AI voice isolation API cost per minute: Runway vs ElevenLabs

Voice isolation costs $0.10 per minute on Runway's API (1 credit per 6 seconds at $0.01 per credit) and $0.12 per minute on ElevenLabs' own API page, both read on 2026-10-02. Sume does not list a voice isolator; its nearest tool, audio detach, only pulls the audio track out of a video for $0.01 per job and does not clean it.

Audio clean-up, dubbing and sound effects are usually priced per minute of audio, so a long file is easy to misprice. The tables below convert each published unit to the same per-minute basis.

What do Runway and ElevenLabs list per minute?

Runway lists voice isolation at 1 credit per 6 seconds of audio, at $0.01 per credit. ElevenLabs lists its own API prices in dollars per minute, including voice changer, sound effects and dubbing. The Runway pricing page read for this post was not checked for those other audio rows, so the table leaves them blank rather than guess.

Only the isolation row is like for like. Use the table to bound a budget, then confirm the exact model you will call.

What does a one-hour recording cost to isolate?

At the listed rates, 60 minutes is $6.00 on Runway and $7.20 on ElevenLabs. A 10-minute podcast episode is $1.00 and $1.20. Isolation is the cheap step in a clean-up pipeline; the large numbers appear when you add dubbing: an hour of ElevenLabs v2 dubbing is $132.00 at its listed $2.20 a minute.

Those figures assume one pass. If your source needs two passes, for example isolate then dub, add the rows.

What does Sume offer instead?

Audio detach takes the audio track of one Sume-hosted video and returns a durable wav or mp3. The public rate is $0.01 per job, reserved up front and captured at the pass's own compute, never above the reservation. It supports up to 1,800 seconds of source and 900 seconds of output, with optional mono and 16 kHz output for speech tools.

It does not remove noise, music or other voices. It is a demux. If you need the speech separated from a mix, price a separate tool; Sume has no isolator I could find in its docs.

How do I keep the Sume side cheap and correct?

Detach once per source video, keep the artifact, and reuse it instead of re-detaching. Each job is a flat cent, so the saving is small, but the real saving is avoiding a re-run on the other vendor's per-minute meter. Timeline audio joins or splits audio files for a flat $0.01 per job when the cleaned tracks come back.

Check the live number in the catalog before you automate: the docs point at the generated rate card, not a hand-typed price.

How should I choose between the two vendors?

Pick on the whole pipeline, not the isolation row. If your audio already lives in Runway credits, the $0.10 per minute isolation rate avoids a second account. If you are already on ElevenLabs, the $0.12 rate is the one to use, and the difference on an hour of audio is $1.20.

Dubbing is where the cost grows. The ElevenLabs page lists three dubbing prices, with the v2 tier at $2.20 per minute. Check Runway's current page for its own dubbing rate and which version it calls before you compare.

On Sume, plan the rest of the chain. Narration is text to speech, music is a flat per-generation job, and the final assembly is a Timeline render at $0.10 per minute. Cleaning dialogue is the step you would buy elsewhere.