Open jiaxiangc opened 3 months ago
@fengyuzzz We have released text-to-image, lcm-lora, ip-adapter, controlnet workflow. Can you help us to make video demos?
Sure,I' m trying the latest workflow
@fengyuzzz We have released text-to-image, lcm-lora, ip-adapter, controlnet workflow. Can you help us to make video demos?
Love resadapter (excited for your A1111 extension). Wanted to mention it works well w/ TCD (Trajectory Consistency Distillation) also. I find that's a better alternative to LCM personally.
Repo if interested in testing: https://github.com/jabir-zheng/TCD
(Removed inaccurate workflow)
@CCpt5 Make sure the sampler_name is right. lcm
equal to tcd with eta=1.0
, which produce a blur result.
https://github.com/dfl/comfyui-tcd-scheduler Here is tcd scheduler.
is there any benefit to put resadapter model strength < 1.0? It seems some of the workflow videos have 0.6 or 0.7.
@dfl For personalized diffusion model, adjusting resadapter model strength can reduce the style conflict.
We will release more workflow, such as lcm-lora, controlnet, ip-adapter to teach users to use ComfyUI-ResAdapter.
@fengyuzzz If possible, you can put more demo video in this issue. I will make a gallery for these video.