Artemis · live demo

The same speech model, before and after Artemis changed its code.

Speak into either tab. Both run the identical model on the identical hardware, and the only difference is one code change Artemis found on its own.

0%
less processing, waiting for speech
Baseline
0ms
0 passes
After Artemis
0ms
0 passes
Processor time spent transcribing, accumulating as you speak. Lower is better.

Baseline transcript

Waiting for speech

After Artemis arrives in larger chunks

Waiting for speech
Open both tabs and press record to begin