Getting Error when i click on Start Button

[chnisar515]

I have fresh installed Roop-unleashed in Winows 10 Pro 64Bit version fully updated. i have Core i7 2.40Ghz 2.40Ghz 16GBRAM and GPU Intel(R) HD Graphics 5500 and GPU 1 NVIDIA GeForce 940M. i also installed fresh Python-3.12.2-amd64 and installed VisualStudio after that i downloaded zip file of Roop-unleashed from here and run the bat file after downloading all the things roop just launched in my default browser Chrome and there i upload the source image and also uploaded the target file which is 36MB video MP4 file and then i click on Start button and after few seconds i got these errors.

Error
[ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Exception during initialization: D:\a_work\1\s\onnxruntime\core\providers\cuda\cuda_call.cc:121 onnxruntime::CudaCall D:\a_work\1\s\onnxruntime\core\providers\cuda\cuda_call.cc:114 onnxruntime::CudaCall CUBLAS failure 3: CUBLAS_STATUS_ALLOC_FAILED ; GPU=0 ; hostname=DESKTOP-VRVRQRL ; file=D:\a_work\1\s\onnxruntime\core\providers\cuda\cuda_execution_provider.cc ; line=168 ; expr=cublasCreate(&cublas_handle_);

Please take look and guide me how i can solve this issue.

[chnisar515]

i forget to tell that i selected setting in (Select post-processing) to (GFPGAN) and after that when i click to start i got that error which i mentioned my above post and now when i select (None) in setting (Select post-processing) so now after that when i click on start button i did not get any error but my output video file didn't reflect any change then i noticed when i draw an image to Source Files its successfully uploaded but on the above Box (Input Faces) there's no face to show even that box.

[awpeoith1254]

When you are in the face swap tab and you load your target file is it showing in the preview tab? If not, wait for it to load before loading your input file. Once you have your input file loaded, be sure to click face swap --if it is a video, you can find a place in the scrubber for where you are expecting to see the face swap occur and see if it is happening. Please note that you might need to refresh the preview on a video as the scrubber is sometimes not at the same frame. Depending on which face swapping type you are using (e.g., first face, all faces, target face) you might need to adjust other settings -- in particular for target face the higher the threshold the less well it has to match the target image to be swapped.

Also for the test run I would just set the max memory to use to 0 in your settings and set post processing to none. Also is you have 'force cpu for face analyzer' then you are not using CUDA or your graphics card. Not sure if that is your goal, if it is change the provider in that section of the settings to CPU vs. cuda.

[lysxelapsed]

Your system isn't suited for these kinds of tasks. Roop-unleashed is RAM- and VRAM-hungry. Using GPEN as enhancer, my system currently allocates (as in: reserves for) over 14GB RAM and over 8GB VRAM (GPU memory) for the swapping process. You have 16G RAM and 2GB VRAM (940M), I doubt it will work even without an enhancer.
The error message refers to roop-unleashed not being able to allocate enough VRAM. First of all it's using your Intel HD Graphics 5500 (GPU 0), you would have to set the environment variable CUDA_VISIBLE_DEVICES="1" in python first. The result would be the same though, since you only have 2GB VRAM.
I'd recommend 32GB RAM and a graphics card with 6GB VRAM at the very least. 10+ GB VRAM if you want to use an enhancer.