Shuttle 3 Diffusion

Shuttle 3 Diffusion is a text-to-image generation model designed to create detailed and diverse images in just four steps. It offers enhanced image quality, understanding of complex prompts, efficient resource usage, and increased detail.

Key features:

• Built on the Flux.1 Schnell architecture, it has been partially de-distilled and trained using specialized methods to improve detail and color reproduction compared to earlier Schnell series models.
• Refiner mode, activated when using more than 10 generation steps, enhances image details without altering the composition.
• Licensed under Apache 2.0, making it more liberal than models like Flux Dev.

Example text prompts showcasing the model’s capabilities:

• “Venerian floating market at dawn, fantasy digital artwork, high detail…”
• “Quiet forest, sunlight piercing through tree canopies, mysterious dark-red-toned lake…”
• “Night jungle waterfall with a blue tint, shimmering fireflies, photographic realism…”

---

The model is a component of the image generation pipeline, consisting of:

• CLIP text encoder: ~123M parameters,
• T5 text encoder: ~5B parameters,
• Transformer: ~12B parameters,
• VAE: ~84M parameters.

Total: ~17B parameters

---

The 'VRAM requirements' line is calculated based on the size of the weights in the specified precision, with an additional +10% coefficient for inference:
  model_parameters_count * bytes_per_parameter * 1.1 / 1024^3 
where model_parameters_count is number of model parameters, bytes_per_parameter is number of bytes used to store one parameter. For 16 bit = 2, 8 bit = 1, 4 bit = 0.5.

Announce Date: 12.11.2024
Parameters: 12B
VRAM requirements: 8.6 GB using 4 bits quantization, 17.3 GB using 8 bits quantization, 34.5 GB using 16 bits quantization
Developer: ShuttleAI
License: Apache 2.0

Use our pre-built public endpoints for free to test inference and explore Shuttle 3 Diffusion capabilities. You can obtain an API access token on the token management page after registration and verification.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.