Skip to content

1.1.3

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 17 Aug 21:42
· 4 commits to main since this release
v1.1.3
b0ef7a3

Latent Tools — v1.1.3

Local-first Windows desktop app for bulk image-dataset prep: AI watermark
removal, format conversion, and uncensored image captioning — all running
entirely on-device on your own GPU.

What's new in v1.1.3

  • Packaging cleanup — removed a sidecar.spec file introduced in v1.1.2
    that was suspected as a cause of installer size, and confirmed not to be
    one. The sidecar is packaged via the same plain PyInstaller CLI invocation
    every release through v1.1.1 used, with the two stale CRT DLLs (see
    v1.1.2's notes) now deleted via a straightforward post-build step instead
    of a .spec-file filter.
  • Everything from v1.1.2 carries forward: the packaged sidecar no longer
    crashes on startup (ModuleNotFoundError, then a msvcp140.dll access
    violation — both fixed in v1.1.2), and the new LT monogram app icon.
  • Installer size clarified, not a bug: each packaged .exe is ~616MB.
    That reflects genuinely bundling PyTorch, CUDA, and the inpainting/
    captioning model stack — v1.1.1's much smaller ~104MB build looked
    "healthy" but was actually broken: a packaging bug meant it never
    included torch at all, so it silently omitted the sidecar's real
    dependencies rather than being efficiently packaged. There's real room to
    shrink this further (unused model backends pulled in transitively,
    dev-only tooling in the packaging environment) — tracked as a pruning
    backlog in HANDOVER.md — but ~600MB is the correct current baseline,
    not a regression.

Highlights (full feature set)

  • Watermark removal — Florence-2 open-vocabulary detection finds
    watermarks/logos/text, LaMa (IOPaint) inpaints them out. Manual mask
    brush/eraser editing with undo/redo if the automatic detection needs a
    nudge.
  • Format conversion — JPEG / PNG / WEBP export with quality, lossless,
    compression-level, background-flatten color, and metadata-retention
    controls. Export presets for LoRA / Archive / Web, plus custom
    localStorage-backed presets.
  • Uncensored image captioning — Qwen2-VL-2B / Qwen2-VL-7B-Instruct, or
    point it at your own local model folder. Custom system prompts and
    trigger-word support.
  • Bulk dataset processing — folder-in, folder-out batch pipeline built
    for preparing training datasets (e.g. LoRA), not just one-off cleanup.
    Single round-trip /process pipeline (normalize → detect → inpaint →
    caption → convert) for throughput.
  • Single Image Editor — Detect → Remove → Caption stepper with a
    zoomable/pannable canvas mask overlay.
  • Live GPU telemetry — real-time GPU name, VRAM usage, and temperature
    in the titlebar and sidebar.
  • Runs fully local — no cloud calls; a Python (FastAPI) sidecar talks
    to the Electron app over 127.0.0.1 only.

Requirements

  • Windows 10/11 (64-bit) — no macOS/Linux build today.
  • A dedicated, CUDA-capable NVIDIA GPU. There is no CPU fallback.
    8GB+ VRAM recommended; 16GB+ if you plan to run the 7B captioning model.
  • ~616MB of free disk space for the download.

Installation

Download and run either:

  • Latent-Tools-Setup-*.exe — NSIS installer
  • Latent-Tools-*.exe (no "Setup" in the name) — no-install portable build

Known limitations

  • Windows-only.
  • First run downloads model weights (Florence-2, LaMa, Qwen2-VL) from
    Hugging Face — expect a delay and disk usage on first launch.
  • No auto-update mechanism yet — check the Releases page for new versions.
  • Installer size (~616MB) has room to shrink further via dependency
    pruning; see the packaging backlog in HANDOVER.md. Not a bug.

Full Changelog: v1.1.2...v1.1.3

Assets 8
Loading