Speech-to-text price per audio hour: xAI, OpenAI, Sume

Per audio hour, xAI lists speech to text at $0.10 over REST, OpenAI's gpt-4o-mini-transcribe at $0.18 and gpt-4o-transcribe at $0.36. Sume's video_inspect transcript costs $0.01 per audio minute, or $0.60 per hour, which is the highest of the five. Sume's value is that the transcript comes with the video work it belongs to, not that it is cheapest.

Rates per hour

Per-minute rates times 60. Vendor rows are public list prices.

Where Sume fits

video_inspect takes a video, returns a transcript along with frames and timing, and bills $0.01 per audio minute. The default reserve is one minute and the maximum duration hint is 600 seconds, so it is built for clips, not for a day of recordings.

If all you need is a transcript of long audio at the lowest cost, the vendor rows above are cheaper. If you need to inspect a clip you generated or received and then act on it, one call and one wallet is the point.

• Long audio: use a dedicated speech-to-text model.
• Short clips with visual context: use video_inspect.
• Always compare accuracy on your own audio, not just price.

A concrete month

Take 200 hours of recordings in a month. xAI REST is $20, OpenAI mini $36, OpenAI 4o $72 and Sume video_inspect $120. The spread is $100 a month, which is a real cost but a small one beside most production budgets.

If those same 200 hours were 12,000 one-minute clips that you also want to inspect visually, the single-call value of a clip-level tool is likelier to outweigh the difference.

Limits

Accuracy, language coverage, diarization and latency differ among these services and are not compared here. The tables show the model amount. A workspace Agent Fee line can be added on top, so treat the job's own usage cost as the truth and check the live catalog before you budget; limits and rows change.