and answers before recognition finishes. Poll GET with the job id until status is completed or failed.

Authenticate with Authorization: Bearer YOUR-API-KEY or with a signed-in session. A key acts as the person who owns it.

Start at two seconds and back off to at most thirty. While status is processing or pending, the transcript is not there yet.

Parameter	Type	Default	Description
`id`	`uuid`	-	UUID of the job. id and job_id are the same identifier.
`mode`	`"transcript" \| "meeting"`	-	transcript (text only) or meeting (text plus notes).
`status`	`"processing" \| "pending" \| "completed" \| "failed"`	-	processing, pending, completed or failed.
`text`	`string \| null`	-	The finished transcript. With anonymization, the redacted text, never the raw original.
`segments`	`array \| null`	-	Chunks with time and optional speaker. Only when requested and not anonymized.
`summaries`	`array`	-	Only in meeting mode: prepared notes per template.
`error_code`	`"AUDIO_TOO_LONG" \| "UPLOAD_INCOMPLETE" \| "AUDIO_UNREADABLE" \| "ASSEMBLY_FAILED"`	-	Only with failed and only when the reason is known: AUDIO_TOO_LONG, UPLOAD_INCOMPLETE, AUDIO_UNREADABLE or ASSEMBLY_FAILED. ASSEMBLY_FAILED means the parts could not be put together, even after several tries. Upload the recording again.

1curl "https://app.anymize.ai/api/v1/transcriptions" \
2  -H "Authorization: Bearer YOUR_API_KEY"

Create a job

Upload a file and start recognition. The answer is 202 with the job id.

Upload audio and start transcription

multipart/form-data with the file and optional switches. Fetch the transcript later via GET.

POST/api/v1/transcriptions

Request: multipart/form-data

Body

Parameter	Type	Default	Description
`file`required	`File`	-	The audio file. MP3, WAV, M4A, OGG, WEBM.
`mode`	`"transcript" \| "meeting"`	`transcript`	transcript or meeting. Default transcript.
`speaker_detection`	`boolean`	-	Speaker detection. Default on for meeting, otherwise off.
`timestamps`	`boolean`	-	Chunks with milliseconds. Always on when speaker detection is on.
`language`	`string`	-	ISO-639-1, or omit / auto to detect after transcription.
`prompt`	`string`	-	Short hint for recognition, at most 1000 characters.
`context_bias`	`string`	-	Comma-separated terms so names are spelled right. At most 100 tokens.
`anonymize`	`boolean`	`false`	Then send through anonymization. Supported: de, en, fr, es, it.

202Accepted

{
  "id": "11111111-1111-4111-8111-111111111111",
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "processing",
  "mode": "transcript",
  "title": "standup.m4a",
  "filename": "standup.m4a",
  "message": "Audio submitted for transcription. Poll GET /api/v1/transcriptions/:id until status is completed or failed."
}

401	`UNAUTHORIZED`	No valid key and no session.
400	`NO_FILE`	The form is missing the file field.
400	`INVALID_FORMAT`	Unsupported audio format.
413	`FILE_TOO_LARGE`	The file is larger than 100 MB. Upload larger recordings in parts.
429	`RATE_LIMIT`	Too many uploads in this minute.

1curl -X POST "https://app.anymize.ai/api/v1/transcriptions" \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -F "file=@standup.m4a" \
4  -F "mode=transcript" \
5  -F "language=de"

A single request carries up to 100 MB, the same for transcripts and meetings. Larger recordings up to 1 GB go up in parts (see below).

Upload large recordings in parts

One request carries at most 100 MB, and big requests like to break off on shaky lines. For longer recordings there is the upload in parts: up to 1 GB and 6 hours, and in the end you still get exactly one job with one transcript.

How it works: open a session, send the parts in order with Content-Range, then call complete. You get the same job ID as with the single upload and poll its status via GET /api/v1/transcriptions/:id.

Send the parts strictly one after another, never in parallel. The server only accepts the part that continues exactly where the last one ended, and otherwise answers 409 with the byte to continue from. If the connection drops, ask for the state via GET and pick up from there.

Limits: up to 1 GB per file, parts up to 8 MiB, up to 6 hours of audio. A longer recording ends as failed with error_code AUDIO_TOO_LONG. An open session expires one hour after its last part.

Open an upload session

JSON with file name, size and type, plus the same options as the single upload. Format, size and language are checked right here, before a single byte is on its way. The response names sessionId, chunkSize and nextExpectedRanges.

POST/api/v1/transcriptions/uploads

Request: application/json

Body

Parameter	Type	Default	Description
`filename`required	`string`	-	The recording's file name, e.g. meeting.wav. It becomes the job's title.
`size`required	`integer`	-	The size of the whole file in bytes, at most 1 GB.
`mimeType`	`string`	-	The MIME type, e.g. audio/wav. Same formats as the single upload.
`mode`	`"transcript" \| "meeting"`	`transcript`	mode, speaker_detection, timestamps, language, prompt, context_bias and anonymize work as in the single upload, here as JSON fields.

201Created

{
  "sessionId": "22222222-2222-4222-8222-222222222222",
  "uploadUrl": "/api/v1/transcriptions/uploads/22222222-2222-4222-8222-222222222222",
  "chunkSize": 8388608,
  "nextExpectedRanges": [
    "0-"
  ],
  "expiresAt": "2026-09-24T10: 15: 00.000Z",
  "maxBytes": 1073741824
}

401	`UNAUTHORIZED`	No valid key and no session.
400	`INVALID_BODY`	Not a JSON object, or filename or size is missing or invalid.
400	`INVALID_FORMAT`	Not a supported audio format (MP3, WAV, M4A, OGG, WEBM).
413	`FILE_TOO_LARGE`	The file is larger than 1 GB.
429	`RATE_LIMIT`	Too many uploads in this minute.
503	`STORAGE_UNAVAILABLE`	Storage is not answering right now. Send the same request again in a moment.

1curl -X POST "https://app.anymize.ai/api/v1/transcriptions/uploads" \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{"filename":"workshop.wav","size":734003200,"mimeType":"audio/wav","mode":"meeting"}'

Upload one part

The raw part as the body, with Content-Range: bytes start-end/total and Content-Length. The 202 response names the next expected byte in nextExpectedRanges. You can safely send a part that was already accepted again.

PUT/api/v1/transcriptions/uploads/:sessionId

Request: application/octet-stream

202Accepted

{
  "sessionId": "22222222-2222-4222-8222-222222222222",
  "receivedBytes": 8388608,
  "nextExpectedRanges": [
    "8388608-"
  ],
  "expiresAt": "2026-09-24T10: 16: 02.000Z"
}

400	`INVALID_RANGE`	Content-Range is missing, malformed or does not match the body's length.
400	`UPLOAD_INCOMPLETE`	The body arrived shorter than its Content-Length, the connection broke off. Send the same part again.
401	`UNAUTHORIZED`	No valid key and no session.
404	`SESSION_NOT_FOUND`	This session does not exist or belongs to another account.
409	`RANGE_MISMATCH`	The part does not continue without a gap. nextExpectedRanges names the byte to continue from.
410	`SESSION_EXPIRED`	The session has expired and its parts are gone. Start the upload again.
411	`LENGTH_REQUIRED`	Content-Length is missing. Without a length the server accepts no part.
413	`FILE_TOO_LARGE`	The part is larger than the session's chunkSize.
503	`STORAGE_UNAVAILABLE`	Storage is not answering right now. Send the same request again in a moment.

1# One part at a time, in order. Here: the second 8 MiB part.
2curl -X PUT "https://app.anymize.ai/api/v1/transcriptions/uploads/22222222-2222-4222-8222-222222222222" \
3  -H "Authorization: Bearer YOUR_API_KEY" \
4  -H "Content-Type: application/octet-stream" \
5  -H "Content-Range: bytes 8388608-16777215/734003200" \
6  --data-binary @part-001.bin

Check a session's state

How many bytes arrived and where to continue. Once the upload is completed, the jobId shows up here too.

GET/api/v1/transcriptions/uploads/:sessionId

200OK

{
  "sessionId": "22222222-2222-4222-8222-222222222222",
  "filename": "workshop.wav",
  "sizeBytes": 734003200,
  "receivedBytes": 83886080,
  "nextExpectedRanges": [
    "83886080-"
  ],
  "chunkSize": 8388608,
  "expiresAt": "2026-09-24T10: 25: 40.000Z"
}

404	`SESSION_NOT_FOUND`	This session does not exist or belongs to another account.
410	`SESSION_EXPIRED`	The session has expired and its parts are gone. Start the upload again.

Abort a session

Cancels the upload and removes all parts right away. Without this call we clean them up ourselves one hour after the last part.

DELETE/api/v1/transcriptions/uploads/:sessionId

200OK

{
  "success": true
}

404	`SESSION_NOT_FOUND`	This session does not exist or belongs to another account.
409	`SESSION_COMPLETED`	The session already became a job and can no longer be aborted.

Complete the upload

Once all parts are up: creates the job and answers 202 with its ID right away. Assembling, converting and transcribing run in the background afterwards. Calling it again returns the same job.

POST/api/v1/transcriptions/uploads/:sessionId/complete

202Accepted

{
  "id": "11111111-1111-4111-8111-111111111111",
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "processing",
  "mode": "meeting",
  "title": "workshop.wav",
  "filename": "workshop.wav",
  "message": "Audio submitted for transcription. Poll GET /api/v1/transcriptions/:id until status is completed or failed."
}

404	`SESSION_NOT_FOUND`	This session does not exist or belongs to another account.
409	`UPLOAD_INCOMPLETE`	Parts are still missing. nextExpectedRanges shows where to continue.
410	`SESSION_EXPIRED`	The session has expired and its parts are gone. Start the upload again.

1curl -X POST "https://app.anymize.ai/api/v1/transcriptions/uploads/22222222-2222-4222-8222-222222222222/complete" \
2  -H "Authorization: Bearer YOUR_API_KEY"

Fetch a job

Status, text, segments and, for meetings, the notes. Audio bytes are not included.

One transcription job

Returns the current state. Poll until status is completed or failed. text is null while it runs.

GET/api/v1/transcriptions/:id

200Completed transcript

{
  "id": "11111111-1111-4111-8111-111111111111",
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "completed",
  "mode": "transcript",
  "title": "standup.m4a",
  "filename": "standup.m4a",
  "language": "de",
  "words": 12,
  "duration_ms": 18400,
  "has_audio": true,
  "speaker_detection": false,
  "anonymize": false,
  "anonymization_applied": false,
  "text": "Guten Morgen. Lasst uns mit dem Stand-up beginnen.",
  "segments": [
    {
      "start_ms": 0,
      "end_ms": 18400,
      "speaker": "",
      "text": "Guten Morgen. Lasst uns mit dem Stand-up beginnen."
    }
  ],
  "created_at": "2026-09-11T08: 15: 00.000Z"
}

200Completed meeting

{
  "id": "11111111-1111-4111-8111-111111111111",
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "completed",
  "mode": "meeting",
  "title": "Q3 Planung",
  "filename": "meeting.wav",
  "language": "de",
  "words": 842,
  "duration_ms": 3120000,
  "has_audio": true,
  "speaker_detection": true,
  "anonymize": false,
  "anonymization_applied": false,
  "text": "(((Speaker 1))) Guten Morgen. (((Speaker 2))) Danke, dass ihr da seid.",
  "segments": [
    {
      "start_ms": 0,
      "end_ms": 2100,
      "speaker": "Speaker 1",
      "text": "Guten Morgen."
    },
    {
      "start_ms": 2200,
      "end_ms": 5400,
      "speaker": "Speaker 2",
      "text": "Danke, dass ihr da seid."
    }
  ],
  "summaries": [
    {
      "template_key": "general",
      "title": null,
      "content": "Stand-up zur Q3-Planung. Naechster Schritt: Budget bis Freitag.",
      "status": "completed",
      "locale": "de"
    }
  ],
  "created_at": "2026-09-11T08: 15: 00.000Z"
}

401	`UNAUTHORIZED`	No valid key and no session.
404	`NOT_FOUND`	No audio job with this id, or it is not yours.

1curl "https://app.anymize.ai/api/v1/transcriptions/11111111-1111-4111-8111-111111111111" \
2  -H "Authorization: Bearer YOUR_API_KEY"

The recording streams via GET /api/v1/jobs/:id/audio, not this response.

processing and pending mean: no text yet. failed means the job is done, without a result.

List jobs

Your audio transcriptions, without the full text. Fetch that with GET on the individual id.

List transcriptions

Audio jobs only. Filter by mode and status. No text and no segments, so the list stays small.

GET/api/v1/transcriptions

Query

Parameter	Type	Default	Description
`mode`	`"transcript" \| "meeting"`	-	Only transcript (kind input or live) or only meeting.
`status`	`string`	-	processing, pending, completed or failed.
`limit`	`integer`	`50`	Count per page, at most 100.
`offset`	`integer`	`0`	Offset for the next page.

200OK

{
  "data": [
    {
      "id": "11111111-1111-4111-8111-111111111111",
      "job_id": "11111111-1111-4111-8111-111111111111",
      "status": "completed",
      "mode": "transcript",
      "title": "standup.m4a",
      "filename": "standup.m4a",
      "language": "de",
      "words": 12,
      "created_at": "2026-09-11T08: 15: 00.000Z"
    }
  ],
  "total": 1
}

401 UNAUTHORIZED No valid key and no session.

1curl "https://app.anymize.ai/api/v1/transcriptions?mode=transcript&limit=20" \
2  -H "Authorization: Bearer YOUR_API_KEY"

Modes and limits

The mode you pick decides what happens after recognition, not the quality of the text.

transcript stores only the text. No automatic notes, no meeting title.

meeting detects speakers by default, assigns a title and generates the default summary.

Formats: MP3, WAV, M4A, OGG, WEBM. Up to 100 MB in one request, up to 1 GB and 6 hours with the upload in parts. Rate limit: 50 uploads per minute.

anonymize=true sends the same job through the pipeline. GET then returns the redacted text, no original segments.

1curl -X POST "https://app.anymize.ai/api/v1/transcriptions" \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -F "file=@meeting.wav" \
4  -F "mode=meeting" \
5  -F "speaker_detection=true" \
6  -F "timestamps=true"

Alias /api/transcribe

The older endpoint stays. It starts the same engine and answers with the historical 202 shape so existing clients can keep polling /api/status/:id.

Legacy upload, same engine

API key only, no session cookie. The same form fields as POST /api/v1/transcriptions, including mode.

POST/api/transcribe

Request: multipart/form-data

Body

Parameter	Type	Default	Description
`file`required	`File`	-	The audio file. MP3, WAV, M4A, OGG, WEBM.
`mode`	`"transcript" \| "meeting"`	-	transcript or meeting. Default transcript.
`speaker_detection`	`boolean`	-	Speaker detection. Default on for meeting, otherwise off.
`timestamps`	`boolean`	-	Chunks with milliseconds. Always on when speaker detection is on.
`language`	`string`	-	ISO-639-1, or omit / auto to detect after transcription.
`prompt`	`string`	-	Short hint for recognition, at most 1000 characters.
`context_bias`	`string`	-	Comma-separated terms so names are spelled right. At most 100 tokens.
`anonymize`	`boolean`	`false`	Then send through anonymization. Supported: de, en, fr, es, it.

202Accepted

{
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "processing",
  "message": "Audio submitted for transcription. Use /api/status/{job_id} to check status."
}

1curl -X POST https://app.anymize.ai/api/transcribe \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -F "file=@standup.mp3" \
4  -F "language=de"

The job id in job_id works for GET /api/status/:id and for GET /api/v1/transcriptions/:id.

Job Status

Endpoint

Check the current status of an asynchronous job.

GET/api/status/{jobId}

Parameters

Parameter	Type	Default	Description
`jobId`required	`string`	-	The job ID you received when creating the job.

Status values

A job goes through the following states:

processingThe job is currently being processed.

completedThe job is complete — result is available.

failedThe job failed. The response contains the status but no separate error details.

Status: Processing

200Response during processing

{
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "processing"
}

Status: Completed

When the job is complete, the response contains the result.

200Response for a completed job

{
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "completed",
  "anonymized_text_raw": "Max [[Person-A1B2C3]] lives in [[Location-D4E5F6]].",
  "original_text": "Max Mustermann lives in Berlin.",
  "metadata": {
    "filename": "document.pdf",
    "language": "en",
    "words": 6
  }
}

Status: Failed

200Response for a failed job

{
  "job_id": "11111111-1111-4111-8111-111111111111",
  "status": "failed"
}

Polling strategy

Query the status at regular intervals until the job is finished.

Tip

Start with a 2-second interval and increase exponentially up to 30 seconds.

1# Poll every 3 seconds until completed
2while true; do
3  RESULT=$(curl -s https://app.anymize.ai/api/status/11111111-1111-4111-8111-111111111111 \
4    -H "Authorization: Bearer YOUR_API_KEY")
5  STATUS=$(echo $RESULT | jq -r '.status')
6  echo "Status: $STATUS"
7  if [ "$STATUS" = "completed" ] || [ "$STATUS" = "failed" ]; then
8    echo $RESULT | jq .
9    break
10  fi
11  sleep 3
12done

Document Export

Endpoint

Export completed transcriptions in various formats.

GET/api/v1/jobs/{jobId}/export

Parameters

Specify job ID, format, and optionally a variant.

Parameter	Type	Default	Description
`jobId`required	`string`	-	The job ID of the completed transcription.
`format`	`string`	`txt`	Output format: pdf or txt. docx returns HTTP 501.
`textType`	`string`	`anonymized`	Export variant (e.g. anonymized or original).

Available formats

pdfPDF — Formatted document with layout and metadata.application/pdf

txtTXT — Plain text without formatting.text/plain

Response headers

The download includes appropriate Content-Type and Content-Disposition headers.

200Example response headers

Content-Type: application/pdf
Content-Disposition: attachment; filename="document_anonymized.pdf"
Content-Length: 245832

Export variants

Choose between the original and anonymized version.

Tip

Use the anonymized variant when sharing with third parties.

Example requests

Export as PDF

1curl -OJ https://app.anymize.ai/api/v1/jobs/11111111-1111-4111-8111-111111111111/export?format=pdf \
2  -H "Authorization: Bearer YOUR_API_KEY"

Error cases

DOCX returns 501. Missing jobs return 404, and incomplete jobs return 400.

501DOCX unavailable (HTTP 501)

{
  "error": "DOCX export not yet implemented"
}

404Not Found

{
  "error": "Job not found"
}