StemWave Extract 0.1.1-beta.1 — Diagnostics and Waveform Fixes

StemWave Extract 0.1.1-beta.1

Beta testing release. Fully offline; both model checkpoints and the runtime are included.

• Settings → Save diagnostic report… exports local timings, system/backend and model/settings details without audio, track names or paths. Nothing is uploaded automatically or placed beside extracted stems.

• FAST | STANDARD | HQ. Standard defaults to the bundled Hugging Face config's overlap 2. Fast reduces overlap for preview speed; HQ adds averaging and has a gold/white hover shimmer with reduced-motion support.

• Waveforms now render directly to each lane; stereo analysis no longer cancels opposite-polarity channels. This addresses the clipping/rendering path and flat-waveform bug, but still needs confirmation on affected Macs.

• Tested on M1 Max / macOS 13.6.6: 15 Rust tests, 13 Python tests, UI checks, real Standard-mode extraction, signature and archive integrity. The reported performance difference versus other apps is not claimed fixed.

• Apple Silicon only; macOS 13.6 or later.

• Unsigned release: ad-hoc signed, without Apple Developer ID or notarization. After moving the app to Applications, right-click → Open. macOS may require approval in Privacy & Security.

• The separation checkpoint is included in this release with redistribution permission confirmed by the publisher. It is not part of the source repository or covered by the app's MIT license.

• Fully offline: Python, pinned packages, FFmpeg, and both model checkpoints are bundled. No first-launch download.

• Runtime and models live inside the app. Preferences/cache and previous-version runtime files may live in ~/Library/Application Support/com.stemwave.extract. Exported audio stays in the folder you choose.

• To uninstall: quit and remove the app, then remove that Application Support folder. Keep or delete your exported audio separately.

• App and stemwave_core copyright © 2026 move37audio, MIT. Third-party terms are listed in engine/licenses/THIRD-PARTY.md and on the download page.

FFmpeg is LGPL 2.1 or later; the corresponding unmodified ffmpeg-7.1.tar.xz source is attached alongside the app. The app uses FFmpeg as a separate program.

StemWave Extract 0.1.0 — Fully Offline (Unsigned)

StemWave Extract 0.1.0

Separate songs into vocals, drums, bass, guitar, piano and other, with acapella and instrumental mixes. Includes Spectral and RGB DJ waveform styles. Extraction folders now contain only WAV audio.

• Apple Silicon only; macOS 13.6 or later.
• Unsigned release: ad-hoc signed, without Apple Developer ID or notarization. After moving the app to Applications, right-click → Open. macOS may require approval in Privacy & Security.
• The separation checkpoint is included in this release with redistribution permission confirmed by the publisher. It is not part of the source repository or covered by the app's MIT license.
• Fully offline: Python, pinned packages, FFmpeg, and both model checkpoints are bundled. No first-launch download.
• Runtime and models live inside the app. Preferences/cache and previous-version runtime files may live in ~/Library/Application Support/com.stemwave.extract. Exported audio stays in the folder you choose.
• To uninstall: quit and remove the app, then remove that Application Support folder. Keep or delete your exported audio separately.
• App and stemwave_core copyright © 2026 move37audio, MIT. Third-party terms are listed in engine/licenses/THIRD-PARTY.md and on the download page.

FFmpeg is LGPL 2.1 or later; the corresponding unmodified ffmpeg-7.1.tar.xz source is attached alongside the app. The app uses FFmpeg as a separate program.