karand120497 / glaze

Glaze is langchain for images. If you're building a text-to-image app with LLMs (Stable Diffusion v2.1/v1.5, ControlNet) or an image-to-text app( ViT GPT2 image captioning), use functions out-of-the-box.

Glaze Chain

This library is a way to abstract your workflows for images. You can use text to image and image to text models, set settings for your images and set guardrails for outputs.

Our principles are 2 fold:

1. Interact how large models interact with images.
2. Make Sherlocks which act as agents to perform tasks.

For text-to-images we support the following models:

1. Stable Diffusion v2.1
2. Stable Diffusion v1.5
3. Stable Diffusion ControlNet

For image-to-text we support

1. ViT GPT2 image captioning.

This works with single/multiple prompts. You can also use 1 or multiple images stored in your local storage to perform.

Leave a comment/ ping us [e-mail] or [e-mail] for feature requests!