Sonic 3.5 to 3.6 on Sume TTS: change model, keep the voice id

To move a Sume TTS Router call from Sonic 3.5 to Sonic 3.6, change model from sonic-3.5 to sonic-3.6 and leave the rest of the body alone. Cartesia's Sonic 3.6 page (read 2026-10-02) says 3.6 is fully backwards compatible with 3.5, and its changelog (read 2026-10-02) says to use Sonic 3.6 with the same voice ids, with nothing to recreate.

What does Cartesia say stays the same?

The changelog entry says there are no changes to the feature surface: prompting guidance, SSML tags, timestamps, codec, and speed and volume controls behave as they do on Sonic 3.5. It also says its human evals confirm voices sound the same as on 3.5, which is Cartesia's claim and not something Sume has measured. The 3.6 page adds that pro voice clones created on older models work on 3.6.

What changed is the speech itself: Cartesia lists more natural pacing, 44 languages, and two new ones, Odia and Urdu.

What do I change on Sume?

Only the model value. Voice selection, language, output_format, timestamps and segmentation are shared by every catalog row. Sume TTS 1.0, the surface without an engine picker, already runs on sonic-3.6 according to Sume's router notes, so a move matters only to calls that pin an older id on the router.

curl -X POST https://api.sume.com/v1/tts-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: move-to-36-001" \
  -d '{
    "model": "sonic-3.6",
    "transcript": "Same line, same voice, newer model.",
    "avatar_handle": "@your_avatar",
    "language": "en"
  }'

How do I check the result?

Generate one line on each model with the same settings and listen before you switch a whole project. Then read model_id on the finished job, as described in Jobs and results, to confirm which engine made the file. Cartesia's older models page (read 2026-10-02) lists sonic-3.5-2026-05-04 as Stable, so there is no deadline on 3.5 there.