Deepkseek Janus Pro: Optimize Image/Video Generation

While working with Text-to-Image or Text-to-Video workflows, you already know the struggle of getting accurate and high-quality results is not always easy. Now, this can be achieved by  DeepSeek’s Janus Pro Model (an upgraded version of Janus) is here to change the game.

They took things up a notch by integrating DeepSeek-LLM (1.5B and 7B) as the base language model, with a sequence length of up to 4096 tokens. Well, it simply means the model understands prompts way better and produces much more refined results.

Source-Deepseek Janus Pro Research Paper

DeepSeek’s research shows that Janus Pro 7B is not just another model, it actually outperforms models like Stable Diffusion 3 Medium, Janus (previous version) and DALLE 3. People can access their research paper for further in-depth understanding. 

The model usage is registered under DeepSeek License that comes under MIT license. Lets, see how we can enhance the quality of our Image/Video generation in ComfyUI.

Installing process

1. First, install ComfyUI if you are new to it. Old user need to update ComfyUI from the Manager by selecting “Update ComfyUI“.

2. There are two methods to install and setup Janus Pro.

Automatic Install(for normal user):

From the manager, select Custom Node manager and serach “Janus Pro” (author-CY-CHENYUE) from the download list and click “Install“.

Manual Install(for professionals):

Move inside your “ComfyUI/custom_nodes” folder and use the following command to install.

git clone https://github.com/CY-CHENYUE/ComfyUI-Janus-Pro

Install the requried dependecies. 

cd ComfyUI-Janus-Pro

……python_embededpython.exe -m pip install -r requirements.txt

3. Currently, 1 billion and 7 billion model variants are available. Download all these files shown below (illustrated image) and create the folder structure as follows:

(a)Download Janus-Pro-1B  and create this “ComfyUI/models/Janus-Pro/Janus-Pro-1B/” folder structure and save it inside Janus-Pro-1B folder.

(b)Now, download Janus-Pro-7B   and create this “ComfyUI/models/Janus-Pro/Janus-Pro-7B/” folder structure and save all files inside Janus-Pro-7B folder.

4. Restart ComfyUI to take effect.

Workflow

1. After installing, the respective workflow (ComfyUI Janus-Pro-workflow.png) can be found inside your “ComfyUI/custom_nodes/ComfyUI-Janus-Pro/workflow” folder.

2. Just drag and drop into ComfyUI. You can modify the provided workflows for your basic Flux, SDXL, StableDiffusion 3.5 text to image and for even HunyuanVideo, CogVideoX, Mochi and LTXVideo workflows.