Datamosh GUI

Interactive timeline-based datamoshing for clip-level I-frame and P-frame manipulation.

Current version: v1.1.5.

Support

If this project helps your workflow, support ongoing development and experiments:

• Patreon: https://www.patreon.com/Seriousshit
• PayPal: https://www.paypal.com/cgi-bin/webscr?cmd=_donations&business=dogme84%40gmail.com&currency_code=USD

Preview

Screenshot:

Sample output video:

Download example clip (AVI)

Changelog

v1.1.5

• Fix: Temp directories from normalization and I-frame injection are now cleaned up when a clip is removed, preventing /tmp accumulation over long sessions
• Fix: Video info probing (fps, frame count, dimensions) moved off the Qt main thread — no more UI freeze after importing clips
• Fix: Timeline scroll wheel now pans; Ctrl+wheel zooms (was inverted vs. shortcut docs)
• Fix: Clip list view now highlights the correct row when a timeline clip is clicked
• Fix: Settings panel controls are now disabled when no clip is selected (prevented silent no-op slider interactions)
• Fix: "Cut At Playhead" context menu item disabled on empty timeline
• Fix: UpdateWorker exceptions no longer permanently block future update checks
• Fix: OpenCV video capture handle always released via finally (prevents file lock on Windows)
• Fix: RenderDialog stops its render worker when the dialog is closed mid-render
• Fix: Drag-reorder MIME data parse is now validated; malformed drops return False cleanly
• Fix: Undo/redo stacks use deque(maxlen=200) for O(1) eviction
• Fix: Timeline hint text contrast raised to meet WCAG AA minimum

What This App Does

Datamosh GUI is a PySide6 desktop editor built for experimental compression-art workflows.

• Build a sequence from multiple clips on a timeline
• Drag clips from bin to timeline, reorder segments, and cut at playhead
• Control keyframe behavior per clip or per timeline segment
• Toggle "Drop first I-frame" on individual cuts/segments
• Duplicate P-frames to push motion-smear and prediction artifacts
• Preview timeline output before final render
• Undo/redo timeline and settings changes

The app keeps AVI bitstream manipulation in mosh.py and uses the GUI as an editor/orchestrator.

Project Layout

• main.py: GUI entrypoint
• gui/: active PySide6 application
• mosh.py: core AVI parser/rewriter + CLI
• tests/: pytest suite
• legacy/: old Tkinter-era code (reference only, not active)

Requirements

• Python 3.10+
• ffmpeg and ffprobe available on PATH

Install Python deps:

pip install -r requirements.txt

Run

python3 main.py

Or use:

./launch.sh

Releases & Local Packaging

• Current release page: https://github.com/willbearfruits/datamosh-gui/releases/tag/v1.1.5
• Download portal: https://willbearfruits.github.io/datamosh-gui/
• Release pipeline: .github/workflows/release.yml
• Packaging details: RELEASES.md

Local build output naming (Linux host):

• local-release-artifacts/Datamosh-<version>-linux-portable.tar.gz
• local-release-artifacts/Datamosh-<version>-linux-installer.deb
• local-release-artifacts/Datamosh-<version>-linux-<arch>.AppImage
• local-release-artifacts/SHA256SUMS-linux.txt

Cross-platform local attempt logs (Linux host):

• local-release-artifacts/windows-build-attempt.log
• local-release-artifacts/macos-build-attempt.log

Build Locally

Use native OS runners for release-quality builds. Linux can package Linux artifacts locally; Windows and macOS should be built on their own platforms.

Linux:

pyinstaller --noconfirm --clean --windowed --name Datamosh \
  --add-data "README.md:." \
  --add-data "LICENSE:." \
  --add-data "VERSION:." \
  main.py
./packaging/linux/build_deb.sh "$(cat VERSION)" "dist/Datamosh" "local-release-artifacts"
./packaging/linux/build_appimage.sh "$(cat VERSION)" "dist/Datamosh" "local-release-artifacts"

Windows (PowerShell, run on Windows):

pyinstaller --noconfirm --clean --windowed --name Datamosh `
  --add-data "README.md;." `
  --add-data "LICENSE;." `
  --add-data "VERSION;." `
  main.py

macOS (run on macOS):

pyinstaller --noconfirm --clean --windowed --name Datamosh \
  --add-data "README.md:." \
  --add-data "LICENSE:." \
  --add-data "VERSION:." \
  --osx-bundle-identifier com.datamosh.gui \
  main.py

Workflow (GUI)

1. Open one or more source clips.
2. Choose import mode/settings (normalize recommended, direct-AVI optional advanced).
3. Wait for ingest/normalization to complete.
4. Drag clips from bin to timeline and reorder as needed.
5. Use playhead + Cut (Ctrl+K) to split timeline segments.
6. Inject a one-frame I-frame clip from image/video (Ctrl+Shift+I) when needed.
7. Toggle Drop I (I) for selected segment when needed.
8. Adjust per-clip/segment settings:
   • Keep extra keyframes after first
   • Duplicate count
   • Duplicate gap
   • Keep/drop specific keyframe indices
9. Preview timeline output.
10. Render final AVI.

Keyboard Shortcuts

• Ctrl+O: open clips
• Ctrl+Shift+O: add clips
• Ctrl+R: render
• Ctrl+Z: undo
• Ctrl+Shift+Z / Ctrl+Y: redo
• Space: play/pause preview
• Left / Right: frame step
• Ctrl+K: cut selected timeline segment at playhead
• Ctrl+Shift+I: inject single I-frame clip from media
• I: toggle drop-first I-frame for selected segment
• Delete: remove selected timeline segment
• F1: shortcut help

CLI (Core Engine)

mosh.py is usable directly for scripted workflows:

python3 mosh.py input.avi output.avi --keep-first 1 --duplicate-count 2 --duplicate-gap 3

Normalization example:

python3 mosh.py input.mp4 output.avi --normalize --normalize-preset balanced

Testing

Headless-safe full suite:

QT_QPA_PLATFORM=offscreen pytest -q

Notes

• Current output target is AVI/Xvid.
• Non-AVI sources are normalized before processing.
• legacy/ exists for historical reference; active code path is main.py + gui/.