VisualClozePipeline-384

VisualClozePipeline-384 is one of the models compatible with diffusion models within the VisualCloze framework for image generation with visual context.

Key Features of VisualCloze:

• Supports diverse tasks within the domain and generalizes to new, unseen tasks.
• Combines multiple tasks in a single step, generating both the target image and intermediate results.
• Enables reverse generation — restoring a set of input conditions from a target image.

This model was trained with a resolution of 384. The resolution parameter ensures that all images are resized to the specified dimensions before processing to avoid out-of-memory errors. For high-resolution image generation, we use the SDEdit technology to upscale the generated results.

---

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

Announce Date: 15.05.2025
Parameters: 12B
Context: 77
Developer: VisualCloze
Diffusers Version: 0.33.0.dev0
License: Apache 2.0

Use our pre-built public endpoints for free to test inference and explore VisualClozePipeline-384 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.