CivArchive
    Unbounded_LCM - Preview Version
    Preview 1
    Preview 2
    Preview 3
    Preview 4
    Preview 5
    Preview 6

    Description

    As a fast LCM model, it comes with the well-received SDXL Simulator , striving for a perfect balance between quality and efficiency.

    Being an LCM model, the plotting settings are very different from the previous SD1.5 model, so be sure to use the recommended parameters.

    Come to the AI mutual help club and make a friend, QQ group number:482570609

    →→→→→→Get more high-quality models←←←←←←


    Boundless_and_Without_Limits

    Selected from the "Shang Shu · Da Yu Mo", it means that no matter how far away it is, it can be reached, just like forging ahead in AI, even the stars and the sea will eventually be reached.  

    First of all, I don't know if you are like me, starting to feel indifferent to the same old internet celebrities, but with no choice, so I created one myself.

    Secondly, this is just the beginning, it is a preview version (earlier than a beta version), I will optimize the model based on your feedback, so please do not have overly high expectations for it, give it a try first, but your feedback will influence the direction of the model.


    Model Features:

    It can do what other realistic large models can do, but why choose it?  

    1. Brand new Asian faces, different from the mainstream internet celebrities, please let me know if there are any lookalikes (theoretically there shouldn't be).

    2. The model focuses on realism and delicacy, such as skin texture, grain details are clear and visible, with significant differences from beauty retouching models of internet celebrities (no good or bad, based on personal preference).

    3. Integrated with LCM_lora optimization to speed up, almost instant output at 512*768 resolution (tested on a 3060_12G graphics card).

    4. Good compatibility with various loras.


    Recommended Parameters:

    It is highly recommended that friends who are not familiar with LCM use the following parameter settings for output. 

    Sampling method: Euler a / DPM++SDE Karras

    Iteration steps: 6-9

    Hint word guidance coefficient (CFG Scale): 1.5

    For those familiar with the plugin, CFG scale fix can be turned on.

    Upscale model recommendation: 4x-UltraSharp


    Hope you have a fun time playing Peace

    FAQ

    Details

    Downloads
    695
    Platform
    ShakkerAI
    Platform Status
    Available
    Created
    1/1/2024
    Updated
    9/6/2024
    Deleted
    -

    Files

    Unbounded.safetensors

    Mirrors

    ShakkerAI (1 mirrors)