HeadshotFlow v1.0.1

HeadshotFlow 1.0.1 renames GibbonPfp and brings the latest headshot editing and batch workflow improvements into a release.

• HeadshotFlow branding in the desktop, launcher, and release packages.
• Reusable processing masks and face detections for faster repeated edits.
• Automatic/CPU processing controls, device diagnostics, and loaded-model release.
• Processing stages, elapsed times, and CLI timing/cache diagnostics.
• Public portrait benchmarks and autosave/workspace recovery.
• Linux file dialogs use the desktop portal, supporting GTK file pickers through the configured portal backend.
• Portable brush test fixtures and retained CI test diagnostics.

Existing preferences and recovery data remain compatible. The command is still gibbonpfp; the repository, package ID, and internal storage names retain their existing identifiers.

Downloads

Package	Target
Windows installer or ZIP	Windows x64
AppImage	Linux x64
macos-x64 DMG	macOS Intel
macos-arm64 DMG	macOS Apple Silicon

Windows ZIP: extract the whole archive and run bin/gibbonpfp.exe. macOS: copy gibbonpfp.app from the DMG to Applications. Linux: make the AppImage executable; without FUSE, launch with APPIMAGE_EXTRACT_AND_RUN=1.

The face and Fast models are bundled. High Quality is a separate download and requires load confirmation. Bundled release inference uses CPU; local builds can use a compatible CUDA runtime. Segmentation quality varies by photo, and uncertain framing requires review unless fully automatic export is selected. RAW compatibility depends on camera support in LibRaw.

Package hashes are in SHA256SUMS.txt. See the README for usage and limitations.

Validation and signing

All four native CI jobs passed at commit a4d0a17: Ubuntu 24.04, Windows Server 2022, macOS 15 Intel, and macOS 15 Apple Silicon. Checks covered CTest, CLI integration, GUI startup, packaging, and relocated-package CLI verification. The downloaded Linux AppImage also passed a local offscreen startup check. Local native Qt screenshots were inspected at 1024 × 720 and 150% scale. Minimum target OS versions were not each tested separately.

Windows packages are unsigned. Both macOS applications are ad-hoc signed and are not notarized.

GibbonPfp 0.1.0

GibbonPfp 0.1.0 is an offline Qt desktop application and CLI for preparing profile pictures.

• Single-photo and batch workflows, with default 360 × 480 portrait output and optional uncapped/custom sizes.
• Face-aware cropping, adjustable headroom, manual framing, and review gates for uncertain photos.
• Perceptual brightness, reference matching, lightweight background removal, mask brushes, and JPEG/PNG output.
• JPEG, PNG, TIFF, WebP, SDR HEIC, and camera RAW input; RAW white balance and highlight controls.
• Presets, sessions, collision-safe exports, reports, and Strider-inspired light/dark themes.

Download	Target
Windows installer or ZIP	Windows 11 x64
AppImage	Ubuntu 24.04+ x64
macos-x64 DMG	macOS 14+ Intel
macos-arm64 DMG	macOS 14+ Apple Silicon

Windows packages are unsigned. Mac applications are ad-hoc signed, not notarized; operating-system publisher checks may require explicit approval to open them. No Python or Qt installation is needed to run the packages.

Linux: make the AppImage executable, then open it. If FUSE is unavailable, run APPIMAGE_EXTRACT_AND_RUN=1 ./GibbonPfp-0.1.0-x86_64.AppImage. Windows ZIP: extract the entire archive and run bin/gibbonpfp.exe. macOS: open the DMG and copy gibbonpfp.app to Applications.

The small face and Fast segmentation models are bundled. The optional High Quality model is a separate 973 MB download and can use about 7 GB RAM on CPU. Photos are processed locally.

Validation includes native core/controller tests, GUI startup, relocated-package checks without Qt SDK search paths, RAW/HEIC decoding, and 16 MP/24 MP CLI inputs. CI uses Ubuntu 24.04, Windows Server 2022, and macOS 15 Intel/ARM runners. The Linux AppImage is additionally checked on EndeavourOS. Minimum target OS versions are not each tested separately.

See README for usage and limitations, and building/distribution for dependencies and signing. SHA256SUMS.txt lists the package hashes.

Verified build: all four native CI jobs passed, commit 153f8ae.