~~~~ Evil model ~ Evil generation ~~~~
💦 Generation Settings:
Sampler: Euler a + Auto or DPM++ SDE Karras
Steps: 30-40
CFG Scale: 3-7
Resolution: 768x1248 or 1248x768 (cinematic quality)
Resolution: 768x1024 or 1024x768 (usually quality)
🍿 Prompting Tips:
ultra detailed image, ultra detailed background
cinematic, masterpiece, 4k,
The model is made up of my favorite models that I use most often.
❤️❤️❤️ Enjoy ❤️❤️❤️
Description
FAQ
Comments (7)
Not to be a nag here, but I notice that about half of the showcase images use Img2Video; this would be great if you had posted an img2vid model. However, img2vid doesn't properly demonstrate what YOUR model can do, it demonstrates what the video generator can do. This has become an increasingly prevalent problem on the platform-- this is the first of many models I intend to address this with.
Yes, that's neat, your pictures move. It'd be neater if they looked more like what your model produced in txt2img and less like an img2video edit.
In fact, it's just a beautiful picture animation. I always specify the picture generation text in the prompt, not the video. So the result itself can be repeated
@SupremeEvil Hmmm upon further inspection and with your comment in mind, I can't help but agree. Apologies for the assumption; it is clear that this is more of an animateDiff sort of motion, not a WAN or Qwen video as was previously assumed. Again, apologies for being presumptuous. I'm clearly off to a bad start in my journey here, but will strive to be more diligent in the future.
@Chronos_Shinomori No. You're right. My animations were created using Wan2.2, BUT! The original images for the animations were created using the exact model shown in the image
@SupremeEvil will it be ok to share the workflow for video generation you use
@jackaline There's really no secret to it. I'm using the standard ComfyUi template for Wan2.2, with a slight modification on Lora. That's it. There's no point in uploading a template that anyone can access.
clean 👍🏽