PixelOkay
Browser workspace
Tools
HomeVideo ToolsContinuity Checker
✦ FILMMAKING INTELLIGENCE / SHOT ANALYSIS

Continuity Checker

Compare single shots or scan adjacent frames for possible changes in wardrobe, props, character blocking, lighting and scene details. Review suggestions visually, mark specific regions and create a continuity report.

◉ Local frame analysis◎ No account required⇩ Exportable review
01 / WORKSPACE

Continuity workspace

Start with your footage or explore a built-in example.

Drop frames or video clips here

Choose multiple images, or capture frames from videos. Nothing is uploaded.

PNG · JPEG · WEBP · GIF snapshot · browser-supported videos
Shot library 0 frames
Select any two frames below to compare · Rename reference A in shot details
Your frames will appear here. Add at least two to begin.
Shot detailsAuto-save scene, setup and take for reference A. Switch A to edit other frames.
Sequence review

Scan consecutive frames in library order with local visual checks. Keeps existing notes; optional AI detection is available for individual pairs only.

0 open across 0 pairs
02 / COMPARISON VIEW

Frame inspection

AReferenceNo frame selected
Add a frame
–
BComparisonNo frame selected
Add a frame
–

Run analysis to see alignment information.

i

Automatic differences are suggestions, not verified errors. Differences in camera angle, movement and intended creative changes can produce false alerts. Select “Different angles” to avoid pixel-level suggestions across dissimilar compositions.

Region inspector PRO REVIEW

Draw a matching region over each shot to compare costume colors, props or set details.

Draw a rectangle on shot A, then switch to B to draw its matching location. You can also add a region using the coordinates below.
AB
No custom regions added.
HOW IT WORKS

A faster continuity pass, with you in control.

Designed for editors, script supervisors, student filmmakers and creators reviewing adjacent takes or reshoots.

01

Import frames

Drop stills or video clips, seek the right moment and capture frames into your local shot library.

02

Compare intelligently

Check individual frames or scan adjacent pairs in order. Optional translation-only alignment reduces slight camera-shift noise; object detection is available for single pairs.

03

Review and hand off

Mark your own wardrobe or prop regions, confirm or dismiss findings, add notes and export a review-ready report.

FREQUENTLY ASKED

Questions before you start

Does the tool recognize the same costume or actor in every shot?

Not reliably. A free general-purpose object detector can identify people and some props, but it cannot establish character identity or identify exact costume items. Wardrobe suggestions use image-region color comparisons and require a human to review.

Will footage leave my computer?

The tool reads media locally. Optional AI object detection downloads TensorFlow.js and COCO-SSD assets from public CDNs after you opt in. A separate optional WebGPU text model downloads WebLLM assets and interprets only measured findings, never image pixels. Inference runs locally, but the CDN receives asset requests. Without AI, frame comparison works without models.

What does auto-alignment do?

For similarly framed shots, an optional translation-only check searches for small horizontal and vertical offsets. It does not fix perspective, zoom or moving subjects, and it does not reposition the preview. You can disable it or use Different angles when shots are dissimilar.

Can I scan a complete sequence?

Use Scan adjacent pairs to compare neighboring frames in shot-library order. The scan runs local visual heuristics sequentially (no optional AI downloads), preserves existing review notes and can be canceled. It does not automatically decide whether adjacent shots depict the same moment. Scene metadata can restrict known cross-scene pairs.

Can I compare different camera angles?

Yes. Choose “Different angles / framing” to suppress misleading pixel change maps. Matching region boxes can be drawn separately on both shots. Object and person detection is still heuristic and not proof of continuity.

Can I save or export findings?

Export a review-only JSON report, CSV findings, or a print-to-PDF report. Portable project JSON also embeds reduced-resolution frame previews so you can reopen a review without separately re-adding the media. Portable files may contain sensitive production imagery. No workspace data is silently uploaded or saved to a server.

LOCAL AI SETUP

Set up AI on this device

First, a model needs to be loaded on your device

AI continuity review uses a WebGPU language model locally. On first use, its files may require a large download. Shot pixels are not sent to the language model; only the current pair’s candidate findings and limited production metadata are provided as text.

  1. 1. Check device
  2. 2. Download / cache
  3. 3. Prepare AI continuity review
  4. 4. Review continuity

Device & model resources

Measurements may be approximate
Device RAM / model memoryChecking device information…

Live system RAM, model RAM and GPU memory are not reliably exposed in browsers.

Browser website storageChecking storage…

Site-wide estimate; includes other PixelOkay tools and is not RAM.

ONE ACTION · ALL RECOGNIZED AI MODEL DOWNLOADS

Free AI model storage across PixelOkay

Scan this website origin for WebLLM and Transformers.js model caches, plus recognizable model files in other cache buckets. This includes caches from other PixelOkay tools, not only this tool. No site data or browser databases are cleared automatically.

Run a scan to see how many model caches can be cleared together.

Browser-native models and unknown IndexedDB/OPFS or private runtime caches may remain. Unknown application/user storage will not be deleted. Other running tools might redownload caches; close their tabs first.

Having trouble downloading or loading AI?

Old or incomplete AI model downloads may be using your browser's storage. First refresh the storage meter or try a smaller model. If setup still fails, inspect and remove recognized AI model caches with confirmation, then download again.

Only model caches are targeted. Other PixelOkay tools may need to redownload their models. Never clear all website data without exporting important chats and projects first.

Select “Check device” to see whether the current model can run.
AI is optional. Frame comparison remains available. COCO-SSD object detection is a separate opt-in model loaded from the Analyze shots workflow, not this text-model selector.
CROSS-TOOL MODEL CACHE CLEANUP

Delete all recognized AI model downloads?

Deleting these caches will cause other PixelOkay tools to download their models again. Loaded AI sessions in other tabs may be disrupted. Unidentified model caches and browser-managed models may remain.

Data-loss caution: This action does not intentionally delete saved chats, projects, drafts, localStorage or IndexedDB. Do not clear website data unless you have backed up work; clearing all website data may permanently erase saved work.

Using Continuity Checker

Compare selected film frames or nearby shot pairs to surface visual changes for a human continuity review. Record scene, shot and take details, then decide which candidates deserve follow-up.

  1. Import or capture reference frames and add the scene, shot and take details that help organize the comparison.
  2. Compare selected frames or scan adjacent pairs; use the region view and limited alignment controls to inspect visible differences.
  3. Review each candidate against the footage, record your decision and export a review report or portable project.

Example: Compare two close frames of a tabletop across takes, inspect a suggested difference near a prop, and record whether it needs follow-up after checking the original shots.

Limits and privacy: Findings are visual candidates, not verified continuity errors or actor/prop identity. Analysis uses a 320 × 180 center crop with translation-only alignment. Optional object detection and text-summary models retrieve external files after setup and may disclose the visitor IP; the text model summarizes existing finding text and does not view frames. Portable projects include compressed previews; reports omit the source video. Manual frame comparison/export was tested, but optional detector inference and real continuity-detection quality remain unverified.

Related tools