@elon_lit: Current image generators start with pure static and use your prompt to steer the denoising. We built a system where the…
Summary
Introducing BIT, a bidirectional image-text diffusion bridge that enables direct transformation from text tokens to images and vice versa, allowing for caption tweaking and structural preservation across modalities.
View Cached Full Text
Cached at: 09/01/26, 03:44 PM
Current image generators start with pure static and use your prompt to steer the denoising. We built a system where the text itself physically transforms into the image. So you can reverse a photo into text, tweak the caption, and morph it back into a new image that natively preserves the original structure. What excites me most is taking this idea to the limit: one can imagine scaling this to an N-way SDE bridge (after doing some gritty multi-marginal math) connecting text, images, audio, 3D meshes, and DNA, basically building Hermann Hesse’s The Glass Bead Game.
Gabe Guo 🦄 (@therealgabeguo): What if text-to-image models started not from noise, but transitioned directly from text tokens to images? 🔠➡️🌆 🤝 What if we unified text-to-image and image-to-text under ONE bidirectional framework? 📚↔️📸
Introducing BIT: Bidirectional Image-Text Diffusion Bridges! With
Similar Articles
Nemotron-Labs-Diffusion-Image: Advancing Masked Discrete Diffusion for High-Resolution Image Synthesis
This paper proposes Nemotron-Labs-Diffusion-Image, a masked discrete diffusion model for high-resolution text-to-image synthesis, introducing a token-editing mechanism and grouped cross-entropy objective to improve token refinement and training efficiency.
Text-to-Image Models Need Less from Text Encoders Than You Think
This paper demonstrates that text-to-image diffusion transformer models primarily rely on token merging and word order from text encoders rather than full contextual embeddings, suggesting that the image model itself decodes complex linguistic structures.
@xichen_pan: Modern text-to-image models are increasingly powered by large pretrained LLMs. But there is a curious mismatch: the LLM…
RepFusion introduces a method to use pretrained multimodal LLMs as noisy representation encoders in diffusion transformers for text-to-image generation, outperforming baselines with similar compute.
@AdinaYakup: SeFi-Image a new text to image foundation model built on Semantic-First Diffusion https://huggingface.co/SeFi-Image - 1…
SeFi-Image is a new open-source text-to-image foundation model built on Semantic-First Diffusion, available in 1B, 2B, and 5B parameter sizes, trained on just 125K A800 GPU hours.
Hierarchical text-conditional image generation with CLIP latents
OpenAI proposes a hierarchical two-stage model for text-conditional image generation using CLIP latents: a prior that generates CLIP image embeddings from text captions, and a diffusion-based decoder that generates images from embeddings. The approach improves image diversity and enables zero-shot language-guided image manipulations.