Webfilmland QC User Guide
Webfilmland QC is a desktop quality-control application for video, audio, caption, and delivery preflight checks. It creates audit-ready HTML, PDF, and JSON reports for technical review and customer delivery evidence.
This guide describes the customer release build distributed as an Apple Developer ID signed, notarized, and stapled macOS .pkg installer. Windows customer releases are distributed as a compiled .exe artifact, not as a source-code archive.
Current Release Evidence
The current macOS customer package is published at:
https://license.webfilmland.com/download/mac- Version:
1.0.26 - Build:
100 - Installer:
Webfilmland_QC.pkg - SHA256:
614bd1d33b0aa041cbc0e69062af85ad26c440300fdf3d98afd754c65656ee79
Release checks completed for this build:
- macOS PKG signed with Developer ID Installer, notarized by Apple, stapled, and accepted by Gatekeeper.
- Unit test suite passed: 27 tests.
- Golden media regression passed: 23 known-good and known-bad QC cases.
- Spec-source audit passed: 26 sources across 14 delivery presets.
- Environment check passed with FFmpeg, FFprobe, OpenCV, PySide6, NumPy, Torch, audio/caption libraries, and PDF reporting available.
Product Positioning
Webfilmland QC is a professional preflight QC assistant for indie films, web masters, marketplace delivery, archive review and post-production teams. It is designed to catch common technical problems before upload, client delivery, archive handoff or external certified QC.
It is not marketed as a legal replacement for Baton, Vidchecker, Harding or a broadcaster-certified final QC workflow. Internal detectors and FFmpeg-based checks provide advisory evidence, while certified results must come from certified external engines where required by the delivery contract.
What Webfilmland QC Can Do
Webfilmland QC can help a professional team find common technical problems before a master is uploaded, delivered to a client, archived, or sent for external certified QC.
Core capabilities:
- Run native desktop preflight checks for video, audio, captions, metadata, and delivery packages.
- Use a native C++ video scanner as the primary video QC engine, with Python handling orchestration, delivery presets, reports, review workflows, and AI advisory checks.
- Check resolution, frame rate, scan type, aspect ratio, codec/container metadata, audio structure, sample rate, bit depth, loudness, true peak, clipping, silence/dropouts, captions, and package sidecar structure.
- Detect or flag visual candidates such as freezes, black frames, flashing/PSE advisory events, thin edge padding, baked bars, output blanking, image shifts, legal-range/YUV issues, HDR metadata issues, and continuity/jump-cut candidates.
- Select source-linked delivery presets for Amazon, Apple/ProRes, EBU R128, Netflix/IMF, generic IMF, web/platform masters, AS-11 UK DPP, IMF Application DPP, ARIB Japan Broadcast, Filmhub, and DCP-oriented checks.
- Generate HTML, PDF, and JSON reports with run ID, app version/build, selected preset, thresholds, official source references, scan coverage, findings, snapshots, review decisions, and recheck evidence.
- Resume long-form work from cancellation checkpoints where supported, so a long scan does not always need to start again from the beginning.
- Separate deterministic measured findings from advisory AI findings and from imported certified evidence.
Good uses:
- Catching obvious delivery blockers before a client sees them.
- Checking export settings before upload to a distributor, platform, archive, or broadcaster-side review process.
- Creating technical evidence for editors, producers, colourists, and post supervisors.
- Preparing focused notes before sending material to a certified QC provider.
Limits:
- It does not guarantee acceptance by Netflix, Amazon, Apple, broadcasters, or distributors.
- It does not replace contractual delivery specifications supplied directly by a client or platform.
- It does not produce a certified PSE PASS unless a recognized external PSE report is imported and accepted as certified evidence.
Downloads
Stable customer download links are served from the Webfilmland license server:
- macOS PKG installer:
https://license.webfilmland.com/download/mac - Windows EXE:
https://license.webfilmland.com/download/windows - Release manifest with version, file size and SHA256:
https://license.webfilmland.com/api/releases/latest
The macOS link points to the latest published signed, notarized, stapled PKG installer. The Windows link points to the latest published compiled .exe after a Windows build has been created and published.
System Requirements
- macOS 11 or newer
- Apple Silicon Mac recommended
- Enough free disk space for source media, reports, and QC snapshots
- Local or fast external storage for long-form HD/UHD masters
The signed app includes its required runtime dependencies, including OpenCV (cv2), FFmpeg/FFprobe, PySide6, NumPy, audio/caption libraries, report generation libraries, and the bundled AI/QC runtime.
Installation
- Download
Webfilmland_QC.pkgfrom the license server. - Open the PKG installer.
- Follow the installer prompts.
- Open
Webfilmland QCfrom Applications.
Because the customer build is Apple Developer ID signed and notarized, macOS Gatekeeper should allow normal launch without unsafe bypass steps.
Windows Release
Windows customers receive a compiled Webfilmland_QC-<version>-build-<build>-windows.exe. Do not distribute the repository, Python files, virtual environment, build folder, or source tree to customers. The executable is built with PyInstaller in one-file/windowed mode so the release artifact is an application executable rather than a folder of source files.
Python desktop apps cannot be made impossible to reverse engineer, but customer builds should avoid shipping readable source files. For stronger source-code protection later, use a native compiler/obfuscation build such as Nuitka commercial mode or a compiled extension boundary for sensitive detector logic.
First Launch
At startup, Webfilmland QC checks that the packaged runtime is complete. If a required component is missing or damaged, the app shows the exact problem before starting QC.
During development or source-based use, the checker can ask before installing missing Python packages. In the signed customer PKG, required dependencies are expected to be bundled already.
Basic Workflow
- Open the app.
- Choose the QC mode:
Video FileAudio FileAudio Stems Folder- Load the source media.
- Optional: load a caption/subtitle file.
- Select the delivery preset.
- Choose a scan mode.
- Run QC.
- Review Live QC progress while the job runs.
- Open the generated HTML/PDF/JSON report.
Scan Modes
Fast Mode
Use Fast mode for quick triage during editing, conform, or early delivery prep.
- Samples video frames
- Runs deterministic metadata, audio, package, and caption checks
- Faster than Final mode
- Can miss isolated one-frame visual defects between sampled frames
Best use: early checks, quick customer screeners, obvious technical mistakes.
Deep ML Mode
Use Deep ML mode when you want sampled deterministic QC plus advisory AI review.
- Samples deterministic video frames
- Runs advisory AI visual/audio/caption checks
- More detailed than Fast mode
- Still not frame-accurate for every visual defect
Best use: stronger preflight before final export or before human QC review.
Final Full Precision Mode
Use Final Full Precision before delivery sign-off.
- Deterministic video checks inspect every decoded frame in the selected range
- Audio, package, captions, metadata, and IMF checks run deterministically
- AI visual checks remain sampled and advisory for speed
- Duplicate AI audio/caption advisory analysis is skipped because deterministic audio and caption QC already ran
Best use: final delivery preflight, long-form masters, customer-facing QC evidence.
Long-Form Runtime Expectations
QC speed depends on codec, resolution, disk speed, enabled checks, CPU, and whether the media is HD, UHD, HDR, IMF, ProRes, H.264, HEVC, or MXF.
General expectations:
- HD Fast mode should usually be faster than real time.
- HD Deep ML mode may be near real time depending on checks.
- HD Final Full Precision should be treated as a full technical pass and may take close to the media runtime or longer.
- UHD/4K, HEVC, HDR, IMF, PSE, and heavy perceptual checks can take much longer.
For a 90-minute film, use Fast or Deep ML during development and Final Full Precision only for final sign-off or selected problem ranges.
Live QC Progress
The Live QC panel prints what the app is doing:
- Startup/runtime checks
- Metadata checks
- Package/sidecar checks
- Audio checks
- Video frame worker progress
- YUV legal-range progress
- AI advisory sampling progress
- Caption checks
- Report generation
If the app is running a long Final Full Precision job, Live QC should continue showing progress so the operator can see that analysis is active.
Delivery Presets
The app includes source-linked delivery presets for:
- Amazon Prime / Mezzanine
- Apple / ProRes Master
- EBU R128 Broadcast
- Netflix / IMF Delivery
- Generic IMF Package Preflight
- Web / Platform Master
Reports include the selected preset, thresholds used, and official/public or partner-official spec-source references where available.
Platform requirements can change. For contractual delivery, always confirm the active client or platform specification before final acceptance.
PSE Certification Policy
PSE findings are deliberately separated into advisory and certified evidence:
- Internal PSE checks:
PSE advisory: non-certified - FFmpeg photosensitivity checks:
PSE advisory: non-certified - Apple/IRIS-style/free/open checks:
PSE advisory: non-certified - Imported certified external evidence: can show
Certified PSE: PASSorCertified PSE: FAIL
The app only shows Certified PSE: PASS when an imported external PSE report from a recognized professional engine, such as Harding, Baton, Vidchecker, Telestream or Aurora, confirms PASS. PEAT or unknown/free tools can be imported as review evidence, but they do not create a certified PASS claim.
Trial and Data-Sharing Option
Paid licenses are issued after order review and payment confirmation. A 90-day research trial may be issued when the user explicitly consents to share selected QC data for detector training and product improvement.
Data sharing must be opt-in. Shared material may include QC reports, measured values, snapshots, review decisions and selected media samples only when the user has the legal right to share that material. Trial users can be asked to remove or anonymize confidential client data before sharing.
IMF Package Preflight
For IMF package checks, select an IMF preset and load the package folder.
The app checks:
- AssetMap presence and references
- PKL presence and references
- CPL presence and TrackFileId references
- CPL timeline math
- OPL references
- MCA audio labeling hints
- Asset size/hash checks where possible
- Optional XSD validation when schema support is available
This is an IMF preflight, not a replacement for a formal studio/platform certification process.
Reports
After QC, the app saves:
- HTML report for review
- PDF report for sharing/archival
- JSON report for integration and audit
Reports include:
- PASS/WARN/FAIL summary
- Delivery preset and version
- Official spec-source evidence
- Scan mode and confidence
- Frame coverage
- Thresholds used
- Findings with snapshots where available
- Review / false-positive evidence when marked
- App version/build and run ID
Reviewing Findings
Use FAIL findings for delivery-blocking issues unless a qualified reviewer marks them as accepted. WARN findings usually require review but may be acceptable depending on the content and delivery spec.
During development and pilot use, keep notes on:
- Correct detections
- False positives
- Missed issues
- Accepted exceptions
- Customer-specific delivery rules
This evidence improves future presets and helps build trust with buyers.
Support Bundle
If a customer reports a crash, wrong result, or unexpected behavior, create a support bundle from the application support tools or by running the included support-bundle command in the development environment.
The support bundle is designed to include:
- App logs
- Crash logs
- Recent reports
- Validation evidence
- Platform metadata
It does not include source media by default.
Troubleshooting
App Does Not Open
Confirm you are using the signed/notarized customer PKG. If macOS still blocks launch, download the PKG again and make sure it was not modified after signing.
QC Seems Slow
Check Live QC. Long Final Full Precision runs can be slow because every decoded frame is inspected. For early review, use Fast or Deep ML mode.
Missing Dependency Message
In a signed customer build this usually means the app package is damaged or an incomplete build was distributed. Reinstall from the latest signed PKG.
Report Does Not Open
The report is still saved on disk even if automatic opening fails. Use the path shown in Live QC or the report preview.
Best Practice
- Use Fast mode during editing and early delivery prep.
- Use Deep ML mode when you want broader advisory review.
- Use Final Full Precision for delivery sign-off.
- Keep the generated JSON report with delivery records.
- Use PDF/HTML reports for customers.
- Review AI findings manually; AI is advisory, not certification.
- For paid delivery, confirm the active customer/platform spec.