1.1.3

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

1.1.2

Assets removed — superseded by v1.1.3. This release's binaries worked correctly and have been deleted only because v1.1.3 supersedes it entirely. Get v1.1.3 instead.

---

Latent Tools — v1.1.2

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 v1.1.2

• Fix: packaged sidecar crashed on startup, every release through v1.1.1.
  Nobody had actually run the packaged .exe and checked GPU status since
  v1.0.0 first shipped — CI only ran the unpackaged test suite, which never
  exercises the frozen binary. Two independent bugs, fixed here:
  • uvicorn.run("app.main:app", ...) used the "module:attr" string form,
    invisible to PyInstaller's static analysis, so it silently dropped the
    app package from the bundle and the packaged sidecar crashed instantly
    with ModuleNotFoundError: No module named 'app'.
  • Past that, the sidecar crashed with OSError: [WinError 1114] ... c10.dll.
    The real fault (only visible via Windows Event Viewer) was an access
    violation inside a stale, PyInstaller-auto-bundled msvcp140.dll that
    shadowed the correct, much newer system copy — torch's c10.dll (built
    against a newer CRT) crashed calling into it.
  • Full technical detail is in HANDOVER.md in the repo.
• New app icon — LT monogram replaces the previous mark across the
  installer, titlebar, and system tray.

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.

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.
• Each packaged .exe was ~616MB — that's the correct cost of bundling
  PyTorch, CUDA, and the model stack (v1.1.1's much smaller build looked
  efficient but was actually broken — see v1.1.3's notes). Not a bug.

Full Changelog: v1.1.1...v1.1.2

1.1.1

Assets removed — the packaged .exe from this release crashed on startup (GPU features never came up), so the binaries have been deleted. Get v1.1.3 instead.

---

Latent Tools — v1.1.1

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 v1.1.1

• Latent Design System migration — Full visual migration from the old
  Deep Neon theme to the shared Latent Design System: dark graphite canvas
  (#0A0A0D), step-up flat surface levels, desaturated Latent Cyan/Violet
  accents, and the signature Latent brand gradient. Titlebar logo now
  renders from the vendored latent-mark.svg asset instead of a
  hand-inlined copy, and the redundant Google Fonts <link> (which
  double-loaded fonts already pulled locally, at the wrong JetBrains Mono
  weights) has been removed. Sidebar width corrected to the Design
  System's 224px.
• Settings modal — New sidebar Settings button opens a Latent Design
  System modal with a Ko-fi support link.
• Security & reliability fixes from an internal code review:
  • shell:open-external IPC now only allows http:/https: schemes.
  • Main-process image cache capped at 10 entries with LRU eviction, so
    long single-editor sessions no longer grow RAM unbounded.
  • Florence-2 detection wrapped in torch.inference_mode().
  • folder:list-images now returns cleanly on permission errors instead
    of throwing an unhandled IPC rejection.
  • Renderer revokes Object URLs on image reload/change instead of
    leaking them.
  • openExternal link handling now routes through a main-process IPC
    channel for reliable link opening.
• appId changed to com.nilsson.latent.tools (was com.latenttools.app),
  matching the com.nilsson.latent.* convention used by sibling Latent
  apps. If you have v1.0.0 installed, the new installer may add a
  second Start Menu/uninstall entry rather than replacing the old one in
  place — uninstall the old version manually if that happens.
• Editor UX fixes — caption button, zoom controls, and canvas import
  behavior tightened up; brush stays live immediately after a removal
  pass instead of requiring a Detect first.
• Repo hygiene — .agents/, .claude/ (except settings.json, which
  carries the secrets-read deny list), and docs/ are now gitignored as
  local dev tooling/reference material rather than shipped repo content.
  Stale superseded design-direction docs were deleted outright.
• Documentation accuracy pass — README, HANDOVER, and CONTRIBUTING
  corrected to match the actual app: Electron version badge, removal of
  an inaccurate "token-authenticated" loopback claim (there is no auth
  token), and System Requirements narrowed to Windows-only with no CPU
  fallback (the packaged build only targets Windows, and the sidecar
  requires CUDA). The GPU requirement is now called out at the top of the
  README, not just in System Requirements.

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.

Installation

Download and run either:

• Latent-Tools-Setup-*.exe — NSIS installer
• Latent-Tools-*-portable.exe — 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.
• The appId rename (see above) means upgrading in place from v1.0.0 may
  leave a duplicate Start Menu/uninstall entry until the old one is
  removed manually.

Full Changelog: v1.0.0...v1.1.1

1.1.0

Assets removed — the packaged .exe from this release crashed on startup (GPU features never came up), so the binaries have been deleted. Get v1.1.3 instead.

---

Latent Tools — v1.1.0

1.0.0

Latent Tools v1.0.0 — First Full Release with Unified Latent Design System

🚀 Highlights & Features

• Unified Latent Design System Rework: Upgraded visual UI to match the official Latent Design System (#0A0A0D near-black graphite canvas, desaturated Latent Cyan #4FD8D0 and Latent Violet #9B7EF5 accents, signature brand gradient, Inter typography, and JetBrains Mono telemetry).
• Single-Round-Trip /process Endpoint: ~45% faster bulk dataset processing via optimized single HTTP roundtrips between Electron main and FastAPI sidecar.
• Bulk Dataset Processor: Folder-based processing, drag-and-drop folder input & output dropzones with real-time thumbnail grid, compression presets, and live log terminal.
• Single Image Editor: Pipeline stepper (Detect → Remove → Caption), interactive mask canvas overlay with undo/redo history, zoom/pan controls, and customizable LoRA training dataset caption generator (Qwen2-VL-2B / 7B / custom model folders).
• Offline & Firewall Isolation: Python FastAPI sidecar bound strictly to 127.0.0.1 loopback with PyInstaller standalone compilation (sidecar.exe) preventing Windows Firewall security prompts.
• UI Scaling: Ctrl + mousewheel zooming (50%–250%) and Ctrl+0 reset shortcut via webFrame.

🛠️ Fixes & Polish

• Fixed mask editing touch-up behavior after inpainting.
• Standardized active/hover sidebar icon stroke and button glow ring styling.
• Resolved brush mode segmented control active state toggling (Add / Erase).
• Integrated brand gradient square logo in titlebar and system tray icon integration.

v0.2.0-beta

Latent Tools — v0.2.0-beta

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 v0.2.0-beta

• Modernized Deep Neon UI Design System — Full visual refresh with
  charcoal-navy ground (#0a0b10), ambient radial gradients, crisp solid
  card surfaces (#12141b), restrained hover rings (--glow-ring), active
  navigation gradient left accent bars, solid gradient .btn-cta primary
  action buttons, 3px range sliders, and JetBrains Mono typography for VRAM,
  temperature, and zoom readouts.
• Fix: brush unusable right after "Remove Watermark" — Previously the
  mask canvas was hidden after an inpaint pass, which silently disabled all
  brush input until you ran Detect again and cleared it. The brush now
  stays live immediately after removal so you can touch up any remaining
  watermark without the Detect detour.
• Documented usage disclaimer — The README and CONTRIBUTING guide now
  state clearly that Latent Tools is intended for removing watermarks you
  have the rights to remove (your own marks, or marks on assets you're
  licensed to edit) — not for stripping copyright or ownership marks from
  other people's work. Responsibility for lawful use rests with the user.
• LICENSE.md added — The MIT license was published in the repo and
  referenced from package.json.
• CONTRIBUTING.md added — Setup steps, workflow expectations (GPU
  exclusivity for performance work, test-first bugfixes), and commit/PR
  conventions for anyone contributing to the project.
• README overhaul — Documents bulk dataset processing, the
  single-round-trip /process sidecar endpoint, the 2B/7B/custom
  captioning model selector, export presets, the Deep Neon UI, and how to
  grab a prebuilt release.

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.
• Deep Neon UI — dark, glassy theme; Ctrl + scroll to zoom the whole
  UI (50%–250%).
• 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
• A CUDA-capable discrete GPU (Florence-2, LaMa, and Qwen2-VL all run on
  GPU; recommend 12 GB+ VRAM)

Installation

Download and run either:

• Latent-Tools-Setup-*.exe — NSIS installer
• Latent-Tools-*-portable.exe — no-install portable build

Known limitations (beta)

• Windows-only for now.
• 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.

Full Changelog: v0.1.0-beta.2...v0.2.0-beta

v0.1.0-beta.3

Latent Tools — v0.1.0-beta.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 since v0.1.0-beta.2

• Fix: brush unusable right after "Remove Watermark" — Previously the
  mask canvas was hidden after an inpaint pass, which silently disabled all
  brush input until you ran Detect again and cleared it. The brush now
  stays live immediately after removal so you can touch up any remaining
  watermark without the Detect detour.
• Documented usage disclaimer — The README and CONTRIBUTING guide now
  state clearly that Latent Tools is intended for removing watermarks you
  have the rights to remove (your own marks, or marks on assets you're
  licensed to edit) — not for stripping copyright or ownership marks from
  other people's work. Responsibility for lawful use rests with the user.
• LICENSE.md added — The MIT license was previously only claimed in the
  README with no license file; it's now published in the repo and
  referenced from package.json.
• CONTRIBUTING.md added — Setup steps, workflow expectations (GPU
  exclusivity for performance work, test-first bugfixes), and commit/PR
  conventions for anyone contributing to the project.
• README overhaul — Now documents bulk dataset processing, the
  single-round-trip /process sidecar endpoint, the 2B/7B/custom
  captioning model selector, export presets, the Deep Neon UI, and how to
  grab a prebuilt release — all previously undocumented.

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.
• Deep Neon UI — dark, glassy theme; Ctrl + scroll to zoom the whole
  UI (50%–250%).
• 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
• A CUDA-capable discrete GPU (Florence-2, LaMa, and Qwen2-VL all run on
  GPU; recommend 12 GB+ VRAM)

Installation

Download and run either:

• Latent-Tools-Setup-*.exe — NSIS installer
• Latent-Tools-*-portable.exe — no-install portable build

Known limitations (beta)

• Windows-only for now.
• 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.

Full Changelog: v0.1.0-beta.2...v0.1.0-beta.3

v0.1.0-beta.2

Latent Tools — v0.1.0-beta.1 (First Beta)

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.

This is the first public beta. Expect rough edges; feedback and bug reports
are welcome.

Highlights

• 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.
• Deep Neon UI — dark, glassy theme; Ctrl + scroll to zoom the whole
  UI (50%–250%).
• 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
• A CUDA-capable discrete GPU (Florence-2, LaMa, and Qwen2-VL all run on
  GPU; recommend 12 GB+ VRAM)

Installation

Download and run either:

• Latent-Tools-Setup-*.exe — NSIS installer
• Latent-Tools-*-portable.exe — no-install portable build

Known limitations (beta)

• Windows-only for now.
• 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.

Full Changelog: https://github.com/erroralex/Latent-Tools/commits/v0.1.0-beta.1

v0.1.0-beta.1

Latent Tools — v0.1.0-beta.1 (First Beta)

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.

This is the first public beta. Expect rough edges; feedback and bug reports
are welcome.

Highlights

• 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.
• Deep Neon UI — dark, glassy theme; Ctrl + scroll to zoom the whole
  UI (50%–250%).
• 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
• A CUDA-capable discrete GPU (Florence-2, LaMa, and Qwen2-VL all run on
  GPU; recommend 12 GB+ VRAM)

Installation

Download and run either:

• Latent-Tools-Setup-*.exe — NSIS installer
• Latent-Tools-*-portable.exe — no-install portable build

Known limitations (beta)

• Windows-only for now.
• 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.

Full Changelog: https://github.com/erroralex/Latent-Tools/commits/v0.1.0-beta.1