Open KintCark opened 3 months ago
Maybe I will consider making a video later when I will have time for it.
For now I can advice some key points regarding the topic:
Try to find and use fp16 quantized models in ORT, like here: https://huggingface.co/Androidonnxfork/test/tree/main/fp16fullonnxsdquantized_in_ort (Note : most of them are available for download from the app directly)
Then place the necessary files to your device internal storage at Download/SDAI/model
directory (like app shows you, see the screenshot below). In app settings select the custom model and save the settings (press the "Setup" button).
You can also try to convert some model you like to ONNX format like in this instruction: https://onnxruntime.ai/docs/tutorials/tf-get-started.html (Note : never tried it myself as I don't have access to powerful PC temporarily that is able to process AI models)
Oh me eithermy pc needs an rtx 3050 12gbvram to run stable diffusion so I just use my phone but is it possible to get text to video local or stable video diffusion one day maybe add an hugging face txt 2 vd model ?
Can you please make a yt video on how to get ort models and install them on Android sdai app I'm looking everywhere for custom ort models but can't find any.