LiteVision-v1

LiteVision-v1 is a lightweight (only about 2GB), fast Stable Diffusion 1.5-based model optimized for low-step inference using LCM scheduler.

  • Built for low-end GPU.
  • Minimal steps.
  • Clean results.

Benchmark test on P100 GPU

image

πŸš€ Key Features

  • πŸ”Ή Based on Stable Diffusion 1.5 architecture
  • πŸ”Ή Optimized for LCM (Latent Consistency Models)
  • πŸ”Ή High-quality output in 4–8 inference steps
  • πŸ”Ή Works with standard StableDiffusionPipeline
  • πŸ”Ή No custom pipeline required
  • πŸ”Ή Fully compatible with Diffusers 0.36.0

⚑ Recommended Settings

For best results:

num_inference_steps = 6
guidance_scale = 1.5
scheduler = LCMScheduler

Lower guidance gives cleaner results.

Higher guidance may introduce instability due to LCM behavior.

🧠 Why LiteVision?

Traditional SD 1.5 models require 20–30 steps for good quality. LiteVision ONLY requires about 6 steps.

πŸ›  Usage

import torch
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(
    "HyHorX/LiteVision-v1",
    torch_dtype=torch.float16
).to("cuda")

image = pipe(
    "cinematic cyberpunk city, ultra detailed",
    num_inference_steps=6,
    guidance_scale=1.5,
    height=512,
    width=512
).images[0]

image.save("litevision_output.png")

⚠ Notes

Designed specifically for LCM scheduler.

Not tuned for traditional DDIM/PNDM high-step sampling.

Use FP16 for optimal performance.

πŸ‘€ Author

HyHorX

License

This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

  • You can't use the model to deliberately produce nor share illegal or harmful outputs or content

  • The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license

  • You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here

Downloads last month
57
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Space using HyHorX/LiteVision-v1 1