NiceVois Agent API ยท developer preview
Let an agent train the voice model
NiceVois gives coding agents one small, predictable job: take permitted audio, train a private RVC v2 model, and return the finished .pth, matching .index, and complete ZIP.
Read current formats and limits.
Create a job and send the audio.
Start once, then check progress.
Give the user PTH, index, and ZIP.
Machine-readable entry points
Agents can discover the public requirements without signing in. Private job creation, status, and artifacts require an access credential.
https://nicevois.com/openapi.jsonhttps://nicevois.com/api/mcphttps://nicevois.com/api/agent/v1/capabilitiesThe contract an agent should follow
- Call
getTrainingRequirementsorget_training_requirements. - Confirm the user owns or controls the voice, authorizes training, and accepts the NiceVois terms.
- Create a job, upload the exact audio bytes to the returned one-time PUT URL, and start training.
- Check the job no more than once every 15 seconds. When complete, return all available artifact links.
Permission comes first. NiceVois is for voices the user owns or has clear permission to train. Agents must never silently infer consent or expose a NiceVois credential in prompts, URLs, logs, or their final answer.
What works today
The versioned Agent REST API and remote MCP tools expose discovery, private job creation, progress, and artifact retrieval. Access is currently a controlled developer preview while account linking is completed.
About drag-and-drop chat
A coding agent that can read a local attachment and make authenticated HTTP requests can use this contract. A normal ChatGPT, Claude, or Gemini conversation still needs that client to provide secure file handoff and NiceVois account linking. The published chat integration will add those pieces; this page does not pretend every chat client supports them already.
Tell us which coding agent or chat client you want to connect. We will provide preview access and integration guidance.
Contact NiceVois