nielsr HF Staff commited on
Commit
c872ce0
·
verified ·
1 Parent(s): 0fabe02

Add pipeline tag, license, and improve model card documentation

Browse files

Hi! I'm Niels, part of the community science team at Hugging Face. This PR improves the UniCharacter model card by:
- Adding the `any-to-any` pipeline tag to improve discoverability for multimodal models.
- Adding the `apache-2.0` license based on standard practices for this repository.
- Moving the ArXiv ID from the YAML metadata to the markdown section to comply with Hub metadata guidelines.
- Adding links to the paper, project page, and GitHub repository.
- Including a sample usage section with code snippets taken from the official GitHub README to help users get started quickly.

Files changed (1) hide show
  1. README.md +53 -30
README.md CHANGED
@@ -1,4 +1,6 @@
1
  ---
 
 
2
  tags:
3
  - UniCharacter
4
  - customized-multimodal-role-play
@@ -6,14 +8,15 @@ tags:
6
  - character-customization
7
  - text-to-image
8
  - image-generation
9
- - arxiv:2605.08129
10
  ---
11
 
12
  # UniCharacter
13
 
14
  UniCharacter is a collection of character-specific checkpoints for **Customized Multimodal Role-Play (CMRP)**, introduced in the paper [Towards Customized Multimodal Role-Play](https://arxiv.org/abs/2605.08129).
15
 
16
- The model is designed to customize a character's persona, dialogue style, and visual identity so that the character can respond consistently across text and image generation settings.
 
 
17
 
18
  ## Repository Contents
19
 
@@ -21,26 +24,52 @@ This repository contains separate checkpoint folders for multiple characters. Ea
21
 
22
  Available character folders include:
23
 
24
- - `Adrien_Brody`
25
- - `Bo`
26
- - `Butin`
27
- - `Chandler`
28
- - `Coco`
29
- - `Furina`
30
- - `Gao_Qiqiang`
31
- - `Hermione`
32
- - `Ichihime`
33
- - `Joey`
34
- - `Leonardo`
35
- - `Mam`
36
- - `Miki_Nikaidou`
37
- - `Mydieu`
38
- - `Pikachu`
39
- - `Rin_Tohsaka`
40
- - `Saber`
41
- - `Will_In_Vietnam`
42
- - `Wukong`
43
- - `YuiYagi`
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44
 
45
  ## Download
46
 
@@ -56,12 +85,6 @@ Download a single character checkpoint folder:
56
  huggingface-cli download Tangc03/UniCharacter --include "Hermione/*" --local-dir UniCharacter
57
  ```
58
 
59
- ## Paper
60
-
61
- This model is associated with the following paper:
62
-
63
- - [Towards Customized Multimodal Role-Play](https://arxiv.org/abs/2605.08129)
64
-
65
  ## Citation
66
 
67
  If you use UniCharacter, please cite:
@@ -69,8 +92,8 @@ If you use UniCharacter, please cite:
69
  ```bibtex
70
  @article{tang2026towards,
71
  title={Towards Customized Multimodal Role-Play},
72
- author={Tang, Chao and Wu, Jianzong and Shi, Qingyu and Tian, Ye and Zhang, Aixi and Jiang, Hao and Zhang, Jiangning and Tong, Yunhai},
73
  journal={arXiv preprint arXiv:2605.08129},
74
  year={2026}
75
  }
76
- ```
 
1
  ---
2
+ license: apache-2.0
3
+ pipeline_tag: any-to-any
4
  tags:
5
  - UniCharacter
6
  - customized-multimodal-role-play
 
8
  - character-customization
9
  - text-to-image
10
  - image-generation
 
11
  ---
12
 
13
  # UniCharacter
14
 
15
  UniCharacter is a collection of character-specific checkpoints for **Customized Multimodal Role-Play (CMRP)**, introduced in the paper [Towards Customized Multimodal Role-Play](https://arxiv.org/abs/2605.08129).
16
 
17
+ [**Project Page**](https://tangc03.github.io/UniCharacter.github.io/) | [**GitHub**](https://github.com/Tangc03/UniCharacter) | [**Paper**](https://arxiv.org/abs/2605.08129)
18
+
19
+ The model is designed to customize a character's persona, dialogue style, and visual identity so that the character can respond consistently across text and image generation settings. Using a unified multimodal model, UniCharacter employs a two-stage training framework containing Unified Supervised Finetuning (Unified-SFT) and character-specific group relative policy optimization (Character-GRPO).
20
 
21
  ## Repository Contents
22
 
 
24
 
25
  Available character folders include:
26
 
27
+ - `Adrien_Brody`, `Bo`, `Butin`, `Chandler`, `Coco`, `Furina`, `Gao_Qiqiang`, `Hermione`, `Ichihime`, `Joey`, `Leonardo`, `Mam`, `Miki_Nikaidou`, `Mydieu`, `Pikachu`, `Rin_Tohsaka`, `Saber`, `Will_In_Vietnam`, `Wukong`, `YuiYagi`
28
+
29
+ ## Quick Usage Example
30
+
31
+ To use these checkpoints, please follow the installation instructions in the [official repository](https://github.com/Tangc03/UniCharacter). Below is an example of the unified inference interface:
32
+
33
+ ```python
34
+ from inference import create_unicharacter_inference
35
+ from pathlib import Path
36
+
37
+ # Initialize the unified inference (modify paths according to your environment)
38
+ inference = create_unicharacter_inference(
39
+ model_path="models/BAGEL-7B-MoT",
40
+ checkpoint_path="<checkpoint_path>",
41
+ vit_checkpoint_path="<vit_checkpoint_path>",
42
+ max_mem_per_gpu="40GiB",
43
+ seed=42,
44
+ )
45
+
46
+ out_dir = Path("test_images/outputs")
47
+ out_dir.mkdir(parents=True, exist_ok=True)
48
+
49
+ # 1) Text-to-image generation (Role T2I)
50
+ res = inference.generate_image("Ichihime chasing a butterfly")
51
+ res["image"].save(out_dir / "t2i_ichihime.png")
52
+
53
+ # 2) Visual understanding / VQA
54
+ res = inference.visual_understanding(
55
+ "data/personalized_data/train/Mahjong Soul-Ichihime/1.png",
56
+ "What's the color of Ichihime's hair?",
57
+ )
58
+ print("VQA:", res["text"])
59
+
60
+ # 3) Knowledge QA
61
+ res = inference.knowledge_qa("When do you born?")
62
+ print("Knowledge QA:", res["text"])
63
+
64
+ # 4) Multimodal role-play
65
+ res = inference.role_play(
66
+ character_name="Ichihime",
67
+ description="",
68
+ opening="",
69
+ user_text="Hi, Ichihime. How are you?",
70
+ )
71
+ print("Role-play:", res["response"])
72
+ ```
73
 
74
  ## Download
75
 
 
85
  huggingface-cli download Tangc03/UniCharacter --include "Hermione/*" --local-dir UniCharacter
86
  ```
87
 
 
 
 
 
 
 
88
  ## Citation
89
 
90
  If you use UniCharacter, please cite:
 
92
  ```bibtex
93
  @article{tang2026towards,
94
  title={Towards Customized Multimodal Role-Play},
95
+ author={Tang, Chao and Wu, Jianzong Manager, Shi, Qingyu and Tian, Ye and Zhang, Aixi and Jiang, Hao and Zhang, Jiangning and Tong, Yunhai},
96
  journal={arXiv preprint arXiv:2605.08129},
97
  year={2026}
98
  }
99
+ ```