Mirli

•
238
•

Generate AI sound effects, music, speech, and voice clones locally on your own GPU with complete privacy and offline support.

Mirli (formerly Soundbox) is an open-source, local alternative to cloud audio generation services like ElevenLabs and Suno. Powered by open-weight models and the native audio.cpp runtime, Mirli allows users to synthesize sound effects, compose instrumental music and full songs with lyrics, and perform text-to-speech or voice cloning directly on their own hardware without accounts or credits. It includes a multi-track timeline studio, a local HTTP API, and a 24-tool Model Context Protocol (MCP) server that empowers AI agents like Claude Code and Cursor to design, mix, and produce rich audio landscapes programmatically.

주요 기능

01Native execution via audio.cpp eliminating complex Python or CUDA environment configurations.
02Multi-track timeline studio for editing, trimming, MIDI control, voice casting, and mixdowns.
03Generates sound effects, full-length music with lyrics, and TTS voice clones entirely offline on local GPUs.
04Integrated 24-tool MCP server and agent skills enabling automated audio generation passes within Claude Code or Cursor.
05Transparent licensing interface showing clear commercial use terms and model sizes before downloading weights.
06232 GitHub stars

사용 사례

01Game developers looking to programmatically generate, test, and insert custom foley, ambience, and sound effects using AI coding agents.
02Content creators and musicians seeking to produce royalty-free background tracks, instrumentals, and songs without subscription costs.
03Software developers integrating local speech synthesis, voice casting, and audio generation pipelines via HTTP API or MCP.