Cartesia Ink STT hours per plan vs Sume STT $0.60 per hour
Cartesia plans include about 9 to 741 hours of Ink speech-to-text. That works out near $0.40 to $0.54 an hour; Sume STT is $0.60 an hour. Concurrency compared.

Cartesia includes Ink speech-to-text hours in its monthly plans: about 1 hour 51 minutes on Free, 9 hours 16 minutes on Pro ($5), 115 hours 44 minutes on Startup ($49) and 740 hours 44 minutes on Scale ($299). Dividing price by hours gives roughly $0.54, $0.42 and $0.40 per hour. Sume STT 1.0 is $0.01 per audio minute, or $0.60 per hour, with no plan.
Plan figures
Hours and concurrency are quoted from Cartesia's page. The derived rate is plan price divided by included hours, and it only applies if you actually use all included hours.
| Plan | Price | Included Ink STT | Concurrent STT requests | Derived $ per hour |
|---|---|---|---|---|
| Free | $0 | About 1h 51m | 8 | n/a |
| Pro | $5 | About 9h 16m | 12 | About $0.54 |
| Startup | $49 | About 115h 44m | 20 | About $0.42 |
| Scale | $299 | About 740h 44m | 60 | About $0.40 |
Sume STT in the same terms
A Sume STT job takes a public HTTPS audio_url, optional language_code and duration_seconds from 1 to 600, then returns text and words[]. There is no plan: you pay $0.01 per audio minute for what you run, so unused capacity costs nothing.
import os, requests
r = requests.post(
"https://api.sume.com/v1/stt-1.0/transcribe",
headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
"Idempotency-Key": "stt-demo-001"},
json={"audio_url": "https://example.com/call.mp3",
"language_code": "en",
"duration_seconds": 300},
)
print(r.status_code, r.json())Reading the trade-off
At steady high volume, a Cartesia plan used fully is cheaper per hour. At low or spiky volume, an unused plan hour is wasted spend, and pay-per-minute can come out lower overall. Compute your own monthly hours before choosing.
Worked example
Here is the same exercise for transcription.
- Need 100 hours a month: Startup's about 115h 44m covers it at $49.
- Need 10 hours: Pro's about 9h 16m falls just short.
- At Sume's $0.01 per minute, 10 hours is $6.00, and 100 hours is $60.00.
Checklist before you commit
Cartesia's per-hour figure is only reached if you use the whole plan. Pay-per-minute has no such threshold.
- Check concurrency (Startup lists 20 STT requests).
- Check latency needs, since Cartesia also targets live use.
- Compare accuracy on your own audio.
Sources
Related posts
More in Comparisons
- Cheapest 720p AI video per second: Veo, xAI, Sume
At 720p, Veo 3.1 lists $0.05 to $0.40 per second, xAI lists grok-imagine-video at $0.05, and Sume lists Grok Imagine Video 1.5 at $0.0125. Read 2026-10-01.
- Claude directory drops MCPB: remote vs local server, and Sume
Claude's directory no longer accepts MCPB desktop extensions. A remote HTTPS server needs no package; here is what that means for Sume.
- Claude Message Batches 100,000 requests vs a Sume bulk run of 100
A Claude Message Batch holds 100,000 requests or 256 MB. A Sume bulk run queues 1 to 100 Format runs with a concurrency window of 1 to 16. Plan accordingly.
- Claude batch canceled or expired results: billed? Sume cancel billing
Claude does not bill canceled or expired batch requests. Sume bills generation a run finished before a cancel, so a cancel stops spend but never refunds.
Written by Sume