Multimodality · 2021
Zero-Shot Text-to-Image Generation (DALL-E)
Showed that a single autoregressive Transformer over combined text and image tokens could generate coherent images from natural-language captions, demonstrating zero-shot text-to-image generation at scale.
Editorial record
Plain-language summary
DALL-E encodes images as discrete tokens and trains a large Transformer to model the sequence of text followed by image tokens, then samples images from captions and reranks them with CLIP. It produced novel, compositional images without task-specific training. It opened the modern text-to-image line later dominated by diffusion methods.
Knowledge graph
Relationships
Descendants
Depends onEvidence: Direct
Attention Is All You Need
DALL-E autoregressively models image tokens with a Transformer
P-530
Applies toEvidence: Direct
CLIP: Learning Transferable Visual Models from NL Supervision
CLIP is used to rerank DALL-E samples
P-530
Source record
Provenance
- Record ID
- P-530
- Record created
- 2026-07-13
- Last reviewed
- 2026-07-14
- Record version
- 2
- https://arxiv.org/abs/2102.12092
- arXiv:2102.12092
Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.