/tldream

A tiny little diffusion drawing app

Primary LanguagePythonApache License 2.0Apache-2.0

tldream

A tiny little diffusion drawing app

total download version Open in Colab

tldream_demo.mp4

Features

  • Support any Stable Diffusion 1.5 model on HuggingFace or local ckpt/safetensors
  • Support multi-language translation: zh, ja, ko, ru

Quick Start

# In order to use the GPU, install cuda version of pytorch first.
# pip install torch==1.13.1+cu117 --extra-index-url https://download.pytorch.org/whl/cu117
pip install tldream
tldream --model runwayml/stable-diffusion-v1-5 --device cuda

If you are not familiar with python/pip, you can try this installer: Windows 1-click Installer

Command line arguments

  • --listen / --no-listen: If true, start server at 0.0.0.0 [default: no-listen]
  • --port INTEGER: [default: 4242]
  • --device TEXT: Device to use (cuda, cpu or mps) [default: cuda]
  • --model TEXT: Any HuggingFace Stable Diffusion model id. Or local ckpt/safetensors path [default: runwayml/stable-diffusion-v1-5]
  • --lang TEXT: Translation language model. ['en', 'zh', 'ja', 'ko', 'ru'] [default: en]
  • --low-vram / --no-low-vram: Use low vram mode [default: no-low-vram]
  • --fp32 / --no-fp32: Use float32 mode (For NVIDIA 16xx GPU) [default: no-fp32]
  • --nsfw-filter / --no-nsfw-filter: Enable nsfw filter [default: nsfw-filter]
  • --cache-dir TEXT: Model cache directory, by default model downloaded to ~/.cache/huggingface/hub
  • --local-files-only / --no-local-files-only: Not connect to HuggingFace server, add this flag if model has been downloaded [default: no-local-files-only]