Every Check Auto QC Runs — Explained
15 automated quality checks across video, audio, AI detection, and platform compliance. Here's exactly what gets inspected when you upload a video.
Video Quality Checks
Scans every frame for visual issues before they reach your audience.
VIDEO / ARTIFACT
Black Frame, Flash Frame & Bad Edit Detection
Three signal failures caught in a single pass. Detects unexpected black screens anywhere on the timeline, rapid light flashes that violate broadcast safety thresholds, and jarring cuts where an ultra-short scene (≤5 frames) is sandwiched between two significant scene changes.
FREEZE FRAME
Freeze Frame Detection
Identifies sequences where motion has completely stopped. A three-stage cascade — signal detection, motion vector analysis, and entropy-based classification — distinguishes true freezes from low-contrast or slow-moving scenes with zero false positives.
ASPECT RATIO
Aspect Ratio Check
Verifies the video's pixel and display aspect ratio against your chosen platform. Catches mismatched DAR/SAR values that cause letterboxing, pillarboxing, or distorted playback on the target device.
BORDER
Border Detection
Detects persistent black bars — letterbox (top/bottom) or pillarbox (left/right) — that shouldn't be present in the final deliverable. Frame-accurate start and end timecodes are provided for each occurrence.
BRIGHTNESS
Brightness Analysis
Flags sequences that fall outside safe luminance ranges — identifying overexposed highlights or crushed underexposed shadows in both SDR and HDR content.
BLUR
Blur Detection
Measures per-frame sharpness using Laplacian variance and flags sequences that fall below your clarity threshold — catching focus pulls, motion blur, or encoding degradation.
Audio Quality Checks
Speech-aware analysis across the full mix — every audio check is enhanced with voice activity detection.
LOUDNESS
Loudness Compliance
Measures integrated program loudness against broadcast and streaming audio standards — EBU R128 (−23 LUFS) for European broadcast, ATSC A/85 (−24 LUFS) for North America. Speech-gated analysis ensures dialogue drives the meter, not music or effects.
CLIPPING
Audio Clipping
Detects true-peak overloads that cause audible distortion. Voice activity detection intelligently skips speech segments during analysis to avoid flagging intentional dynamic peaks.
NOISE
Background Noise Analysis
Samples ambient noise exclusively during non-speech segments to calculate an accurate signal-to-noise ratio — identifying hiss, hum, or room noise that could affect perceived quality without contamination from the dialogue itself.
SILENCE
Silence Detection
Flags unexpected audio gaps where the signal drops below threshold — catching missing audio tracks, edit errors, or dropout events with frame-accurate timecodes.
SPEECH
Speech Clarity Analysis
Multi-dimensional perceptual evaluation that automatically classifies the background environment — music, ambient noise, or silence — then scores speech clarity accordingly. Music backgrounds are scored by vocal prominence over the mix; noise backgrounds are scored by SNR. UGC-adapted thresholds ensure music-backed speech isn't penalised unfairly.
AI Visual Detection
Powered by Auto QC's PlanD Engine — a hybrid pipeline combining GPU-accelerated OCR and large language model analysis to catch errors no signal processing can find.
AI / OCR / LLM
Subtitle & Text Check
The PlanD Engine scans every scene containing on-screen text — subtitles, lower thirds, titles, and graphics. It detects errors that only become visible when you can read the content.
TYPO DETECTION
Context-aware spell check across 109 languages. Catches misspellings, brand name errors, and punctuation inconsistencies in subtitles and graphics.
SAFE ZONE CHECK
Verifies that subtitles and overlay graphics stay within platform-defined safe zones. Includes presets for YouTube Shorts, Instagram Reels, TikTok, and more.
SUBTITLE EXTRACTION
Extracts all hard-coded subtitle text with ≤120ms timestamp precision. Uses hybrid sampling — 1fps baseline plus ROI-dense sampling for high-activity scenes.
SUBTITLE REFINEMENT
Uses LLM semantic analysis to clean and improve extracted subtitle quality — correcting OCR artefacts and improving readability.
Logo & Brand Detection
Scans every sampled frame for brand logos and on-screen products. Potential copyright and endorsement risks get a timecoded marker before you deliver — no configuration required.
AI / VISION
Frame-by-frame rights risk scanning
Powered by Grounding DINO — a zero-shot object detection model that identifies brand logos and on-screen products without pre-training on specific brands. Results appear in the same QC timeline alongside all other detected issues.
LOGO COPYRIGHT RISK
Detects brand logos and trademarks that may require licensing clearance. Amber timeline markers with a dashed bounding box that tracks the logo across frames in the video player.
BRAND PRODUCT RISK
Flags full branded products in frame — laptops, phones, packaging — that could imply brand endorsement. Orange markers, separate from logo copyright flags.
Platform Compliance
5 built-in platform presets. Select your delivery target and Auto QC verifies format, encoding, loudness, and aspect ratio in a single pass.
YouTube
Resolution · Aspect ratio · Frame rate · Loudness
TikTok
Vertical format · Codec · Loudness
Aspect ratio variants · Codec · Loudness
Broadcast
Codec · Loudness · Frame rate · Full spec
Custom
Fully configurable
Every Issue, Pinpointed on the Timeline
QC results aren't just a list — they're mapped to exact timecodes, so you can jump straight to the problem and fix it fast.
Every detected issue is annotated with its exact start and end timecode
Click any marker in the timeline to jump the player to that moment
Export a broadcast-grade PDF report for client sign-off and delivery documentation