Skip to content

Releases

Sitr v0.6.0 - Smart clothing, animated images, and PDF protection

Choose a tag to compare

@tantour tantour released this 16 Nov 16:56

Highlights

  • Optional smart clothing fill extends garment colors over covered body skin. Hair uses black coverage with a smoother outline and a narrower face-aware edge.
  • Custom rules support individual and combined body areas, multiple percentage conditions, and configurable actions. Expansion follows each person's size in frame. The settings editor has clearer area controls and fabric-fill placement.
  • Animated GIF and WebP use frame-matched video protection; single-frame files use image settings. Gallery previews keep inactive animations covered without keeping every decoder running.
  • PDF protection uses the existing image policies, processes visible content first, prepares the next three pages, and retains the previous two pages. Local and network PDFs have loading feedback, editable picture boxes, tile reconstruction, and scan-based picture detection without another AI layout model.
  • Video updates improve motion alignment, tracking continuity, mask delivery, and brief-gap handling. This release does not guarantee a particular frame rate on every site or GPU.
  • Fixes cover tab resume, dynamic Google Images previews, small thumbnails, image backgrounds, and stale clothing overlays during reprocessing.

Installation and updates

Download sitr-0.6.0-chrome.zip, verify it against SHA256SUMS.txt, extract it, and load its folder from chrome://extensions with Developer mode enabled. Update an existing unpacked installation in the same folder, reload the extension, and refresh open tabs or reopen PDFs. The installed extension is named Sitr.

Chrome 148 or later is required. Automatic PDF handling requires Chrome 151 or later; the popup also provides Open PDF reader. PDFs use Images policies through their separate PDFs switch. PDF handling uses a custom reader and does not implement all native Chrome PDF features.

Limits and privacy

Processing stays on the device. Network PDF/image retrieval contacts the source website and may use that site's cookies; there is no inference upload. Downloaded PDF bytes and the reading window consume memory, and preparing pages increases memory and background work. Picture boxes, skin segmentation, appearance estimates, and motion masks remain approximate. Scans can include text or decorations in a box, and content can still be missed. Manual PDF box corrections last for the current document session.

Model weights and hashes remain unchanged from 0.5.2. PDF.js 6.5.438 and its notices are included. Extension permissions are unchanged. The matching source and model archives accompany the Chrome archive for modification and license compliance.

Validation

See release validation for checks performed on the release build. Test results do not guarantee complete filtering or compatibility with every site or PDF.

Sitr v0.5.2 — Part thresholds and visual video controls

Choose a tag to compare

@tantour tantour released this 20 Oct 07:45

Choose when an image part should cover the entire part or body, split chest and
belly independently, and control video coverage with the same visual editor as
images. Optional Advanced video mode adds independent body-skin, face and hair
masks while keeping the existing person detector.

New and improved

  • Per-part image thresholds: select an area and percentage, then Hide entire
    part or Hide entire body. The percentage is exposed body skin divided by the
    estimated visible part area, including clothing. A rule triggers strictly
    above its percentage. Set the part to Show for conditional-only coverage;
    ordinary Show/Hide actions still apply.
  • Threshold face action: body rules offer Black (the default) or Show. Show
    reopens the semantic face and uniquely associated face box after expansion,
    overriding that person's ordinary face effect while the rule triggers. Black
    takes precedence if multiple triggered body rules disagree.
  • Chest and belly: separate approximate halves use existing shoulder/hip
    landmarks and dense anatomy to follow body orientation, including sideways,
    lying-down and upside-down cases. Saved torso choices migrate to both areas;
    missing or ambiguous anatomy retains Unresolved torso.
  • Advanced video mode: off by default and explicitly marked heavy on
    performance
    . MediaPipe runs on each analyzed video frame alongside either
    YOLO silhouettes or fast person boxes. Body skin excludes faces; face and hair
    have separate Show/Hide choices. All three masks share one semantic pass.
    Videos do not run BodyPix part segmentation or image thresholds.
  • Advanced video performance: remove the inherited 4–8 Hz cap, analyze
    available frames as quickly as inference permits, reuse reliable person motion for at
    most 250 ms between detector passes, and capture large face inputs only when
    classification is due. Request high-performance MediaPipe graphics and report
    the active semantic GPU adapter. Add a playing-video benchmark.
  • Visual video editor: Female/Male tabs, clickable person diagrams, keyboard
    area selection, action buttons and advanced Show all / Hide all. Separate
    expansion respects shown regions. Standard whole-person effects are retained
    when advanced mode is switched off; image/video policies stay independent.
  • Refreshed README, user guide, English/Arabic screenshots and regression checks.
    Model weights, artifact hashes, required permissions and Chrome minimum are
    unchanged from 0.5.1.

Install or update

Download sitr-0.5.2-chrome.zip and SHA256SUMS.txt, verify its SHA-256,
and extract into a permanent folder. In chrome://extensions, enable Developer
mode and Load unpacked the folder containing manifest.json. Requires
Chrome 148+; models and runtimes are included.

To update, replace the files in the same extension folder, click Reload,
and refresh website tabs. This preserves the unpacked extension identity and
saved settings. Wait for startup preparation to finish. Advanced video mode
remains off until enabled.

Developers: sitr-0.5.2-source.zip contains matching source;
sitr-0.5.2-models.zip contains pinned models, original export inputs and
licenses. The 0.5.1 model files are also compatible because hashes are unchanged;
0.5.0 and older bundles do not meet the restored BodyPix output pins. Publish
all three ZIPs and SHA256SUMS.txt together under tag v0.5.2.

Validation

Initial release preparation passed TypeScript, 189 unit tests, all 18
browser checks
, and refreshed English/Arabic UI screenshots. The performance
follow-up expands unit coverage to 194 tests and reruns affected browser
checks; see the validation record for current evidence. npm audit reported zero vulnerabilities.
Checks include both packaged video detectors with independent semantic masks,
threshold face overrides, cached policy edits, model fallback and local WebRTC
streams. Archives include matching source, pinned models and SHA-256 checksums.

Performance, privacy and limits

Advanced video adds a semantic inference pass and can substantially slow analysis,
especially on CPU or with several videos. Face/hair selections reuse that pass;
they do not start additional models. Standard mode preserves the existing
whole-person video path. This release introduces no new model weights, network
services, analytics or accounts; media processing stays on device.

Skin ratios, anatomy and appearance labels are fallible estimates of visible
content. Uncertain/missing anatomy can prevent a threshold from triggering.
Mask expansion can cover nearby clothing, and motion can outrun sampled video
masks. These controls do not guarantee complete coverage or zero-frame exposure.
Meet validation uses local WebRTC fixtures rather than an actual meeting.

See release validation, area behavior,
user guide, and the
README screenshots.
Original Sitr files use AGPL-3.0-only; third-party assets retain their licenses.

Full Changelog: v0.5.1...v0.5.2

Sitr v0.5.1 — Separate coverage, ready models

Choose a tag to compare

@tantour tantour released this 15 Oct 07:07

Independent groin coverage, improved upper-thigh boundaries, and models prepared
ahead of browsing work. The popup now shows actual startup progress and elapsed
seconds in English and Arabic.

New and improved

  • Groin (pelvis) is a separate image-area choice for Female and Male. Hide
    groin and thighs while leaving the torso shown, or choose each independently.
  • Upper thighs use hip-to-knee landmarks to reduce the gap near the pelvis.
    Existing BodyPix weights expose original pose/short-offset outputs, with no
    retraining or additional model family. Skin-only coverage excludes clothing.
    Cropped, overlapping, or low-confidence anatomy can remain uncertain.
  • All model variants run synthetic warmup at browser/extension startup and stay
    loaded across pages and setting changes. Engine recovery repeats preparation.
  • The popup shows completed steps and elapsed time, then Ready for images and
    videos
    . Limited-feature and failure states offer Retry. Progress measures
    work completed, not estimated time remaining.
  • Saved groin choices inherit the former torso choice. Updated English/Arabic
    controls, screenshots, user guide, model provenance, and regression checks.

Install or update

Download sitr-0.5.1-chrome.zip and SHA256SUMS.txt, verify the checksum,
and extract into a permanent folder. Open chrome://extensions, enable
Developer mode, and Load unpacked the folder containing manifest.json.
Requires Chrome 148+; models and runtimes are included.

To update, replace files in the same extension folder, click Reload, and
refresh website tabs. Keeping the folder preserves settings and the unpacked
extension identity. Watch preparation progress in the popup.

Developers: sitr-0.5.1-source.zip is the matching source;
sitr-0.5.1-models.zip contains pinned models, original export inputs, and
licenses. Use the new model ZIP: 0.5.0 BodyPix files have different output
heads and hashes and cannot build this revision. Publish all three ZIPs and
SHA256SUMS.txt together under tag v0.5.1.

Performance and validation

Preparation took approximately 18–22 seconds on the local test machine.
Model lookup/loading overhead afterward was below 2 ms in tested first
frames and setting changes. Total inference still varies; identical first/last
image latency is not guaranteed. Retaining every model increases idle RAM/GPU
memory. Restart, extension reload, or engine recovery needs preparation again.

Validation covers TypeScript, 167 unit tests, the browser regression suite,
English/Arabic UI, live progress and recovery, modest clothed male anatomy
fixtures, model hashes, documentation, and release archives. Full details are
in docs/VALIDATION.md in the source ZIP.

Advanced segmentation remains image-only and off by default. Anatomy and
appearance estimates can be wrong; filtering does not guarantee complete
coverage or zero-frame exposure. Meet checks use local WebRTC fixtures rather
than an actual meeting. Media processing stays local, with no account, analytics,
or inference service. Original Sitr files use AGPL-3.0-only; third-party
assets retain their documented licenses.

Full Changelog: v0.5.0...v0.5.1

Sitr v0.5.0 — Choose what to cover

Choose a tag to compare

@tantour tantour released this 14 Oct 12:12

The second Sitr release makes coverage choices independent for Female and Male.
Use the interactive person diagram to show or hide image areas, with separate
face and hair choices. Blur and Pixelate reveal face strength directly below
the buttons, independently for each gender.

New and improved

  • Optional Advanced segmentation for images, powered by local BodyPix models.
    Fast adds less processing; Quality is slower with finer detail. Body skin
    still comes from MediaPipe, and video never runs the anatomy model.
  • Individual torso, upper-arm, forearm, hand, thigh, lower-leg, foot, neck, hair,
    and face choices. Torso combines chest and belly; neck boundaries are approximate.
  • Correct coverage on both upper arms and forearms. Shown people and areas resist
    mask expansion, fixing unwanted hair/neck coverage and black borders.
  • A wider popup, visible Cover area controls, and clearer English/Arabic UI.
    Skin-exposure rules and the shared face-effect mode are removed. Existing
    settings migrate; saved exposure rules are disabled.
  • Body-region colors, backend/status/timing diagnostics, cached policy redraws,
    and fallback when an optional model fails.
  • Live stream/track handling and rotating analysis of Google Meet tiles and
    presentations, with controls applying only to your local view.
  • Updated screenshots, user guide, provenance, and fifteen-model release bundle.

Download and install

Download sitr-0.5.0-chrome.zip and SHA256SUMS.txt. Verify the checksum,
extract the ZIP into a permanent folder, then open chrome://extensions, enable
Developer mode, and use Load unpacked on the folder containing manifest.json.
Requires Chrome 148+. All models and inference runtimes are included.

For an existing installation, replace the files in the same extension folder,
click Reload, and refresh website tabs, including Meet. Keeping the same folder
preserves the unpacked extension identity and settings.

sitr-0.5.0-source.zip is the matching source; sitr-0.5.0-models.zip contains
the verified weights and original export inputs for rebuilding. The old 0.4.2
model bundle cannot build this version. Publish all three ZIPs and the checksums
together under tag v0.5.0.

Validation

TypeScript, all 148 current unit tests, all 15 browser checks, model hashes,
release structure, and documentation links passed locally. npm audit reported
zero vulnerabilities. Browser checks include CSS backgrounds, both segmentation
variants, cached edits, independent face strengths, video compatibility, and
local WebRTC Meet fixtures.

Performance and limits

Advanced segmentation is off by default. In the measured warm WebGPU fixture on
an Intel i5-8250U/UHD 620, its added stage was approximately 34 ms for Fast
and 94 ms for Quality. Both models together add roughly 9.84 MB of packaged
weights. Cold startup can take seconds; actual results depend on the device and
image. Area and face-strength edits reuse cached inference.

Overlapping parts and appearance estimates can be wrong. This release does not
guarantee complete coverage or zero-frame exposure. Meet regression tests use
local WebRTC fixtures, rather than an actual meeting. See the README, user guide,
and benchmark document for supported media and practical limits.

Sitr processes media locally, with no account, analytics, or inference service.
Original project files remain licensed under AGPL-3.0-only; bundled third-party
models and dependencies retain their documented licenses.

Full Changelog: v0.4.2...v0.5.0

Sitr v0.4.2 — Private image and video filtering

Choose a tag to compare

@tantour tantour released this 06 Oct 06:44

Sitr · ستر is a Chrome extension that covers selected people in images and videos using AI processed locally on your device.

Features

  • Separate controls for images and videos.
  • Skin and hair coverage for images, plus whole-person coverage.
  • Black masks, blur, and pixelated effects.
  • English and Arabic interfaces, with light and dark themes.
  • Website exceptions, advanced controls, and diagnostics.
  • No accounts, analytics, or inference server.

Installation

Requires desktop Chrome 148 or newer. Other Chromium-based browsers may work but have not been verified. Firefox, Safari, and mobile browsers are currently unsupported.

  1. Download sitr-0.4.2-chrome.zip and SHA256SUMS.txt.
  2. Verify the ZIP’s SHA-256 checksum, then extract it into a permanent folder.
  3. Open chrome://extensions and enable Developer mode.
  4. Click Load unpacked and select the extracted folder containing manifest.json.
  5. Refresh your website tabs and open Sitr to choose your settings.

All required models are included in the Chrome ZIP.

Downloads

  • sitr-0.4.2-chrome.zip — ready-to-load extension.
  • sitr-0.4.2-source.zip — corresponding source code.
  • sitr-0.4.2-models.zip — model bundle for building from source.
  • SHA256SUMS.txt — archive checksums.

Known limitations

AI detection can miss people or misclassify appearance. Video coverage may lag behind movement, and some protected or unreadable media is unsupported. Complete coverage is not guaranteed.

See the README for supported media, privacy details, and configuration guidance.

Licensed under GNU AGPL v3; bundled third-party components retain their respective licenses.