Kolors A large model of Wenshengtu that understands Chinese better

Model Introduction

Kolors is a large-scale text-to-image generation model based on latent diffusion developed by the Kuaishou Kolors team. After training on billions of text-image pairs, Kolors has shown significant advantages over open-source and closed-source models in terms of visual quality, complex semantic accuracy, and text rendering of Chinese and English characters. In addition, Kolors supports both Chinese and English input, and has shown strong performance in understanding and generating Chinese content. The generation effect is comparable to that of Midjourney-v6, and supports text input of up to 256 characters.

How to run

1. Clone and run the container

2. When the container is in the "Running" state, copy the API address and open it in the browser

3. After opening the link, you can see the following interface

4. Click below to upload a picture and enter the text prompt, click Generate Image The result is generated

You can also modify the relevant parameters as needed

Height: Modify the height of the generated image
Width: Modify the width of the generated image
Inference Steps: The number of denoising steps used when generating images. Usually, using the default number of inference steps (e.g. 50 steps) can produce high-quality images. If you need to quickly preview the generated effect, you can use a smaller number of steps; if you pursue the highest quality results, you can use a larger number of steps.
Guidance Scale: A hyperparameter that controls the degree to which the model obeys the text prompt when generating images. When the value is larger (for example, greater than 7), the generated images may be visually closer to the description of the text prompt, with higher quality and consistency. When the value is smaller (for example, less than 7), the generated images may show more diversity, and the model is less dependent on text prompts when generating images, allowing more creativity and variation.
Images per Prompt: Modify the number of images generated by the model.

5. Generate results

Discussion and Exchange

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓