Kolors
Kolors is a bilingual text-to-image generation model developed by Kuaishou Technology, designed with native understanding of both Chinese and English languages for prompt-driven image creation.
Key Highlights
Strong Chinese Language Support
Offers deep understanding of Chinese language nuances, idiomatic expressions, and cultural concepts through ChatGLM-based text encoder.
LLM-Based Text Understanding
Uses a large language model-based text encoder instead of CLIP, interpreting complex prompts and abstract concepts with superior accuracy.
Apache 2.0 Open Source
Released under Apache 2.0, one of the most permissive open-source licenses, freely permitting all types of use including commercial.
Competitive Visual Quality
Offers strong photorealism and color accuracy that competes with SDXL and approaches FLUX.1 quality in some evaluations.
About
Kolors is a large-scale text-to-image model developed by the Kuaishou Technology team (known for the Kwai short video platform), released as open source in July 2024. Built on a latent diffusion architecture with approximately 2.6 billion parameters, the model stands out with its strong capabilities in Chinese text understanding and rendering. Drawing on Kuaishou's extensive visual data repository, Kolors has become one of the most notable Chinese-origin open-source image generation models in the global AI landscape.
In terms of technical architecture, Kolors adopts the U-Net-based latent diffusion approach. The model's most important technical feature is its use of the ChatGLM large language model as the text encoder — providing a major advantage particularly in understanding Chinese prompts. Thanks to ChatGLM's bilingual (Chinese-English) capabilities, the model demonstrates strong prompt adherence in both languages. The 2.6-billion parameter structure is comparable in scale to SDXL (3.5B). During training, Kuaishou's massive visual database was leveraged, and the model was optimized with bilingual text-image pairs. It operates at a native resolution of 1024x1024 pixels and supports multiple aspect ratios.
In terms of quality, Kolors delivers impressive results particularly when used with Chinese prompts. It surpasses most competitors in Chinese character rendering, art styles specific to Chinese culture, and accurate representation of Asian facial features. Photorealism and digital art quality are comparable to SDXL levels. Prompt adherence in complex compositions is strong, and color vibrancy is noteworthy. Human anatomy and detail accuracy are generally high. In benchmark tests, it achieves competitive scores among open-source models, particularly excelling in evaluations focused on Asian-market content.
Kolors is used by Chinese-speaking creative professionals, agencies producing content for the Chinese market, designers focusing on Asian aesthetics, game developers, and AI researchers. It is valuable in scenarios such as images containing Chinese text, Chinese culture-based illustrations, Asian character designs, e-commerce product visuals, and social media content targeting Chinese-speaking audiences. It is also used internally at Kuaishou for content generation for the company's video platform and related products.
Kolors is open-source under the Apache 2.0 license and downloadable from Hugging Face. It is compatible with the Diffusers library and can be run on ComfyUI. Running locally requires 8-12GB VRAM, making it accessible on consumer hardware. LoRA fine-tuning support is available, and various community-developed adapters can be used for style customization. Commercial use is permitted, and the license terms provide flexibility for developers.
In the competitive landscape, Kolors occupies a unique position with its strong competency in Chinese language support and Asian aesthetics. While Western-origin models like SDXL and FLUX.1 lead in general quality, Kolors has distinct advantages in Chinese prompt understanding and generating styles specific to Chinese culture. It also competes with other Chinese-origin models and Tencent's offerings in the growing Chinese AI image generation market. Its use of the ChatGLM text encoder presents an interesting technical approach to multilingual model development, influencing future research in multilingual image generation and demonstrating the importance of culturally-aware training data.
Use Cases
Chinese Content Creation
Creating marketing, e-commerce, and social media content for the Chinese market by generating high-quality images with Chinese prompts.
Cultural Content Generation
Cultural content production by creating visuals incorporating traditional Chinese culture, festivals, and art styles.
Bilingual Visual Projects
Supporting international projects by generating visuals with consistent quality in both Chinese and English prompts.
Open Source Research
Usage as a base model for researching the impact of LLM-based text encoders on image generation performance.
Pros & Cons
Pros
- Excellent bilingual support for both Chinese and English text understanding and generation
- Achieves highest MPS (Multi-dimensional Human Preference Score) and human satisfaction scores in evaluations
- Uses multimodal large language model for caption refinement, enabling fine-grained semantic understanding
- Strong visual appeal and photorealistic quality with superior text faithfulness in benchmarks
Cons
- Long text rendering is error-prone; accuracy drops significantly with longer text inputs
- Struggles with emotional subtlety and nuanced concepts like sarcasm or figurative language
- Diffusion-based randomness means outputs vary significantly between runs with limited user control
- Text rendering in English can be unreliable, sometimes producing Chinese characters despite English prompts
Technical Details
Parameters
8B
Architecture
Latent Diffusion with ChatGLM encoder
Training Data
proprietary (Kuaishou internal dataset)
License
Apache 2.0
Features
- ChatGLM Text Encoder
- Chinese-English Bilingual Support
- Apache 2.0 License
- 2.6B Parameters
- 1024x1024 Resolution
- Open Source Weights
Benchmark Results
| Metric | Value | Compared To | Source |
|---|---|---|---|
| Parametre Sayısı | 8B (ChatGLM3 text encoder) | SDXL: 6.6B | Kolors GitHub |
| FID Score (COCO-30K) | 9.85 | SDXL: 12.20 | Kolors Paper (arXiv) |
| Çince Prompt Desteği | Doğal Çince anlama | SDXL: Sadece İngilizce | Kolors GitHub |
| Maksimum Çözünürlük | 1024x1024 | — | Kolors GitHub |
Available Platforms
Frequently Asked Questions
Related Models
Adobe Firefly
Adobe Firefly is a commercially safe AI image generation model developed by Adobe, distinguished by being trained exclusively on licensed Adobe Stock content, openly licensed material, and public domain works. This training approach directly addresses the copyright concerns that surround most AI image generators, making Firefly uniquely suited for commercial and enterprise use where legal compliance is essential. Integrated natively into Adobe's Creative Cloud applications including Photoshop, Illustrator, and Adobe Express, Firefly powers features like Generative Fill, Generative Expand, and Text Effects, enabling seamless AI-assisted workflows within tools that millions of creative professionals already use daily. The model generates high-quality images across diverse styles with strong prompt adherence and particularly excels at producing content that feels commercially polished and brand-appropriate. Adobe provides an IP indemnification program for enterprise customers, offering legal protection against copyright claims related to Firefly-generated content. The model supports text-to-image generation, style transfer, text effects, and generative editing features. It is accessible through Adobe applications, the dedicated Firefly web interface, and an API for developers. Content creators, marketing teams, advertising agencies, and enterprise design departments value Firefly for its legal safety, seamless integration with existing Adobe workflows, and consistent professional output quality. While it may not achieve the artistic flexibility or raw creative potential of models like Midjourney, its commercial safety and professional tool integration make it indispensable for businesses requiring legally defensible AI-generated content.
Adobe Firefly 3
Adobe Firefly 3 is the third generation of Adobe's commercially safe generative AI model family, released in April 2024 as the backbone of AI features across Adobe Creative Cloud applications including Photoshop, Illustrator, and Adobe Express. The model delivers significant improvements over Firefly 2 in photorealistic quality, prompt adherence, and creative versatility. Adobe Firefly 3 was trained exclusively on licensed Adobe Stock content, openly licensed material, and public domain content, making it one of the few enterprise-grade AI image models that provides full intellectual property indemnification to commercial users. The model generates images with dramatically improved detail, more natural lighting and shadows, richer textures, and better human rendering compared to its predecessor. Firefly 3 powers features like Generative Fill and Generative Expand in Photoshop, Text to Image generation in Adobe Express, and vector generation capabilities in Illustrator. The model supports Structure Reference and Style Reference controls that allow users to maintain consistency across multiple generations. Available through Adobe's applications, the Firefly web interface, and the Firefly API for enterprise integration, the model serves creative professionals, marketing teams, and enterprise content producers. Firefly 3 supports various aspect ratios and outputs at resolutions suitable for both digital and print workflows. Adobe's commitment to Content Credentials ensures all Firefly-generated images carry metadata indicating AI origin, supporting content authenticity standards.
DALL-E 2
DALL-E 2 is OpenAI's second-generation image generation model that pioneered accessible AI image creation when it launched in 2022, introducing millions of users to the possibilities of text-to-image generation. Built on a diffusion model architecture with CLIP-based text understanding, DALL-E 2 generates images at 1024x1024 resolution from natural language descriptions. The model introduced several innovative capabilities that were groundbreaking at its release, including inpainting for editing specific regions of an image, outpainting for extending images beyond their original boundaries, and variations for creating alternative versions of existing images. DALL-E 2 demonstrated that AI could generate creative, coherent, and visually appealing images from simple text descriptions, sparking the entire consumer AI image generation revolution. While it has been superseded in quality by its successor DALL-E 3 and competitors like Midjourney v6 and FLUX.1, DALL-E 2 remains available through the OpenAI API at significantly reduced pricing, making it a cost-effective option for applications where maximum image quality is not the primary concern. The model offers reliable performance for basic image generation, simple editing tasks, and prototype creation. Developers building applications with high-volume image generation needs, educators creating visual materials, and hobbyists exploring AI art on a budget continue to use DALL-E 2. Its historical significance as one of the first widely accessible AI image generators that brought text-to-image technology to mainstream awareness cannot be overstated.
DALL-E 3
Historical model profile: DALL-E 3 was removed from the OpenAI API on May 12, 2026. The capabilities below describe this earlier model, not the current OpenAI image service. DALL-E 3 is OpenAI's earlier text-to-image generation model, deeply integrated with ChatGPT to provide an intuitive conversational interface for creating images. Unlike previous versions, DALL-E 3 natively understands context and nuance in text prompts, eliminating the need for complex prompt engineering. The model can generate highly detailed and accurate images from simple natural language descriptions, making AI image generation accessible to users without technical expertise. Its architecture builds upon diffusion model principles with proprietary enhancements that enable exceptional prompt fidelity, meaning images closely match what users describe. DALL-E 3 excels at rendering readable text within images, understanding spatial relationships, and following complex multi-part instructions. The model supports various artistic styles from photorealism to illustration, cartoon, and oil painting aesthetics. Safety features are built in at the model level, with content policy enforcement and metadata marking using C2PA provenance standards. DALL-E 3 is available through the ChatGPT Plus subscription and the OpenAI API, making it suitable for both casual users and developers building applications. Content creators, marketers, educators, and product designers use it extensively for social media graphics, presentation visuals, educational materials, and rapid concept exploration. As a closed-source proprietary model, it prioritizes safety, accessibility, and seamless user experience over customization flexibility.
Quick Info
Latent Diffusion with ChatGLM encoder