说明

https://github.com/daocodedao/StreamDiffusionExt.git

源项目 https://github.com/cumulo-autumn/StreamDiffusion Jan 8, 2024 commit id: 8ff959a3ae6536f33d6f0fbf809e9b620e346978

安装

环境

python3.10 -m venv venv_steam_diff
source venv_steam_diff/bin/activate


# 查看 cuda 版本
cat /usr/local/cuda/version.json
ls -l /usr/local | grep cuda
/usr/local/cuda/bin/nvcc --version


# cuda11.8
pip3 install torch==2.1.0 torchvision==0.16.0 xformers --index-url https://download.pytorch.org/whl/cu118


pip install -r requirements.txt

python setup.py develop easy_install streamdiffusion[tensorrt]
python -m streamdiffusion.tools.install-tensorrt

安装 npm

sudo apt update
sudo apt install -y nodejs
sudo apt install -y npm

https://weekendprojects.dev/posts/fixed-npm-install-error-missing-required-argument-1/
curl https://raw.githubusercontent.com/creationix/nvm/master/install.sh | bash
source ~/.profile
nvm install node 

项目启动

文字图片

# 启动 txt2img 端口： 9090
cd /data/work/StreamDiffusion
./start-txt-img.sh

图片图片

# 启动 txt2img 端口： 9091
cd /data/work/StreamDiffusion
./start-img-img.sh

vid2vid

# 启动 vid2vid 端口： 9092
cd /data/work/StreamDiffusion
./start-vid-vid.sh  

example

vid2vid

cd /data/work/StreamDiffusion
source venv_steam_diff/bin/activate 
python examples/vid2vid/main.py --input resource/input/奢奢夫人.mp4 --output resource/output/outvid.mp4 --prompt "girls with red hair" --model_id ./models/Model/kohaku-v2.1.safetensors 

python examples/vid2vid/main.py --input resource/input/2088_1705490330.mp4 --output resource/output/2088_1705490330_out.mp4 --prompt "girls with red hair" --model_id ./models/Model/kohaku-v2.1.safetensors 

txt2img

cd /data/work/StreamDiffusion
source venv_steam_diff/bin/activate 
python examples/txt2img/single.py --output ./resource/output/outtxt.png --prompt "1girl with brown dog ears, thick frame glasses" --model_id_or_path ./models/Model/kohaku-v2.1.safetensors

# turbo
python examples/txt2img/single.py --output ./resource/output/outtxt.png --prompt "1girl with brown dog ears, thick frame glasses" --model_id_or_path ./models/sd_turbo/sd_turbo.safetensors

img2img

参考 https://github.com/radames/Real-Time-Latent-Consistency-Model/tree/main?tab=readme-ov-file https://huggingface.co/collections/latent-consistency/latent-consistency-model-demos-654e90c52adb0688a0acbe6f

cd /data/work/StreamDiffusion
source venv_steam_diff/bin/activate 
python examples/img2img/single.py --input ./resource/input/in.png --output ./resource/output/outimg.png --prompt "girls with red hair" --model_id_or_path ./models/Model/kohaku-v2.1.safetensors


./start-gradio-img-img.sh

常见问题

ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost.

export no_proxy="localhost, 127.0.0.1, ::1"