2026-01-05: V2.0 adds a workflow for using ZIT for generation (Lora links included in ZIT workflow)
I have had a couple of asks for my workflows/prompts - so I figured I would share them here. My videos are all I2V of some kind (InfiniteTalk / Wan2.2 I2v / Wan 2.2 FLF) - these are the workflows that generate the first image.
They aren't particularly neat, definitely not optimised, but might be helpful for getting started and/or inspiration.
Happy to hear comments and criticisms or answer any questions
Description
FAQ
Comments (10)
Hi! I really enjoy your work, and thank you for engaging with your fans. I have a question I think a lot of newbies ask: what do I need to do to create videos like this from scratch, given that my only experience is with images in Forge?
Thanks - and always happy to help! I haven't any experience with Forge, although did use Automatic1111 initially. Biggest piece of advice is to get comfortable with ComfyUI - its much more flexible - all my videos here are generated in it (In the talking videos I am using the standard InfiniteTalk workflow), and there is often support for low vram.
@bebobular Thanks!
Thanks for this! I love how you're using InfiniteTalk, I've been having trouble wrapping my head around anything sound-related...would you be able to share a workflow for that or where to start? Apologies if that is included here and I missed it.
Thanks - and thanks for the compliment! I'm not doing any thing profound with InfiniteTalk - I am pretty much using this workflow as it stands: https://github.com/kijai/ComfyUI-WanVideoWrapper/blob/main/example_workflows/wanvideo_I2V_InfiniteTalk_example_03.json
The only addition is the Bouncing Boobs Lora: https://civitai.com/models/1343431?modelVersionId=2191270
You are doing great work with the BE Lora! If you had to improve that Lora - what would you want it to do different?
@bebobular That example workflow helps, thank you!
Thanks, that lora and the M4CROM4STI4 one do a lot of work. I would love a version of the BE lora that was trained to stretch clothes and squish the breasts instead of revealing them...similar to this bursting t2v lora you're using, but for i2v BE stuff.
Messy or not. Appreciate you sharing. I like looking at people's workflows. I struggle with the speed of Wan's image process, but the quality is nice when well configured.
Thanks - and I agree with you wrt Wan, it isn't fast, but I find its quality is very good (in limited experiments, I find it better than ZIT, albeit much slower). But configuration of Wan is finicky, and I find I have to, as a final step, refine using just the low noise and Instareal Lora to get a good quality final result.
Thanks for sharing your workflow. If possible can you please share the lora you used
Thanks - just updated the workflow to include a workflow for ZIT, it has links to the lora's in it. But let me know if there is any specific one you are looking that isn't included there.