3 ms·
Their models range from 70mb to 3gb. The largest model is smaller than the optimised stable diffusion. Not sure what the inference speed is like, haven't tried
by StevenWaterman 4y ago
Their models range from 70mb to 3gb. The largest model is smaller than the optimised stable diffusion. Not sure what the inference speed is like, haven't tried it myself yet.
- IceWreck 4y agoI just tested it myself. Its fast enough on colab, couple of seconds but not sure if its fast enough to transcribe realtime audio yet.
- MacsHeadroom 4y agoColab is using one of the larger models. Tiny probably runs in realtime on a single core of an RPi.
- lynguist 4y ago"small" runs in realtime on Macbook Air M1 CPU.