Image
Bagel
ByteDance's open-source unified multimodal model for understanding, generation and image editing.
★ 6.2k starsApache-2.0
View guide →DeepSeek's Janus series - unified multimodal models that both understand and generate images
Janus is an open-source image project by deepseek-ai (deepseek-ai/Janus on GitHub). It has 17,754 GitHub stars and 2,232 forks, is released under the MIT license, and was last updated on GitHub on 1 February 2025.
Pick a Janus variant and size (for example Janus-Pro-7B or the 1.3B models) that fits your GPU.
Get the code, which includes inference examples and demos for each model in the series.
Create a Python 3.8+ environment and install the package in editable mode.
Load the weights from Hugging Face and run multimodal understanding or text-to-image generation.
One model answers questions about images and generates images from text.
Separate visual pathways for understanding and generation avoid the conflict that limited earlier unified models.
An optimized training strategy, expanded data and larger scale improve instruction following and generation stability.
Integrates rectified flow generation within the LLM framework for efficient unified modeling.
Local Gradio apps (and hosted Hugging Face Spaces demos) for interactive testing.
A FastAPI demo exposes the same functions as an API server.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/deepseek-ai/Janus.git
cd JanusInstall Janus and its dependencies into your Python environment.
pip install -e .Launch the Janus-Pro Gradio demo, or the FastAPI demo for API access.
python demo/app_januspro.py
# API variant: python demo/fastapi_app.pyIt is maintained by DeepSeek AI as an open-source research release on GitHub.
The code is MIT licensed; the model weights are under DeepSeek's model agreement (LICENSE-MODEL in the repo), so read that before commercial use.
No API key or fee: models run locally on your GPU after a free Hugging Face download.
Compared with separate specialist models (a VLM plus an image generator), Janus unifies both in one model; compared with hosted APIs, it is fully self-hosted but needs your own GPU.
Clone the repo, run pip install -e ., download a checkpoint from Hugging Face, and start the Gradio demo.
ByteDance's open-source unified multimodal model for understanding, generation and image editing.
Real-ESRGAN — practical blind super-resolution for real-world photos and video: restores and upscales degraded images with models trained on pure synthetic data, BSD-3-Clause
The Swiss Army Knife of Offline AI. Chat, see, speak, and generate images on your phone or Mac — GGUF LLMs, vision, Whisper speech-to-text, Stable Diffusion,…
Enhance images in JavaScript — upscale, denoise, deblur and dehaze with AI
Simple Flask web app that upscales low-res images with AI
Browser-only AI interface for UI mockups, covers, logos and social images
The most-starred other Image projects in the LibreAI catalogue — the realistic shortlist if Janus is not the right fit.
Stable Diffusion web UI
The node-graph powerhouse for precise, professional AI image and video generation.
High performance self-hosted photo and video management solution
Focus on prompting and generating
🆙 Upscayl - #1 Free and Open Source AI Image Upscaler for Linux, MacOS and Windows
Free and open — grab the source, star the repo, and start creating.