Introduction

This workflow built on ComfyUI is powerful and innovative. It brings new possibilities to image creation with the help of Tencent's open source InstantCharacter project. Users only need to upload a picture containing a person or object and enter the corresponding description prompt word, and the workflow can generate a new picture that is consistent with the people and objects in the uploaded picture, but meets the description of the prompt word. This means that whether it is converting the style of the picture, changing the scene, or adding specific elements, it can be easily achieved, which greatly improves the efficiency and creative space of image creation, allowing users to quickly obtain high-quality pictures that meet their needs.

The paper qualitatively compares InstantCharacter against other personalization methods also built upon the FLUX foundation, highlighting its advantages:

• Vs. OminiControl and EasyControl: These methods struggle significantly with maintaining character identity consistency. The generated characters often show noticeable differences from the reference image, failing to effectively preserve key identity features.
• Vs. ACE++: While slightly better at identity preservation, ACE++ only maintains partial features, particularly in simple scenarios. Its primary weakness lies in text controllability, especially when dealing with complex, action-oriented prompts, often resulting in inaccurate or unnatural outputs.
• Vs. UNO: UNO excels at identity consistency, but often over-preserves it. This comes at the cost of reduced editability. It finds it difficult to flexibly alter the character's pose or actions according to the text prompt or to integrate the character naturally into new backgrounds, sometimes resembling a simple "copy-paste" effect.
• InstantCharacter's Advantages: In contrast, InstantCharacter achieves an excellent balance between identity consistency, text controllability (especially for complex actions and scene integration), and overall image fidelity. It reliably "remembers" the character's appearance while flexibly and accurately creating high-quality images based on text instructions, even for complex prompts like "a character riding a bicycle" or "playing the piano in a specific location." Experiments show its performance surpasses the aforementioned open-source methods and is comparable to the powerful, closed-source GPT4o. Furthermore, InstantCharacter demonstrates good compatibility with style LoRAs, enabling stylized character generation easily.

https://github.com/Tencent/InstantCharacter

https://github.com/jax-explorer/ComfyUI-InstantCharacter

NOTICE：Animal character are relatively unstable.

Recommended machine：Ultra Pro

Workflow Overview

Part 1 : Get hf_token

1.Click on the user's avatar.

2.Click the Access Tokens button.

3.Click the Create new token button.

4.Click the Read button.

5.Enter the name of the token in the input box.

6.Click the Create token button.

7.After clicking the Create token button, the hf_token will pop up. Copy and keep the token. The token will be used in the workflow.

Part 2 : InstantCharacter workflow parameters

1.Enter the hf_token created in Part 1.If you enable the load_8bit function, you can use less graphics card memory to generate videos, but it will increase the time it takes to generate images.

2.Upload image.

3..Input prompt word.

4.Modify the image resolution by adjusting the 'height' and 'width' parameters.