wazza-voice
wazza-voice
Wazza reads a script in a voice of the workspace (voice), reads a scene in several voices as one
take (scene), designs a new voice from a description (voice-design), and performs a recording
again in another voice (voice-change). Always list the voices first: a voice is named exactly as
list_voices gives it.
Connection
Use the Wazza MCP server (tools list_voices, list_tools, estimate_run, start_run,
get_run, upload_file, complete_upload). The Wazza plugin for Claude Code connects it; other
clients add https://api-pro.wazza.ai/v1/mcp to their MCP config. Without MCP, use the REST API at
https://api-pro.wazza.ai/v1 with Authorization: Bearer $WZ_KEY (GET /v1/voices). Never
print, write or ask for a key in the conversation or in a file.
A local recording goes up with upload_file, then the bytes to the upload URL it returns, then
complete_upload. Use the file id as audio_url. A key starting wz_test_ runs free and returns
a sample file: it checks the setup, not the read.