Cartesia Ink at $0.39 an hour vs ElevenLabs Scribe v2 at $0.22

Cartesia's pricing page lists Ink speech-to-text at $0.39 an hour on the Scale plan. ElevenLabs lists Scribe v2 at $0.22 an hour, so Ink costs about 1.77 times as much (0.39 / 0.22). ElevenLabs' streaming model, Scribe v2 Realtime, is also listed at $0.39 an hour, which matches Ink's price exactly.

Listed prices

Scribe v2 is the pre-recorded model on ElevenLabs' page, and Scribe v2 Realtime is the streaming one, listed at about 150 ms. Cartesia's page gives one Ink price and does not split it into batch and streaming rows, so the sources do not settle which ElevenLabs row is the closer match on features. Check the plan details on the page.

What Cartesia's page ties to the price

Cartesia's pricing page prices a lot beyond Ink, and the Ink price depends on the plan.

• Plans: Free $0 with 20K credits, Pro $5 with 100K, Startup $49 with 1.25M, Scale $299 with 8M.
• The $0.39 an hour figure is stated for the Scale plan; the page text read for this post gives no Ink price for the other plans.
• Agents are listed at $0.06 a minute.
• Cartesia's blog index mentions Ink-2 with keyterm prompting and configurable turn detection. Keyterm prompting is useful when you transcribe product names and jargon.

Cost of 100 hours

Arithmetic on the per-hour column.

What Sume bills and why Scribe shows up

Sume's Video inspect transcribes a hosted clip when you send transcribe: true, at $0.01 per audio minute, or $0.60 an hour. Sume's pricing docs describe this rate as Scribe's list price times 1.25. That does not reconcile with the $0.22 an hour listed on 2026-10-05 ($0.0037 a minute times 1.25 is about $0.0046), so budget from the live rate in the Sume catalog, not from this comparison. Sume does not ship Cartesia Ink or live transcription. It does use Cartesia Sonic for text-to-speech, so a project can pair Sonic voiceover with Sume's clip transcription.

If you need keyterm prompting or live turn detection, use Cartesia or a streaming vendor directly. If you want cheapest pre-recorded transcripts, the table above says Scribe v2, but test it on your audio first, since none of the sources gives accuracy figures.