ElevenLabs Scribe v2 Realtime pricing vs Sume STT jobs

ElevenLabs lists Scribe v2 Realtime at $0.39 per hour, against $0.22 per hour for Scribe v2. Sume does not offer a live speech-to-text stream: Sume STT 1.0 is an async job priced at $0.01 per audio minute, which is $0.60 for an hour of audio.

Vendor figures are from the ElevenLabs API pricing page, read 2026-09-30. Sume figures are from the Video inspect docs and the OpenAPI description of the STT route.

What does the realtime price add up to?

Per-hour prices convert to per-minute by dividing by 60. The table shows only arithmetic on the listed numbers.

What does Sume offer instead of a live stream?

You submit audio and get a job back. With mode: async the response returns right away with status_url and result_url; the finished result carries text and words[] word-level timings in seconds from the start of the audio. Use a signed webhook_url or poll the status URL for completion. See Jobs and results for the lifecycle.

Can I wait on the request instead?

Only briefly. mode: sync and mode: subscribe are aliases for the same bounded wait of up to wait_timeout_seconds, with a maximum of 30. If the job is not done, poll status_url rather than resubmitting. That is a bounded HTTP wait, not a stream of partial transcripts; see mode subscribe is a sync alias.

What are the size limits?

Omit duration_seconds and Sume reserves one minute; the maximum is 10 minutes per request. If your use case is a live call or captions while someone is speaking, this is the wrong shape. If it is recorded audio, an async job fits; the real-time speech-to-text API post covers the boundary.