Which Sume speech endpoint: TTS 1.0, TTS Router, STT or Music?

Use POST /v1/tts-1.0/generate for a managed voice, POST /v1/tts-router/generate when you must name a Sonic model, POST /v1/stt-1.0/transcribe to turn a recording into text, and POST /v1/music-router/generate for a music bed. All four are async jobs with the same status and result URLs. None streams.

The four endpoints

Pick by the question you are asking

• "I want a voice reading my script." TTS 1.0. It always runs sume/tts-1.0 and has no engine picker; sending model returns 400.
• "I need a specific Sonic version, pinned." TTS Router, with a model from GET /v1/tts-router/models.
• "I have a recording and need words and times." STT 1.0. Omit language_code for auto-detect.
• "I need a track under a scene." Music Router, with sume/music-auto, lyria-3.5 or lyria-3-pro.

What none of them do

The TTS Router docs list its non-goals: Eleven and OpenAI engines, streaming TTS, and a routing preset. STT is a file job, not a live stream. Music takes no duration or negative prompt, so length and exclusions go in the brief.

STT can appear on the dev API before the production OpenAPI snapshot lists it, so confirm the path on the current reference before you build on it.

The same shape for all four

Every one returns a job envelope. Poll status_url until result_ready is true, then read result_url. mode: sync waits at most 30 seconds; a timed-out wait means keep polling, not resubmit. Send an Idempotency-Key on every submit.

Over MCP the same families are tts_create, stt_create and music_create, with tts_source_get and tts_source_verify_spine as free reads.

curl https://api.sume.com/v1/jobs/job_123/status \
  -H "Authorization: Bearer $SUME_API_KEY"

curl https://api.sume.com/v1/jobs/job_123/result \
  -H "Authorization: Bearer $SUME_API_KEY"