Export

GET /transcriptions/{id}/export renders the transcript and returns the file in the response body. Generation is synchronous: the call returns when the file is ready.

curl -sS --fail-with-body \
  "https://api.scribewren.com/v1/transcriptions/$TX_ID/export?format=srt" \
  -H "Authorization: Bearer $SCRIBEWREN_API_KEY" \
  -o interview.srt

On success this is the one endpoint that does not use the JSON envelope. The body is the file. On failure you get the normal JSON error envelope, so check the HTTP status before you save the body.

Query parameters

Parameter	Required	Notes
format	Yes	txt, docx, pdf, srt or vtt. Case-insensitive. Anything else is 400 VALIDATION_ERROR with reason invalid_enum.
timestamps	No	true or false. Adds timestamps to txt, docx and pdf. Default false. srt and vtt always carry timestamps and ignore this option.
speakers	No	true or false. Adds speaker names. When you leave it out, it is on if the transcript has more than one speaker, and off otherwise.

timestamps and speakers accept only the exact strings true and false. Any other value is VALIDATION_ERROR with reason invalid_boolean.

Response

Format	Content-Type
txt	text/plain; charset=utf-8
docx	application/vnd.openxmlformats-officedocument.wordprocessingml.document
pdf	application/pdf
srt	application/x-subrip; charset=utf-8
vtt	text/vtt; charset=utf-8

The response also carries these headers:

HTTP/1.1 200 OK
Content-Type: application/x-subrip; charset=utf-8
Content-Disposition: attachment; filename="interview.srt"; filename*=UTF-8''interview.srt
Cache-Control: no-store
X-Content-Type-Options: nosniff

Content-Disposition gives the suggested file name twice: a plain ASCII filename and a percent-encoded UTF-8 filename*. Prefer filename* when your client understands it.

The name is built from the transcription’s title plus the format extension. The characters / \ : * ? " < > | and control characters are removed, and the name is cut to 150 bytes. If you save the file yourself, you can ignore the suggested name and choose your own.

Cache-Control: no-store tells caches and proxies not to keep the file. Exports are generated again on every call, so a call after the transcript changed returns the new text.

Errors

Status	error.code	When
400	VALIDATION_ERROR	format is missing or unknown, or a boolean option is invalid. See error.details.fields.
404	NOT_FOUND	The transcription does not exist or belongs to another account.
409	INVALID_STATE_TRANSITION	The transcription has not succeeded. error.details.reason is not_ready.
429	RATE_LIMITED	You are exporting too fast. Wait for Retry-After seconds. The export policy is listed in Rate limits.
503	EXPORT_FAILED	The file could not be generated. error.details.format names the format. It is safe to retry after a short wait.

Exports are allowed on a restricted account, because they only read data. See Authentication.

Subtitles and long files

A transcription that is truncated exports only the part that was transcribed. Compare durationMs with sourceDurationMs on the transcription before you publish subtitles for a long file.