GradiumBENCHMARKS

Data read on November 19, 2026

Gradium TTS vs Deepdub dd-etts-3.3

Side by side on the four measures Coval publishes for every model. Lower is better on all four.

Gradium TTSDeepdub dd-etts-3.3lower
Median time to first audio64 ms56 mswithin 10 ms
P95 time to first audio (slow turns)172 ms86 msDeepdub dd-etts-3.3
Leading silence (median)0 ms2 mswithin 10 ms
Word error rate (mean)1.7%1.6%Deepdub dd-etts-3.3

Measured by Coval from US East on the same day for both models. A gap under 10 ms is within day-to-day variation.

Source : Coval · Read on November 19, 2026 · gradium-tts-beta-202609 · dd-etts-3.3