Clovercut v0.1.7

Clovercut 0.1.7

Desktop release for Linux x86_64/aarch64, Windows x86_64/ARM64 and macOS Intel/Apple Silicon. Mobile packages are not included.

Desktop timeline and interface

• Give automatic track heights a hierarchy: 64 px main video, 48 px other video/audio, 36 px stickers and 28 px text, captions and adjustment tracks. Manual track sizes remain available.
• Expand the ruler/scrubbing strip and make vertical grid lines optional, disabled by default. Clip selection does not move the playhead.
• Open overflowing timelines on their actual main video track, with audio below it, and retain each timeline's chosen vertical position during the session.
• Use Up/Down to navigate previous/next cuts across tracks, preserving fractional frame-rate precision and customizable shortcuts.
• Add 4:5 and 3:4 frame presets in project creation, project settings and the monitor.
• Simplify library category paging to one arrow, remove redundant pane selectors and tighten titlebar spacing. Highlight selected category and inspector labels with the accent colour.
• Enlarge audio cards and show their source, category, duration and licence information.

Audio libraries and model downloads

• Start the bundled local YouTube Audio API connector on demand. Open a dedicated browser for Studio sign-in and reuse the saved session in the background. No VPS is required; Chrome, Chromium or Edge must be installed. Expired sessions require signing in again.
• Keep separate YouTube music and sound-effect searches. Preserve download/preview validation and project-owned audio copies across providers.
• Route supported downloads through system/environment proxies, retain resumable model downloads and provide translated failure messages without exposing proxy credentials or signed URL queries.
• Offer browser download and manual import after a model download fails. Verify the pinned SHA-256 before installation; wrong or cancelled imports preserve both the user's original and any partial download.
• Update Kokoro archive sizes and digests to match the upstream model files.

Enhance and saved voices

• Process only the used source span when enhancing a clip, with a safety margin, correct audio alignment, cancellation and atomic publication. Reuse repeated frames and compatible processed copies.
• Choose between GPU and ONNX Enhance engines using measured execution. Reject incompatible weights and fall back to ONNX if the selected GPU engine later fails.
• Import saved voices from audio/video with metadata and a suggested sample start. Cancel stale analysis when a recording changes or its dialog closes.

Playback and rendering fixes

• Reuse compatible cached PCM spans immediately after cuts while preserving stream selection, gain curves and nonlinear speed-ramp handling.
• Avoid concurrent SVG temporary-file races and invalidate cached SVG/SVGZ pictures when linked images change.
• Preserve overflowing title glyphs, shadows and transformed text across preview and export.
• Correct Windows shader sampling and the colour matrix of untagged ProRes media.
• Avoid the macOS shutdown path through whisper/Metal static destructors after save/export handling.

Languages and desktop downloads

• Translate new model-download, saved-voice and shortcut messages across all 25 languages and simplify timeline tooltips.
• Improve Windows ARM download detection on the website and refresh desktop documentation.

Existing projects remain supported. Manual track sizes, caption grouping, typed-track placement, generated-media ownership and the existing speed-ramp engine are retained. Automatic linkage between tracks, per-word rich text and transparent-canvas/alpha export are not part of this release.

See audio setup and the upstream adaptation record for details.