DALL-E 2, my impressions

Hello friends

I think many of you knows DALL-E already but I want to spend some words anyway. DALL-E is the text to image generator developed by OpenAI (the owners of chatGPT), that actually has implemented Image to image generation and image editing. The image editing feature is in beta testing and it works a kind of the Stable Diffusion model, with adding to the original frame more image content or masking a part of the image generated and modify the selection we made with it.

Now I go to talk about the practical part: DALL-E provides with 50 credits on sign up and 15 credits per month. Each credit can generate 4 variations of up to 1024×1024 pixels images, they allow commercial use of the images generated and they sell 115 credits for 15 us$.

In the early stages of DALL-E there was a waiting list and the commercial use of the images generated was not allowed, now there is no more waiting list and they released the images we generate with them for commercial use.

The prompt used is different from the Stable Diffusion one, I show you:

this is generated with the DALL-E “proper” prompt:

“a young japanese woman holding a smartphone in the style of Tiziano oil painting”  (this is how DALL-E prompt works)

“ink drawing, young chinese woman, intricate, line art, no background, black and white, perfect anatomy, centered image, ” (this is an image generated with DALL-E with a Stable Diffusion prompt, much poorer output image than the right model one)

this image uses the very same prompt of the above image, generated with the proper model (Stable Diffusion 1.5 DreamShaper 3.2 finetuned model)

DALL-E  is not so generous with the image generations with the free account, but I am grateful to them anyway, they are more generous than before, when they did not allow commercial use of the images generated with them. With drawing hands performs a lot better than Stable Diffusion which does not know how to draw hands, it get confused very easily with fingers and hand shapes.

I prefer using Stable Diffusion but I like DALL-E too, actually I like it a lot 🙂 Of coures NSFW is not allowed on DALL-E too….

Never mind which image generation tool I use I enjoy all of them a lot and I hope with this blog mo make you want to try to make stuff with AI, maybe you would enjoy it too

hugs

Alessandro