Scribe v2 Realtime $0.39 vs batch $0.22: 37 hours vs Sume STT

ElevenLabs lists Scribe v2 at $0.22 per hour and Scribe v2 Realtime at $0.39 per hour, a premium of about 77 percent ($0.39 / $0.22 = 1.77). For 37 hours of meetings that is $14.43 in realtime and $8.14 in batch. Sume STT would be $22.20 for the same 37 hours ($0.01 x 60 x 37).

Numbers come from ElevenLabs' API pricing page and Sume's public catalog, both read on 2026-10-09.

The three prices side by side

Scribe v2 and Scribe v2 Medical share the $0.22 hourly rate on the ElevenLabs page. Realtime is a separate line at $0.39. Add-ons are priced apart: keyterm prompting is $0.050 per hour and transcript editing is $0.070 per hour.

Sume's catalog lists one STT price: $0.01 per audio minute. That is $0.60 per hour.

When realtime is worth the extra

Realtime exists so that text appears while someone is speaking. If the only thing you do with the transcript is read it after the meeting, the batch price is the better deal, and you save about $6.29 across these 37 hours ($14.43 - $8.14). If a person watches captions live or a voice agent reacts mid-sentence, batch cannot do that job.

Sume sits on the batch side of this split. Its STT job takes a public HTTPS audio URL and reserves up to 10 minutes of audio, so a 60-minute meeting is six jobs, $0.60 in total.

Before you choose

Run a short sample of your own meetings through each option. Compare the transcripts and the cleanup time, not just the price, since a cheaper transcript that needs an hour of fixing is not cheaper. If you later need captions burned into a replay clip, Sume's word timings feed its caption job directly, which is a reason to keep the audio in one place.

• Need live text: Scribe v2 Realtime or another streaming service.
• Need after-the-fact text: batch pricing is lower on this page.
• Need transcript plus captioned video: Sume keeps both in one workflow at a higher per-hour rate.

What the add-ons do to the comparison

The keyterm line in the table assumes you turn on keyterm prompting at $0.050 per hour, which raises Scribe v2 from $0.22 to $0.27 per hour. Over 37 hours that is $1.85 extra ($0.05 x 37). It is only worth paying if your meetings are full of product names or jargon that the base model gets wrong.

Sume's STT request takes a language hint but no keyterm list, so there is no equivalent line to add. If your audio is full of terms the model misses, plan for a correction pass on the Sume transcript, or use the caption job's script_text option when you already have the intended wording and only need it aligned to the audio.