https://github.com/gsbm/autoassets

🟠🌸 On-browser textured 3D models generation from existing images or textual descriptions
https://github.com/gsbm/autoassets

3d-generation 3d-model generative-ai image-to-3d image-to-mesh text-to-3d text-to-mesh

Last synced: 2 months ago
JSON representation

🟠🌸 On-browser textured 3D models generation from existing images or textual descriptions

• Host: GitHub
• URL: https://github.com/gsbm/autoassets
• Owner: gsbm
• License: gpl-3.0
• Created: 2025-08-02T12:21:37.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2026-02-06T20:18:22.000Z (5 months ago)
• Last Synced: 2026-02-07T05:39:52.391Z (5 months ago)
• Topics: 3d-generation, 3d-model, generative-ai, image-to-3d, image-to-mesh, text-to-3d, text-to-mesh
• Language: CSS
• Homepage: https://gsbm.github.io/autoassets
• Size: 3.17 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          


 Autoassets


Autoassets is a web application that allows users to quickly generate 3D models from simple textual or visual descriptions using state-of-the-art AI models. The platform integrates with Hugging Face Spaces and leverages models such as Stable Diffusion 3 Medium, Stable Diffusion XL, FLUX.1, Qwen Image, Chroma, InstantMesh, and TRELLIS to create high-quality 3D assets from user prompts or images.

## Documentation

### Parameters

Below are all the parameters available in Autoassets, with explanations of what each does:

- **Prompt**: The main textual description of the 3D model you want to generate. Used as the primary input for image generation models.

- **Image Upload**: Allows you to upload a PNG or JPEG image as a visual reference for 3D model generation (used with the "Image" input type).

- **Image Sampler**: Selects the AI model used for image generation (e.g., Stable Diffusion 3 Medium, SDXL, FLUX.1). Determines the style and quality of the generated image.

- **Mesh Builder**: Selects the AI model used for mesh generation (e.g., InstantMesh). Converts generated images into 3D meshes.

- **Additional Prompt**: Extra text to enhance or guide the generation (e.g., "(3D model), (overall view), (white background), studio photo, blender, 3D render, large view, (single object)").

- **Negative Prompt**: Text describing what you do NOT want in the generated model (e.g., "longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality").

- **Sample Steps**: Number of steps used in mesh generation (affects detail and quality). Range: 30–75. Higher values may improve quality but take longer.

- **Use Random Seed**: If checked, a random seed is used for generation, making results less predictable. If unchecked, you can specify a seed for reproducible results.

- **Seed**: A number (0–1,000,000) used to initialize the random number generator for reproducible outputs. Only enabled if "Use Random Seed" is unchecked.

- **Width**: Width (in pixels) of the generated image. Range: 256–1024.

- **Height**: Height (in pixels) of the generated image. Range: 256–1024.

- **Guidance Scale**: Controls how closely the output matches the prompt. Higher values make the model follow the prompt more strictly. Range: 1–20.

- **Number of Inference Steps**: Number of steps for the image generation process. Higher values can improve quality but increase generation time. Range: 10–100.

- **Hugging Face Access Token**: (Optional) Your Hugging Face API key. Providing this can increase your GPU request limits for faster or more frequent generations.

Some models (like SDXL) may have additional advanced parameters:

- **Guidance Scale Refiner**: Guidance scale for the refiner stage (SDXL only).

- **Number of Inference Steps Refiner**: Number of inference steps for the refiner stage (SDXL only).

- **Apply Refiner**: Whether to use the refiner stage for improved image quality (SDXL only).

## Legal

### Licenses

Autoassets and its dependencies are licensed as follows:

- **Autoassets**: GPL-3 License (see `LICENSE` file).

- **Stable Diffusion XL**: MIT License.

- **Stable Diffusion 3 Medium**: MIT License.

- **FLUX.1-schnell**: MIT License.

- **Qwen Image**: Apache License 2.0.

- **Chroma**: Apache License 2.0.

- **BLIP 3o**: Apache License 2.0.

- **InstantMesh**: Apache License 2.0.

- **TRELLIS**: MIT License.

Users retain ownership of the 3D models they generate, subject to the licenses and terms of use of the generative models provided by Hugging Face, SDXL, and InstantMesh.

### Terms of Service

- No user registration or login is required.

- All user data is sent to Hugging Face for processing and may be stored as analytics by them.

- The platform is intended for experimental and educational purposes only.

- Users are responsible for ensuring their use of generated models complies with all relevant licenses and laws.

- For more details, see [Terms of Service](https://gsbm.github.io/autoassets/legal.html).