Hugging Face's logo

AIDC-AI
/
Ovis-Image-7B

Text-to-Image
Diffusers
Safetensors
English
Chinese
OvisImagePipeline
image generation
Model card Files
xet
Community
3
Flourish commited on
Commit
4948ef8
·
verified ·
1 Parent(s): 2922f72

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +224 -3
README.md CHANGED
@@ -1,3 +1,224 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - en
5
+ - zh
6
+ base_model:
7
+ - AIDC-AI/Ovis2.5-2B
8
+ pipeline_tag: text-to-image
9
+ library_name: diffusers
10
+ tags:
11
+ - image generation
12
+ ---
13
+
14
+ # Ovis-Image
15
+
16
+ <div align="center">
17
+ <img src=https://cdn-uploads.huggingface.co/production/uploads/637aebed7ce76c3b834cea37/3IK823BZ8w-mz_QfeYkDn.png width="30%"/>
18
+ </div>
19
+
20
+ <p align="center">
21
+ <a href="https://arxiv.org/abs/2506.23044"><img src="https://img.shields.io/badge/arXiv_paper-2506.23044-b31b1b.svg" alt="arxiv"></a>
22
+ <!-- <a href="https://github.com/AIDC-AI/Ovis-U1/blob/main/docs/Ovis_U1_Report.pdf"><img src="https://img.shields.io/badge/Paper-Tech_Report-b31b1b" alt="paper"></a> -->
23
+ <a href="https://github.com/AIDC-AI/Ovis-Image"><img src="https://img.shields.io/badge/GitHub-AIDC--AI/Ovis--Image-blue?style=flat&logo=github" alt="code"></a>
24
+ <a href="https://huggingface.co/spaces/AIDC-AI/Ovis-Image-7B"><img src="https://img.shields.io/badge/🎨_HF_Spaces-AIDC--AI/Ovis--Image--7B-lightblack" alt="demo"></a>
25
+ <a href="https://huggingface.co/AIDC-AI/Ovis-Image-7B"><img src="https://img.shields.io/badge/🤗_Model-AIDC--AI/Ovis-Image-7B-yellow" alt="model"></a>
26
+ </p>
27
+
28
+ <p align="left">
29
+ Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering, designed to operate efficiently under stringent computational
30
+ constraints.
31
+ </p>
32
+
33
+
34
+ <p align="center">
35
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/636f4c6b5d2050767e4a1491/10uxbIz-NJf1d716eFZ6O.png" width="95%">
36
+ <br>
37
+ <em>The overall architecture of Ovis-Image (cf. Fig.2 in our report).</em>
38
+ </p>
39
+
40
+ ## 🏆 Highlights
41
+
42
+ * **Strong text rendering at a compact 7B scale**: Ovis-Image is a 7B text-to-image model that delivers text rendering quality comparable to much larger 20B-class systems such as Qwen-Image and competitive with leading closed-source models like GPT4o in text-centric scenarios, while remaining small enough to run on widely accessible hardware.
43
+ * **High fidelity on text-heavy, layout-sensitive prompts**: The model excels on prompts that demand tight alignment between linguistic content and rendered typography (e.g., posters, banners, logos, UI mockups, infographics), producing legible, correctly spelled, and semantically consistent text across diverse fonts, sizes, and aspect ratios without compromising overall visual quality.
44
+ * **Efficiency and deployability**: With its 7B parameter budget and streamlined architecture, Ovis-Image fits on a single high-end GPU with moderate memory, supports low-latency interactive use, and scales to batch production serving, bringing near–frontier text rendering to applications where tens-of-billions–parameter models are impractical.
45
+
46
+
47
+ ## ✨ Showcase
48
+
49
+ Here are some examples demonstrating the capabilities of Ovis-Image.
50
+
51
+ <figure>
52
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/636f4c6b5d2050767e4a1491/UFmGxXISVOtOISoDON5wj.png" alt="Ovis-Image examples">
53
+ <figcaption style="text-align: center;"></figcaption>
54
+ </figure>
55
+
56
+
57
+
58
+ ## 📦 Installation
59
+
60
+ Ovis-Image has been tested with Python 3.10, Torch 2.6.0, and Transformers 4.57.1. For a full list of package dependencies, please see `requirements.txt`.
61
+
62
+ ```bash
63
+ git clone [email protected]:AIDC-AI/Ovis-Image.git
64
+ conda create -n ovis-image python=3.10 -y
65
+ conda activate ovis-image
66
+ cd Ovis-Image
67
+ pip install -r requirements.txt
68
+ pip install -e .
69
+ ```
70
+
71
+ ## 🛠️ Inference
72
+
73
+ For text-to-image, please run
74
+ ```bash
75
+ python ovis_image/test.py \
76
+ --model_path AIDC-AI/Ovis-Image-7B/ovis_image.safetensors \
77
+ --vae_path AIDC-AI/Ovis-Image-7B/ae.safetensors \
78
+ --ovis_path AIDC-AI/Ovis-Image-7B/Ovis2.5-2B \
79
+ --image_size 1024 \
80
+ --denoising_steps 50 \
81
+ --cfg_scale 5.0 \
82
+ --prompt "A creative 3D artistic render where the text \"OVIS-IMAGE\" is written in a bold, expressive handwritten brush style using thick, wet oil paint. The paint is a mix of vibrant rainbow colors (red, blue, yellow) swirling together like toothpaste or impasto art. You can see the ridges of the brush bristles and the glossy, wet texture of the paint. The background is a clean artist's canvas. Dynamic lighting creates soft shadows behind the floating paint strokes. Colorful, expressive, tactile texture, 4k detail." \
83
+
84
+ ```
85
+
86
+
87
+ Alternatively, you can try Ovis-Image directly in your browser on [![Hugging Face Space](https://img.shields.io/badge/🎨_HF_Spaces-AIDC--AI/Ovis--Image--7B-lightblack)](https://huggingface.co/spaces/AIDC-AI/Ovis-Image-7B)
88
+
89
+
90
+ ## 📊 Performance
91
+
92
+
93
+ **Evaluation of text rendering ability on CVTG-2K.**
94
+
95
+ | Model | #Params. | WA (2 regions) | WA (3 regions) | WA (4 regions) | WA (5 regions) | WA (average) | NED↑ | CLIPScore↑ |
96
+ | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
97
+ | Seedream 3.0 | - | 0.6282 | 0.5962 | 0.6043 | 0.5610 | 0.5924 | 0.8537 | 0.7821 |
98
+ | GPT4o | - | 0.8779 | 0.8659 | 0.8731 | 0.8218 | 0.8569 | 0.9478 | 0.7982 |
99
+ | SD3.5 Large | 11B+8B | 0.7293 | 0.6825 | 0.6574 | 0.5940 | 0.6548 | 0.8470 | 0.7797 |
100
+ | RAG-Diffusion | 11B+12B | 0.4388 | 0.3316 | 0.2116 | 0.1910 | 0.2648 | 0.4498 | 0.7797 |
101
+ | FLUX.1-dev | 11B+12B | 0.6089 | 0.5531 | 0.4661 | 0.4316 | 0.4965 | 0.6879 | 0.7401 |
102
+ | TextCrafter | 11B+12B | 0.7628 | 0.7628 | 0.7406 | 0.6977 | 0.7370 | 0.8679 | 0.7868 |
103
+ | Qwen-Image | 7B+20B | 0.8370 | 0.8364 | 0.8313 | 0.8158 | 0.8288 | 0.9116 | 0.8017 |
104
+ | Ovis-Image | 2B+7B | **0.9248** | **0.9239** | **0.9180** | **0.9166** | **0.9200** | **0.9695** | **0.8368** |
105
+
106
+
107
+ **Evaluation of text rendering ability on LongText-Bench.**
108
+
109
+ | Model | #Params. | LongText-Bench-EN | LongText-Bench-ZN |
110
+ | :--- | :---: | :---: | :---: |
111
+ | Kolors 2.0 | - | 0.258 | 0.329 |
112
+ | GPT4o | - | **0.956** | 0.619 |
113
+ | Seedream 3.0 | - | 0.896 | 0.878 |
114
+ | OmniGen2 | 3B+4B | 0.561 | 0.059 |
115
+ | Janus-Pro | 7B | 0.019 | 0.006 |
116
+ | BLIP3-o | 7B+1B | 0.021 | 0.018 |
117
+ | FLUX.1-dev | 11B+12B | 0.607 | 0.005 |
118
+ | BAGEL | 7B+7B | 0.373 | 0.310 |
119
+ | HiDream-I1-Full | 11B+17B | 0.543 | 0.024 |
120
+ | Qwen-Image | 7B+20B | 0.943 | 0.946 |
121
+ | Ovis-Image | 2B+7B | 0.922 | **0.964** |
122
+
123
+
124
+ **Evaluation of text-to-image generation ability on DPG-Bench.**
125
+
126
+ | Model | #Params. | Global | Entity | Attribute | Relation | Other | Overall |
127
+ | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
128
+ | Seedream 3.0 | - | **94.31** | **92.65** | 91.36 | 92.78 | 88.24 | 88.27 |
129
+ | GPT4o | - | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | 85.15 |
130
+ | Ovis-U1 | 2B+1B | 82.37 | 90.08 | 88.68 | 93.35 | 85.20 | 83.72 |
131
+ | OmniGen2 | 3B+4B | 88.81 | 88.83 | 90.18 | 89.37 | 90.27 | 83.57 |
132
+ | Janus-Pro | 7B | 86.90 | 88.90 | 89.40 | 89.32 | 89.48 | 84.19 |
133
+ | BAGEL | 7B+7B | 88.94 | 90.37 | 91.29 | 90.82 | 88.67 | 85.07 |
134
+ | HiDream-I1-Full | 11B+17B | 76.44 | 90.22 | 89.48 | 93.74 | 91.83 | 85.89 |
135
+ | UniWorld-V1 | 7B+12B | 83.64 | 88.39 | 88.44 | 89.27 | 87.22 | 81.38 |
136
+ | Qwen-Image | 7B+20B | 91.32 | 91.56 | **92.02** | **94.31** | **92.73** | **88.32** |
137
+ | Ovis-Image | 2B+7B | 82.37 | 92.38 | 90.42 | 93.98 | 91.20 | 86.59 |
138
+
139
+
140
+ **Evaluation of text-to-image generation ability on GenEval.**
141
+
142
+ | Model | #Params. | Single object | Two object | Counting | Colors | Position | Attribute binding | Overall |
143
+ | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
144
+ | Seedream 3.0 | - | 0.99 | 0.96 | **0.91** | **0.93** | 0.47 | **0.80** | 0.84 |
145
+ | GPT4o | - | 0.99 | 0.92 | 0.85 | 0.92 | 0.75 | 0.61 | 0.84 |
146
+ | Ovis-U1 | 2B+1B | 0.98 | **0.98** | 0.90 | 0.92 | **0.79** | 0.75 | **0.89** |
147
+ | OmniGen2 | 3B+4B | **1.00** | 0.95 | 0.64 | 0.88 | 0.55 | 0.76 | 0.80 |
148
+ | Janus-Pro | 7B | 0.99 | 0.89 | 0.59 | 0.90 | **0.79** | 0.66 | 0.80 |
149
+ | BAGEL | 7B+7B | 0.99 | 0.94 | 0.81 | 0.88 | 0.64 | 0.63 | 0.82 |
150
+ | HiDream-I1-Full | 11B+17B | 1.00 | **0.98** | 0.79 | 0.91 | 0.60 | 0.72 | 0.83 |
151
+ | UniWorld-V1 | 7B+12B | 0.99 | 0.93 | 0.79 | 0.89 | 0.49 | 0.70 | 0.80 |
152
+ | Qwen-Image | 7B+20B | 0.99 | 0.92 | 0.89 | 0.88 | 0.76 | 0.77 | 0.87 |
153
+ | Ovis-Image | 2B+7B | **1.00** | 0.97 | 0.76 | 0.86 | 0.67 | **0.80** | 0.84 |
154
+
155
+
156
+
157
+ **Evaluation of text-to-image generation ability on OneIG-EN.**
158
+
159
+ | Model | #Params. | Alignment | Text | Reasoning | Style | Diversity | Overall |
160
+ | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
161
+ | Kolors 2.0 | - | 0.820 | 0.427 | 0.262 | 0.360 | 0.300 | 0.434 |
162
+ | Imagen4 | - | 0.857 | 0.805 | 0.338 | 0.377 | 0.199 | 0.515 |
163
+ | Seedream 3.0 | - | 0.818 | 0.865 | 0.275 | 0.413 | 0.277 | 0.530 |
164
+ | GPT4o | - | 0.851 | 0.857 | **0.345** | **0.462** | 0.151 | 0.533 |
165
+ | Ovis-U1 | 2B+1B | 0.816 | 0.034 | 0.226 | 0.443 | 0.191 | 0.342 |
166
+ | CogView4 | 6B | 0.786 | 0.641 | 0.246 | 0.353 | 0.205 | 0.446 |
167
+ | Janus-Pro | 7B | 0.553 | 0.001 | 0.139 | 0.276 | **0.365** | 0.267 |
168
+ | OmniGen2 | 3B+4B | 0.804 | 0.680 | 0.271 | 0.377 | 0.242 | 0.475 |
169
+ | BLIP3-o | 7B+1B | 0.711 | 0.013 | 0.223 | 0.361 | 0.229 | 0.307 |
170
+ | FLUX.1-dev | 11B+12B | 0.786 | 0.523 | 0.253 | 0.368 | 0.238 | 0.434 |
171
+ | BAGEL | 7B+7B | 0.769 | 0.244 | 0.173 | 0.367 | 0.251 | 0.361 |
172
+ | BAGEL+CoT | 7B+7B | 0.793 | 0.020 | 0.206 | 0.390 | 0.209 | 0.324 |
173
+ | HiDream-I1-Full | 11B+17B | 0.829 | 0.707 | 0.317 | 0.347 | 0.186 | 0.477 |
174
+ | HunyuanImage-2.1 | 7B+17B | 0.835 | 0.816 | 0.299 | 0.355 | 0.127 | 0.486 |
175
+ | Qwen-Image | 7B+20B | **0.882** | 0.891 | 0.306 | 0.418 | 0.197 | **0.539** |
176
+ | Ovis-Image | 2B+7B | 0.858 | **0.914** | 0.308 | 0.386 | 0.186 | 0.530 |
177
+
178
+
179
+
180
+ **Evaluation of text-to-image generation ability on OneIG-ZN.**
181
+
182
+ | Model | #Params. | Alignment | Text | Reasoning | Style | Diversity | Overall |
183
+ | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
184
+ | Kolors 2.0 | - | 0.738 | 0.502 | 0.226 | 0.331 | 0.333 | 0.426 |
185
+ | Seedream 3.0 | - | 0.793 | 0.928 | 0.281 | 0.397 | 0.243 | 0.528 |
186
+ | GPT4o | - | 0.812 | 0.650 | **0.300** | **0.449** | 0.159 | 0.474 |
187
+ | CogView4 | 6B | 0.700 | 0.193 | 0.236 | 0.348 | 0.214 | 0.338 |
188
+ | Janus-Pro | 7B | 0.324 | 0.148 | 0.104 | 0.264 | **0.358** | 0.240 |
189
+ | BLIP3-o | 7B+1B | 0.608 | 0.092 | 0.213 | 0.369 | 0.233 | 0.303 |
190
+ | BAGEL | 7B+7B | 0.672 | 0.365 | 0.186 | 0.357 | 0.268 | 0.370 |
191
+ | BAGEL+CoT | 7B+7B | 0.719 | 0.127 | 0.219 | 0.385 | 0.197 | 0.329 |
192
+ | HiDream-I1-Full | 11B+17B | 0.620 | 0.205 | 0.256 | 0.304 | 0.300 | 0.337 |
193
+ | HunyuanImage-2.1 | 7B+17B | 0.775 | 0.896 | 0.271 | 0.348 | 0.114 | 0.481 |
194
+ | Qwen-Image | 7B+20B | **0.825** | **0.963** | 0.267 | 0.405 | 0.279 | **0.548** |
195
+ | Ovis-Image | 2B+7B | 0.805 | 0.961 | 0.273 | 0.368 | 0.198 | 0.521 |
196
+
197
+ ## 📚 Citation
198
+
199
+ If you find Ovis-Image useful for your research or applications, please cite our technical report:
200
+
201
+ ```bibtex
202
+ @article{wang2025ovis_image,
203
+ title={Ovis-Image Technical Report},
204
+ author={Wang, Guo-Hua and Zhao, Shanshan and Zhang, Xinjie and Cao, Liangfu and Zhan, Pengxin and Duan, Lunhao and Lu, Shiyin and Fu, Minghao and Zhao, Jianshan and Li, Yang and Chen, Qing-Guo},
205
+ journal={arXiv preprint arXiv:2506.23044},
206
+ year={2025}
207
+ }
208
+ ```
209
+
210
+ ## 🙏 Acknowledgments
211
+
212
+ The code is built upon [Ovis](https://github.com/AIDC-AI/Ovis) and [FLUX](https://github.com/black-forest-labs/flux). We thank their authors for open-sourcing their great work.
213
+
214
+ ## 📄 License
215
+
216
+ This project is licensed under the Apache License, Version 2.0 (SPDX-License-Identifier: Apache-2.0).
217
+
218
+ ## 🚨 Disclaimer
219
+
220
+ We used compliance checking algorithms during the training process, to ensure the compliance of the trained model(s) to the best of our ability. Due to complex data and the diversity of language model usage scenarios, we cannot guarantee that the model is completely free of copyright issues or improper content. If you believe anything infringes on your rights or generates improper content, please contact us, and we will promptly address the matter.
221
+
222
+ ## 🔥 We are hiring!
223
+
224
+ We are looking for both interns and full-time researchers to join our team, focusing on multimodal understanding, generation, reasoning, AI agents, and unified multimodal models. If you are interested in exploring these exciting areas, please reach out to us at [email protected].