eleven_turbo_v2_5 deprecated: use Flash; Sume TTS Router ids

The replacement for eleven_turbo_v2_5 is eleven_flash_v2_5, per the ElevenLabs models page: its deprecated table lists Flash as the suggested model, and says Flash is functionally equivalent with lower latency on average. On the Sume side the TTS Router takes an explicit catalog id, and TTS 1.0 has no engine picker, so there is no vendor model id to migrate.

Vendor facts are from the ElevenLabs models page read 2026-10-01; Sume facts from the OpenAPI schema in the API reference.

What did ElevenLabs deprecate?

The page lists three deprecated ids with a replacement each. It says it recommends the Flash models over Turbo in all use cases.

How does Sume name speech models?

TTS 1.0 has no engine picker: you send a transcript and a voice selector and Sume chooses the engine. The TTS Router is the explicit path. Its required model field is an id from GET /v1/tts-router/models, and the schema lists sonic-3.6, sonic-3.5, sonic-3, sonic-latest and sonic-preview. The field is not accepted on TTS 1.0.

An unknown id fails with 400 model_not_found and the catalog URL, so a retired or misspelled id surfaces as a typed error instead of a silent swap.

What should I change when an id is retired?

Treat the model id as configuration, not a literal in code. Read the router catalog, pick an id, and store it in settings. Do not send ElevenLabs ids such as eleven_flash_v2_5 to Sume; they are not in the router's list. For the same idea on speech-to-text, see a stable model id for STT.

const res = await fetch("https://api.sume.com/v1/tts-router/models", {
  headers: { Authorization: "Bearer " + process.env.SUME_API_KEY },
});
const catalog = await res.json();
console.log(JSON.stringify(catalog, null, 2));