Docs / Providers
Cartesia (Sonic)
In one sentence: a low-latency-focused brand whose voice lists come with free preview audio; pricing sits in the upper-middle range.
In one sentence: a low-latency-focused brand whose voice lists come with free preview audio; pricing sits in the upper-middle range.
What This Is
Cartesia's Sonic speech model is characterized by extremely low latency — nearly no wait from request to first sound. It displays as Cartesia International in Soniva.
Selectable Models & Pricing
| Model | Credits / 1,000 chars | Notes |
|---|---|---|
| sonic-3.5 | 58 | Latest; recommended |
| sonic-3.5-2026-05-04 | 58 | A dated snapshot of 3.5; won't change when official updates arrive |
| sonic-3 | 58 | Previous generation |
| sonic-latest | 58 | Always points to the official latest (may change anytime) |
Official upstream price: 1 credit / char, Pro plan from $5 / 100k credits. Price source: https://www.cartesia.ai/pricing (verified 2026-08-06)
All four models are the same price, so whichever you pick doesn't affect the charge — just use
sonic-3.5.The difference between
sonic-latestandsonic-3.5-2026-05-04is "whether to auto-upgrade with the official release": the former changes the moment official updates, the latter stays forever that day's version. If you care that "what I hear today sounds the same tomorrow," pick the dated one.
About This Price
Soniva converts at Cartesia's smallest usable paid plan (Pro) unit price — a conservative pick, because Cartesia's Startup / Scale large-customer plans have lower unit prices (~¥269/M vs ~¥360/M). Once Soniva's actual volume reaches a higher tier, this price will be adjusted down.
Voices & Free Previews
Cartesia's voice list is pulled live from upstream by Soniva; this document doesn't hard-code a count.
Like ElevenLabs, Cartesia voices come with official pre-made preview audio — click and it plays directly, without going through Soniva's backend, no credits deducted, no count limit, every voice playable. So it isn't on the "5 free previews per day" whitelist, because it simply doesn't need to be.
One Technical Note
Cartesia's speech-rate control uses an official experimental field
(__experimental_controls.speed), with a different value scheme than other brands' "multipliers"
(it's a relative value of -1 to 1, not a 0.75x / 1.5x multiplier).
Soniva converts the speed tier you select in the UI into this range, but because it's an experimental interface, extreme speeds (0.75x or 2x) may be less stable than other brands. If you notice odd speed behavior, returning the speed to 1x usually fixes it.
When to Choose Cartesia
| Scenario | Recommended? |
|---|---|
| Want to hear all voices free before deciding | ✅ Every voice is directly playable |
| English content, want natural but not ElevenLabs' price | ✅ 58 vs 105 — nearly half the price |
| Long Chinese text | ⚠️ 58 credits/1,000 chars is pricey; domestic brands are better value |
| Need precise speech-rate control | ⚠️ See the experimental-interface note above |
Related Documents
- Full Credit Rate Table
- ElevenLabs (the other brand with built-in free previews)
- Eight Brands at a Glance
Was this article helpful?
Anonymous feedback is enabled when you allow analytics and the service is available. You can also report an issue by email.
The email includes this article. You can also copy the details and send them to soniva.cde@gmail.com