Sign In

Almost Anything v1.0

116
1.7k
55
Verified:
SafeTensor
Type
Checkpoint Trained
Stats
1,240
Reviews
Published
Apr 24, 2024
Base Model
SD 1.5
Training
Steps: 20
Epochs: 50
Hash
AutoV2
56A0A8D647
  • edit for V1*

    Here it is, finally out of beta and I am extremely pleased with the results. This updated model offers greater versatility while continuing to prioritize realism. While primarily designed for 512x512 images, it can also handle 1024x1024 images with some trial and error depending on the complexity of the prompt and settings.

    This model is compatible with LCM and can be paired with SD Turbo, though it may not be the best option for high quality images, it is great for rough drafts. Feel free to experiment with the settings and let me know your feedback!


  • I also included both fp16 and fp32 versions of the model.

  • For creating 1024x1024 images, I personally recommend using DPM++ 3M SDE GPU for the sampler and Karras for the scheduler, with 50 steps and HiRes fix enabled. If you are interested in my exact settings for this, feel free to ask!

  • Lastly, please share images of your creations - I would love to see them! ๐Ÿ˜€

    * For V.05 - Beta
    --------------
    I've been working on this model for a while and figured I'd release a Beta of it. It still has a long way to go but has potential. Please upload pics you made with it so I can see what needs to be improved. So far this model focuses on skin textures and lifelike realisum. It can do NSFW and yes... It can kinda do buttholes which I know most models struggle with hah! If anyone has any questions dont hesitate to reach out :)

I personally use:

DPM++ 3M SDE GPU for the sampler

And Karra's for the Scheduler.

Steps normally around 40 and CFG scale you can play with a lot on this model, sometimes 1.00 works sometimes 5 works better.

I normally have mine around 2.75-4.00

Again its a beta so let me know!