← Back to scoreboard
Tavus
closed · proprietaryphoenix-1 (CVI, us-east, tested 2026-04-30)last tested 2026-04-30T14:00:00Z
endpoint · Conversational Video Interface (CVI), model phoenix-1, region us-east
homepage · https://www.tavus.io
All six probes vs. field median
Values normalized to [0,1]. IFC normalized as 1 − |v − 1|.
Raw scores
| Probe | Value | Unit |
|---|---|---|
| GR — Generation Rate | 0.021 | ratio |
| XPD — Cross-Run Pixel Diversity | 0.012 | embedding-dist |
| AVC — Audio→Motion Causality (off-mouth) | 0.180 | pearson-r |
| LWA — Listening-Window Activation | 1.080 | ratio |
| EVD — Emotional-Valence Differentiation | 0.310 | fisher-ratio |
| IFC — Idle-Loop Compression | 0.180 | ratio-vs-human |
Frame strips
Identical-input replays. Hashes computed by the harness.
support-package-lookup
✗ identicalComposite history
Disclosure
Tested via the public Conversational Video Interface API. Region: us-east. Two identical-audio runs returned bit-identical video hashes.