S1–S7 for system presets and V1–V5 for clones.
Provider codes at a glance
System preset tiers are
S1 through S7, including Cantonese presets on S7.
Copy voiceId and provider together from API responses when you need to resubmit or edit content (especially media translation lines). See Voices.
Where you send provider
Clone engines (V1–V5)
Pass V1 through V5 when creating a cloned voice or when media translation uses voice cloning. These match the clone options on vmeg.ai.
Choosing a clone engine
Text to speech: voiceId and batch rules
Text to speech accepts up to 50 lines in data[] per request.
Use voiceId from Voices. Do not send provider on TTS requests.
Batch rules:
Every voiceId must exist. If an ID is not in your system or clone library, the API returns 400 and points you to POST /openapi/v1/assets/voice/basic/list or POST /openapi/v1/assets/voice/clone/list.
All lines in one request must use compatible voices — same preset tier (S*) for system voices, or the same clone engine (V1–V5) for clones. For example, you cannot mix S1 and S2 presets, or V1 and V2 clones, in one request.
timbreRefAudio is clone-only. If any line uses timbreRefAudio, every voice in that batch must be a clone engine (V1–V5). Do not mix timbreRefAudio with system preset voiceIds in the same batch.
See Error codes → TTS voice validation for example messages.
Language locale codes are listed in Supported languages.
provider selects the engine family, not the language.