Platform Support

Platform Support

Zegplayer is developed on macOS (Apple Silicon) and Windows x64. Both are usable today. Linux builds and runs but its convenience layer isn't wired up yet.

Status at a glance

Capability	macOS	Windows	Linux
App boots, DB, MCP server	✅	✅	✅ (needs CI verification)
Load local video, transcribe	✅	✅	✅
YouTube import (uses YouTube's own captions when available)	✅	✅	✅
In-app Install buttons	✅ (Homebrew)	✅ (direct download)	❌
yt-dlp / whisper-cli auto-install on first launch	❌ (use the Install buttons)	✅ (both, no user action)	❌
Whisper speech model pre-fetched at first launch	✅ (once whisper.cpp is installed)	✅	✅ (once whisper.cpp is installed)
Whisper fallback to openai-whisper (Python)	✅	✅	✅
Configurable openai-whisper model size	✅	✅	✅
AV/proxy MITM tolerance (yt-dlp SSL bypass)	n/a	✅ (automatic)	n/a
Model/data paths (HOME-based)	✅	⚠️ HOME often unset	✅
opencode MCP config path	✅	❌ wrong location	✅
Local-provider on-disk detection	✅	⚠️ /Applications-only	⚠️
Obsidian install/detect	✅	⚠️ degraded	⚠️ degraded

Remaining blockers on Windows/Linux

• #3 HOME/Homebrew assumptions break model & tool paths on Windows.
• #4 opencode MCP config path is hardcoded ~/.config (POSIX).

Degraded (works, but worse)

• #5 Local-provider on-disk detection only scans macOS /Applications.
• #6 Obsidian install/detection is macOS-only.
• #7 Executable discovery lacks Windows .exe / Program Files & Linux paths.
• #8 CI: build + smoke-test artifacts on all three OSes.
• #9 Docs: per-platform manual install fallback until installers land.

Working today on Windows/Linux

If you install ffmpeg, whisper-cli, yt-dlp, and a model server manually and keep them on your PATH, the core flow — load video, transcribe locally, generate notes — works. What differs per platform is the convenience layer (in-app installers, silent first-launch downloads) and a few integration paths, not the core pipeline.

How each platform gets its binaries

• macOS — Homebrew, triggered by the Install buttons in Connected Services (brew install whisper-cpp / yt-dlp). Nothing is downloaded silently at launch. whisper.cpp stopped publishing a macOS arm64 release archive, so the app no longer attempts a direct download; it points you at brew instead.
• Windows — direct download from the tools' official GitHub releases into %APPDATA%\Zegplayer\bin\, both on first launch and from the Install buttons. ffmpeg is the exception (too large) — install it with winget/scoop/choco.
• Linux — manual, via your distribution's package manager.

On every platform, once a whisper-cli is present the app pre-fetches the ggml speech model in the background at launch, so your first transcription doesn't stall on a ~466 MB download. It is skipped when you're running the Python openai-whisper, which manages its own model cache.