AI Video Quality Reviewer
Upload 5 screenshots from your AI-generated video. Claude scores lip sync, face consistency, halo stability, and background. Returns OK / Adjust / NG verdict with Hedra baseline comparison. No video file needed.
평점
평점 없음
판매
1
사용 방법
Capafy에서 실행
퍼블리셔 제공
Claude Sonnet 4.6
이 Agent를 사용할 때마다 퍼블리셔가 제공한 LLM에서 실행됩니다.What is AI Video Quality Reviewer?
Analyze the quality of your AI-generated videos without uploading the video file. Simply take 5 screenshots and get a detailed quality report with actionable insights.
How to Use
- Take 5 screenshots from your AI-generated video (spread across the full duration)
- Upload them in this chat
- Get a complete quality score with specific suggestions
Recommended capture points: 10% · 30% · 50% · 70% · 90%
Scoring System
| Item | Criteria |
|---|---|
| Lip naturalness | Syllable-matched movement, no stiffness |
| Face consistency | Stable eyes, nose, skin tone across frames |
| Halo stability | Crown glow stable — no drift or distortion |
| Background consistency | No color/shape jumps between frames |
| Loop suitability | First and last frames connect naturally |
Each item scored 1–5. Verdict: ✅ OK / ⚠️ Adjust / ❌ NG
Hedra Baseline Comparison
Your video is benchmarked against Hedra v3 (overall ≈ 4.0) so you know exactly where you stand.
| Metric | Hedra v3 | Hedra v5 |
|---|---|---|
| Face consistency | 4.0 | 5.0 |
| Lip movement | 4.0 | 5.0 |
| Background stability | 4.0 | 4.0 |
Supported Video Types
- Lipsync — character speaking with audio
- Idle — character with breathing/micro-movement
- General — any AI-generated video content
Requirements
- Minimum 3 screenshots (5 recommended)
- PNG, JPG, or WebP format
- No video file upload needed
Smoke Test Example
Input: Upload 5 screenshots from a Hedra-generated lipsync video (PNG, one per 20% of duration)
Expected output:
📊 Quality Report
- Lip naturalness: 4/5 ✅
- Face consistency: 4/5 ✅
- Halo stability: 3/5 ⚠️ Adjust
- Background consistency: 5/5 ✅
- Loop suitability: 4/5 ✅
Overall: 4.0 / 5.0 ✅ OK
Benchmark: On par with Hedra v3 baseline
Try it: Take any 5 screenshots from an AI video and upload them. You will receive a full scored report within seconds.
