Sume TTS Router: send sonic-3.6, sonic-latest or sonic-preview?
The Sume TTS Router has five Sonic ids. sonic-3.6 is stable, sonic-latest resolves to it, and sonic-preview is a beta that rejects pro voice clones.

Send sonic-3.6 when you want a pinned, stable model, sonic-latest when you want to follow the newest stable one, and sonic-preview only for testing. In Sume's TTS Router, sonic-latest is an alias that resolves to sonic-3.6 today, and sonic-preview is a beta channel that can change without notice.
All five ids are Cartesia Sonic pass-through rows. They share the TTS 1.0 character price, and the router does not include other vendors in v1.
The five ids
From the Sume TTS Router docs.
| Id | Status | Notes |
|---|---|---|
| sonic-3.6 | Current stable | The model TTS 1.0 uses |
| sonic-3.5 | Explicit earlier version | Pin for continuity |
| sonic-3 | Explicit earlier version | Pin for continuity |
| sonic-latest | Alias | Resolves to sonic-3.6 |
| sonic-preview | Beta | Rejects pro voice clones with voice_model_mismatch |
Choosing in practice
Think about what happens when the vendor ships a new model.
- Pin sonic-3.6 for client work where the voice must not drift between months.
- Use sonic-latest for internal drafts, so you hear improvements without a code change.
- Keep sonic-preview out of anything customers hear, because its output and availability can change without notice.
- Keep older ids only if a script has been approved on them and must be reproduced.
Price and limits are the same across ids
Billing is the TTS 1.0 character book: $0.0475 per 1,000 characters (list 38 micro-dollars per character times 1.25), rounded up to the cent per job, with a 20,000-character cap. Read the numbers from GET /v1/tts-router/models rather than hard-coding them.
If a request to sonic-preview returns voice_model_mismatch, the voice is a pro clone. Send the same job to sonic-3.6 instead.
A migration habit
Store the model id with each finished job in your own database. When a provider retires an old Sonic version, you can then list which of your assets used it. Sume's metadata field on the request is kept with the job and is not forwarded to the provider, so it is a good place to write the id and your project name.
Sources
Related posts
More in Models
- Swap the lead in a finished AI episode: H3 Max Recast for 20 s
H3 Max Recast replaces the people in a 5-30 s source with 1-4 photos, keeping motion and sound. A 20 s source is $7.50 at 768p and $11.25 at 1080p on Sume.
- MiniMax H3 Max on Sume: video, lip-sync and recast, which id to call
Three Sume ids carry the MiniMax H3 Max name: minimax-h3-max for video, a lip-sync route and h3-max-recast. What each takes, its window and its rate.
- Two reference images, one clip: Omni's cat-and-yarn pattern on Sume
Google's docs show two images, a cat and yarn, producing one video. Here is the Sume request with IMAGE_REF tokens and what changes with a brand product.
- Video edit prompt: say what stays, then what changes (Omni Flash)
A prompt pattern for Gemini Omni Flash 1.1 video edit on Sume: one change per pass, an explicit keep clause, and what the video_url route fixes for you.
Written by Sume