Generating Images from Long Text
lastmod 2026-11-10created 2019-02-122 edits
GILT (Generating Images from Long Text) is deep learning research that takes long text rather than a short caption, such as an entire cooking recipe, as input and generates a matching image. Its architecture encodes the long text into a conditioning vector and passes it to a GAN generator, and it demonstrated experiments producing photos of finished dishes from a recipe dataset. It is one of the early attempts to extend input length in text-to-image generation.
GILT - Generating Images from Long Text