Local-first music system / desktop
SLISIC
A local-first desktop music system for saving network sources as local audio, learning audio-style model artifacts, and editing waveform ranges in real audio seconds.
Network sources become local audio.
Slisic resolves extractor-backed URLs with yt-dlp, selects audio-only formats, extracts m4a through the managed FFmpeg path, commits stable files and DB music rows, writes raw manifest evidence, normalizes DB titles, and queues common-tail trim.
A predictive topology that moves with listening.
Slisic learns relationships between tracks directly from their audio. It organizes the library into nearby styles and broader listening regions, then weighs the current track, recent history, repetition pressure, and which regions remain reachable. The result is a queue that stays coherent without getting trapped in the same sound.
High-performance waveform rendering engine.
The editor maps audio seconds through a visual-padding viewport, requests FFmpeg-backed waveform summaries, and reads only demand-scoped tiles while keeping padding outside editable audio.
Loudness normalization from measured evidence.
Playback targets -18 LUFS only when the track carries finite non-zero loudness evidence. Slisic measures the exact audio range with FFmpeg, persists integrated LUFS, true peak, LRA, short-term body loudness, and presence, then projects a guarded gain plan without turning missing evidence into a playback block.
One local library, heard through another device.
Enable Remote Share in Slisic and enter its single connection code in the browser. The relay preserves code ownership and coordinates WebRTC, while HLS media moves between Host and viewer over the P2P supply path. A persistent forward reserve feeds one native audio timeline, so signaling recovery never becomes a second player.
Local library. Managed evidence.
Open source. Local files, model artifacts, and waveform cache stay on this machine.