text-to-image synthesis
PyTorch implementation of OpenAI's DALL-E 2 text-to-image model using a CLIP-based diffusion prior and diffusion decoder, with training code.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Broad, actively maintained diffusion library; choose for a standardized multi-model ecosystem instead of a single DALL-E 2 codebase.
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.
Efficient high-resolution linear-diffusion transformer; choose for modern architecture and Apache-2.0 licensing.
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
Few-step consistency distillation; choose when generation speed is more important than DALL-E 2's slower prior/decoder pipeline.
Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference
Bitwise autoregressive image synthesis; choose if you prefer autoregressive modeling over diffusion.
[CVPR 2025 Oral]Infinity ∞ : Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis
Original latent diffusion research code; choose it for the classic LDM architecture and full training flexibility.
High-Resolution Image Synthesis with Latent Diffusion Models
Latent text-to-image diffusion model with a huge ecosystem; choose for easy deployment, community support, and looser compute needs than DALL-E 2.
A latent text-to-image diffusion model
Unified multimodal model with text-to-image generation; choose when you need understanding, generation, and editing in one system.
An unified model that seamlessly integrates multimodal understanding, text-to-image generation, and image editing within a single powerful framework.
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Compact port of the older DALL-E Mini; choose for tiny/fast CPU inference when lower visual quality is acceptable.
min(DALL·E) is a fast, minimal port of DALL·E Mini to PyTorch
Legacy CLIP-guided diffusion; choose for a minimal not-to-image codebase, but fidelity lags DALL-E 2.
A CLI tool/python module for generating images from text using guided diffusion and CLIP from OpenAI.
Stale DALL-E 2 replication/checkpoint repo; use only if you specifically want its long-abandoned pretrained weights.
Pretrained Dalle2 from laion
First-generation DALL-E in mesh-tensorflow; choose only if you are tied to TF/TPU and need the original DALL-E.
Open-AI's DALL-E for large scale training in mesh-tensorflow.
These projects were analysed and named DALLE2-pytorch among their alternatives. The relationship is not symmetric — how DALLE2-pytorch rates them is a separate judgement, made when DALLE2-pytorch is analysed in its own right.
calls DALLE2-pytorch “DALL-E 2 style prior-and-decoder text-to-image if you prefer that architecture over latent diffusion”
High-Resolution Image Synthesis with Latent Diffusion Models
calls DALLE2-pytorch “Use it if you specifically want DALL-E2's prior/decoder architecture in PyTorch rather than latent diffusion, accepting an unofficial and unmaintained codebase.”
A latent text-to-image diffusion model
calls DALLE2-pytorch “Choose when you want to train/experiment with an unCLIP/DALL-E-2-style architecture instead of Sana; research code with no maintained pretrained weights.”
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer