Objective Quality Benchmarks (VMAF / SSIM)
Radical honesty. We benchmark actual streams received over WhatsApp networks.
Scientific Integrity Principle:
Unlike proprietary software that invents fictional quality percentage increases for marketing hype, WhatMedia refuses to display fabricated scores. Until our comprehensive testing corpus is fully finalized and open-sourced, all metrics display "Mesures en cours" (Measurements in progress).
1. End-to-End Evaluation Protocol
To measure real-world performance, we do not compare local test files: we measure the video as received on the viewer's device after traversing WhatsApp's cloud ingest pipeline.
Master Reference
Uncompressed 4K ProRes/HDR master clip
Status Ingest
Posted to real WhatsApp status via test rig
Stream Capture
Extracted from viewer recipient device
VMAF Calculation
Frame-by-frame scoring via Netflix libvmaf
2. Standardized Reference Test Corpus
| File / Scene | Source Spec | Direct Upload | With WhatMedia | Status |
|---|---|---|---|---|
| clip_01_hdr_sunset_landscape.mov | Landscape 4K HDR (Dolby Vision 60fps) | Mesures en cours | Mesures en cours | Testing |
| clip_02_talking_head_indoor.mp4 | Indoor face 1080p 30fps SDR | Mesures en cours | Mesures en cours | Testing |
| clip_03_sports_fast_motion.mp4 | Action sports 4K 60fps SDR | Mesures en cours | Mesures en cours | Testing |
| clip_04_night_neon_city.mov | Low-light urban 4K HDR | Mesures en cours | Mesures en cours | Testing |
| clip_05_screencast_code_text.mp4 | Fine text & UI 1080p 60fps | Mesures en cours | Mesures en cours | Testing |
3. Open Source Reproducibility
The entire benchmark toolchain is open source under tools/benchmarks. Any
video engineer or researcher can inspect our methodology or reproduce results with their
own test vectors.