diffusion-based high-resolution image synthesis
High-resolution image synthesis using latent diffusion models: research and training code for diffusion models operating in a learned latent space conditioned on text, images, layout, etc.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
few-step latent diffusion model when sampling speed matters, but requires consistency-distilled training instead of LDM's full framework
Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference
a maintained unified inference/training API for LDM-family and other diffusion checkpoints instead of the original monolithic scripts
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.
progressive multi-scale fusion to extend existing diffusion models to megapixel outputs without retraining
Let us democratise high-resolution generation! (CVPR 2024)
DALL-E 2 style prior-and-decoder text-to-image if you prefer that architecture over latent diffusion
Implementation of DALL-E 2, OpenAI's updated text-to-image synthesis neural network, in Pytorch
efficient high-resolution diffusion transformer with current checkpoints and faster sampling than the original LDM pipeline
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
bitwise autoregressive high-resolution image synthesis if you want an AR alternative to diffusion-based LDM
[CVPR 2025 Oral]Infinity ∞ : Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
a ready-to-use GUI/CLI inpainting app powered by diffusion, choose it when you need a user-facing tool rather than LDM research code
Image inpainting tool powered by SOTA AI Model. Remove any unwanted object, defect, people from your pictures or erase and replace(powered by stable diffusion) any thing on your pictures.
These projects were analysed and named latent-diffusion among their alternatives. The relationship is not symmetric — how latent-diffusion rates them is a separate judgement, made when latent-diffusion is analysed in its own right.
calls latent-diffusion “Original latent diffusion research code; choose it for the classic LDM architecture and full training flexibility.”
Implementation of DALL-E 2, OpenAI's updated text-to-image synthesis neural network, in Pytorch
calls latent-diffusion “The original latent diffusion research codebase; choose it for reproducing the LDM papers rather than for production.”
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.
calls latent-diffusion “Choose when you want the original LDM research/training codebase rather than a modern text-to-image/text-to-video system.”
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
calls latent-diffusion “Choose to study or extend the original latent diffusion architecture, not for a production toolbox.”
OpenMMLab Multimodal Advanced, Generative, and Intelligent Creation Toolbox. Unlock the magic 🪄: Generative-AI (AIGC), easy-to-use APIs, awsome model zoo, diffusion models, for text-to-image generation, image/video restoration/enhancement, etc.