text and input image without video

[drmbt]

from the original repo it looks like there is a condition for running inference without a vkps_path, is there a way to do this without your comfy implementation? What would be the best approach to do supply audio and an image but without any kind of input video?

[Mikerhinos]

That's the use case I also have : just use a generated portrait and generated TTS audio as inputs, no video, like with Sad Talker. V-Express seems to allow it in Scenario 2 example (the one with Taylor Swift result).