BREAKING
HeyGen Agent Triples VAE Decoder Speed
0x
Inference speedup
0s
fp32 baseline
0s
Optimized time
Fewer Trials, Better Result
Manual trials105
Auto-Optim trials18
How the Speedup Was Built
fp32 eager14.47
bf168.42
+CUDA graph7.42
break-free4.82
final4.79
Quality Gate Held the Line
Quality gate61.3 dB
Parity vs fp32 reference
Cleaner PSNR output
Stop condition99.8% GPU
8 rounds no gain
Convolution-bound limit
Toolkit Released on GitHub
AI NEWS BLITZ
HeyGen says an autonomous AI agent tripled its video decoder speed on its own.