Async File Transcription

TL;DR: Click the upload icon in the header to transcribe an existing recording. Drop in a file (up to 2 GB, 5 hours) or paste a public URL. The file transcribes in the background — you can close the tab and come back; the session shows up in your sidebar when it's done. Credits are only charged on success; failed jobs are refunded automatically.

When to use async transcription

Use async transcription when you already have a recording and don't need the live-transcript experience. Common cases:

• Meeting recordings exported from Zoom, Google Meet, Teams, etc.
• Podcast episodes or interview recordings
• Voice memos from your phone
• Lecture recordings
• Any video file where you want the speech content as text

For recording as it happens, use the Record button on the home page — see Recording & Transcription.

Starting an async transcription

1. Click the upload icon in the top-right of the header (next to the notifications bell)
2. Choose File to upload from your device, or URL to transcribe a publicly-accessible audio/video URL
3. Drop or pick the file (or paste the URL) and click Transcribe
4. The dialog closes and a new session appears in your sidebar with an "In progress" indicator

You can start another recording, open a different session, or close the browser — the job keeps running on the server. A notification fires when it finishes.

Supported formats and limits

URLs must be publicly reachable (no auth headers, no presigned expiry shorter than the transcription window). If you need to transcribe a private file, upload it directly.

How it works

1. The browser requests a signed upload URL from the server
2. The file is uploaded directly to object storage (the server never proxies the bytes)
3. The server hands the storage URL (or your pasted URL) to the transcription engine
4. When the engine finishes, the transcript is attached to the session and a notification is sent

Unlike live recording, async jobs keep the source audio in private storage for the duration of the transcription so the engine can fetch it. It is deleted once the transcript is saved. If you cancel mid-job, the upload is removed immediately.

Credits

Async transcription uses the same rate as live recording (1 credit per minute, rounded up). A few differences:

• You're charged based on the actual audio duration, not wall-clock time
• If the job fails (bad file, engine error, etc.) your credits are refunded automatically
• You can cancel a job while it's running — partial credit is not charged

Cancelling a job

Open the in-progress session in the sidebar and click Cancel. The upload is deleted, the engine job is cancelled if still running, and no credits are deducted.

Troubleshooting

"Unsupported format" — convert the file to one of the formats above. FFmpeg works well: ffmpeg -i input.mov -vn -c:a libmp3lame output.mp3.

"Audio longer than 300 minutes is not supported" — split the file into multiple parts and upload separately.

"Upload failed" / network errors — re-open the dialog and retry; the previous attempt is cleaned up automatically.

URL transcription fails — confirm the URL is reachable from a public browser (open it in an incognito window). Private or presigned URLs with short expiry don't work.

If you run into anything else, see Support.