Peluncuran DALL-E

OpenAI memperkenalkan DALL-E (Januari 2021), DALL-E 2 (April 2022), DALL-E 3 (September 2023) — model text-to-image yang mengubah AI image generation.

DALL-E 1 (2021): autoregressive, VQ-VAE. DALL-E 2 (2022): diffusion, CLIP-conditioned. DALL-E 3 (2023): ChatGPT integration. Pendorong mainstream text-to-image.

Print

Peluncuran DALL-E Series

Ringkasan

OpenAI memperkenalkan DALL-E sebagai seri model text-to-image yang mengubah AI image generation.

DALL-E 1 (Januari 2021)

  • VQ-VAE-2 + autoregressive Transformer
  • 12 miliar parameters
  • Hasil: sederhana, komposisi terbatas
  • Paper: “Zero-Shot Text-to-Image Generation”

DALL-E 2 (April 2022)

  • Diffusion model (CLIP-conditioned)
  • 3,5 miliar parameters
  • Hasil: jauh lebih realistis
  • 4x lebih cepat dari DALL-E 1
  • 512×512 → 1024×1024

DALL-E 3 (September 2023)

  • Terintegrasi dengan ChatGPT
  • Prompt expansion otomatis (ChatGPT menulis prompt detail)
  • Hasil: sangat tinggi
  • Tersedia di Bing Image Creator, ChatGPT Plus

Pesaing

  • Imagen (Google, Mei 2022)
  • Stable Diffusion (Stability AI, Agustus 2022) — open-source
  • Midjourney (Juli 2022) — komunitas artist
  • Adobe Firefly (2023) — terintegrasi Creative Cloud
  • Flux (Black Forest Labs, 2024) — open-source

Dampak

  • Text-to-image menjadi mainstream
  • Adobe, Canva, Microsoft Designer integrasi
  • Industri kreatif — desainer, ilustrator
  • Hak cipta — perdebatan tentang data training
  • Misinformasi — deepfake, propaganda

DALL-E memicu revolusi generative image 2021-2024.

Connected to

Not yet written

The following pages are referenced but don't exist yet — they'd make good future additions.

  • /concepts/generative-ai
  • /sources/openai

References

  1. Wikipedia

Type at least 2 characters to search.

Press to navigate, to open, esc to close.