Speech-to-text audio too large? Sume STT takes up to 10 MB, hosted
Sume's stt_create needs a public HTTPS audio URL on the Sume media host, 10 MB at most. What fits: 16 kHz mono wav versus mp3, and how to cut a long file.
Sume's stt_create tool takes a public HTTPS audio_url, preferably on the Sume media host, and the tool schema caps the file at 10 MB; a non-Sume host fails with unsupported_audio_source. A 16 kHz mono 16-bit wav is about 32 KB per second, so 10 MB holds a little over five minutes of wav, and roughly twice that as 128 kbps mp3. The words always come back in words[], and speech-to-text is listed at $0.01 per audio minute in the catalog.
Size by format
Sume's audio detach guide calls 16000 plus channels: "mono" the speech-to-text shape and says the default wav is pcm_s16le. That is 16,000 samples a second at 2 bytes each. Its mp3 option is 128 kbps, which is 16,000 bytes a second. These are calculated from those published figures, not measured.
| Format | Bytes per second | Approximate length in 10 MB |
|---|---|---|
| wav, 16 kHz mono, 16-bit | 32,000 | About 5 minutes |
| mp3, 128 kbps | 16,000 | About 10 minutes |
| wav, 48 kHz stereo, 16-bit | 192,000 | About 52 seconds |
Cutting a long recording
Detach the audio of a long video once, then split it into ranges with timeline audio, which returns up to 20 ranges per job at $0.01 per job and keeps each as its own durable file. Transcribe each range separately and add the range's start time to its word times when you stitch the results back together.
Audio detach itself returns at most 900 seconds per job, so a recording longer than 15 minutes needs a range anyway.
Sources
Related posts
More in Developers
- Split one voiceover into 40 scene tracks: two timeline audio jobs
Timeline audio split takes up to 20 ranges per job. Forty scene tracks from one voiceover is two split jobs, $0.02, with ranges built from cue times in Python.
- Spotify playlist cover upload: base64 JPEG under 256 KB from Sume
Spotify's playlist cover endpoint takes base64 JPEG data, 256 KB max. Make a 1:1 image on Sume, shrink it in Pillow, check the encoded size, then PUT it.
- Spring AI MCP client request-timeout 20s vs Sume jobs_wait
Spring AI's MCP client defaults request-timeout to 20s, shorter than a 50s Sume jobs_wait. Raise it with a customizer or keep waits short and re-issue them.
- Store the Idempotency-Key before you submit: a SQLite intent table
Write key and body to SQLite, submit with that key, then store the job id. A crash between steps replays the same key and returns the original Sume job.
Written by Sume