Meta has launched an AI image generator. We compare it with Midjourney and DALL-E 3
The new "Imagine with Meta" tool transforms English text prompts into visual images. Users need a Meta account and should activate a VPN set to the United States for access outside the US.
✅ Imagine is trained on 1.1 billion public images from Facebook and Instagram. The neural network generates 4 image options from a text prompt. All with the watermark "Imagined with AI" to avoid issues with content authenticity.
🧑💻 Testing
It handles simple prompts well; with more complex ones, the difference with Midjourney and DALL-E 3 becomes apparent.
During the experiment, we gave each neural network 4 different prompts and selected the best option from the generated images.
Decide for yourself who won:
🔥 — DALL·E 3
👍 — META
❤️ — Midjourney
P.S. You can test Imagine with Meta yourself on the website; Midjourney and DALL-E 3 are available in @GPT4Telegrambot
