I just bought Vikpea yesterday and my first run is with TensorRT It is very fast. Faster than other AI Videoenhancer apps I’ve used. However, I’m not sure if TensorRT reduces the enhancement quality as usually you sacrifice quality for speed.
You have to enable it in the export queue screen and it only takes effect during the next processing not the current batch, which is not intuitive at all. Also only the Generative Restoration model is fast. The rest are very slow. But the Generative Restoration model is really good especially for 1x enhance, where it beats every single Topaz model I tried for 4k videos. At 2x upscale though, Topaz models are a bit better. But it depends on the video.
TensorRT is a model inference acceleration framework provided by NVIDIA. It significantly improves GPU performance and speeds up video enhancement model processing while maintaining video quality. If your graphics card supports it, please enable this feature for even faster processing speeds.
Yes, I had noticed that. And clicked on it. And it say’s “Enabling TensorRT will take effect when you process the next task”. So just clicking on it when it’s GREEN turns it “ON”; and its greyed out when it is “ON”?