Logo
LoginGet Started
Profile

shallowdream204 / bitdance

Generate fast, high-resolution, photorealistic images from text prompts using an advanced autoregressive model for efficient, high-quality results.

0.01
GitHub
License
Weights

Model Input

Input

Text prompt for image generation.

The size of the generated image. Will be snapped to the nearest supported resolution.

Additional Settings

Customize your input with more control.

Min: 10 - Max: 100

Number of diffusion sampling steps per decoding step. Higher values (e.g. 50) improve quality at the cost of speed.

Min: 1 - Max: 15

Classifier-free guidance scale. Higher values follow the prompt more closely.

Random seed for reproducibility. The same seed and prompt will produce the same image.

Min: 1 - Max: 4

Number of images to generate.

Safety checker can only be disabled on API call

The format of the generated image.

You need to be logged in to run this model and view results.
Log in

Model Output

Output

preview
preview
Generated in 330.658 seconds
Logs (1 lines)

Model Example Requests

Examples

msDF2fhas9e8YhWwtKYkz2lcPOKZmQMToYQ843xsIcHlIPZjA1JDAYuo6cSehj9

Model Details

Model Details

BitDance is a powerful autoregressive model designed for generating high-resolution, photorealistic images from text descriptions. It leverages advanced techniques to produce fast, high-quality visuals, making it an excellent tool for artists, designers, and creators who need to bring their textual ideas to life.

The model's core capability is transforming a descriptive text prompt into a detailed image. You can guide the generation process with a high degree of control through various parameters.

### Key Controls

* **Prompt**: The primary input. The more descriptive and detailed your prompt, the more tailored the resulting image will be. Specify subjects, settings, artistic styles, lighting, and composition. * **Image Size**: Choose from a variety of aspect ratios, including square, portrait (4:3, 16:9), and landscape (4:3, 16:9), to fit your specific needs. * **Inference Steps & Guidance Scale**: Fine-tune the image generation process. `num_inference_steps` controls the trade-off between speed and quality, with higher values producing more refined images. `guidance_scale` determines how strictly the model adheres to your prompt. * **Seed**: Use a specific seed to ensure reproducible results, allowing you to generate the same image again or make subtle tweaks from a known starting point. * **Number of Images**: Generate up to four images from a single prompt to explore different variations and select the best outcome.

### Getting the Best Results

For optimal results, write clear and specific prompts. For instance, instead of "a cat," try "A close-up portrait of a fluffy calico cat wearing a tiny top hat, cinematic photography style, soft natural lighting." Experiment with the `guidance_scale` to find the right balance between prompt adherence and creative freedom. Using the `seed` parameter is highly recommended when you want to iterate on a promising result.

To run via the Javascript client, use the following code: ```javascript import { modelrunner } from "@modelrunner/client";

const result = await modelrunner.subscribe('shallowdream204/bitdance', { input: { "prompt": "A vibrant, detailed painting of a mystical forest at twilight, with glowing mushrooms and ethereal creatures, in the style of fantasy concept art", "image_size": "landscape_16_9", "num_inference_steps": 30, "guidance_scale": 8.0, "num_images": 2 } });

console.log(result); ```