Case study
ActVoice
An accessible audio drama studio: scenes, speech synthesis, sound cues, and server-side rendering through REST/MCP.
Audio drama from scenes, lines, and sound cues
ActVoice is a server-side audio drama studio designed for people and AI agents. People use a web interface or API; agents use MCP or REST. A project defines characters, scenes, dialogue, ambience, and cues, which the server assembles into audio.
Documented implementation
The README lists Edge TTS, local RHVoice, and an Edge-to-RHVoice fallback. SQLite stores projects and render jobs. Output includes MP3, WAV, and a manifest recording voices and line timing.
Sound cues can be anchored to a scene’s start or end, or placed before or after a line. The server measures speech duration and resolves cue positions. Dialogue can therefore change without manually recalculating the entire timeline.
The MVP does not require a built-in language model: an external agent can prepare the structure. User-provided OpenAI keys are marked as a future mode in the README, not presented here as a finished feature.
The project connects AI workflow design, API integration, and accessible creative tools. Its documented architecture is not a runtime test of a particular deployment.