Dalle2, MidJourney, Stable Diffusion - Quick recap of the 3 major text to image models

You can use them all today, cost varies between free (bring your own hardware) to 3¢ per image

2022 is the year of ‘text to image models’. You can enter a prompt like ‘black and white photograph of a teddy bear skating in Paris’ and get images back.

Things change fast in the world of AI image generation currently. So here is the basic info on all 3 major text to image models and how you can use them.

Dalle2

Website: https://openai.com/dall-e-2/

Use it today?: Yes. They removed the waitlist.

Cost: $0.032 per image or $3.20 for 100 images

API: Updated, yes as of Nov 4th, 2022. Previously did not offer. Microsoft got access for their Designer Tool, but no one else has. Using methods to get around that is against their terms of service and could get you banned.

Creator: OpenAI https://openai.com/

Midjourney

Website: https://www.midjourney.com/home/

Use it today?: Yes. Open beta.

Cost: $10 a month for 200 images. Or $0.05 an image. High plans have unlimited jobs for slower performance.

API: No API announced yet.

Creator: Midjourney https://www.midjourney.com/home/#about

StableDiffusion

Website: https://github.com/CompVis/stable-diffusion

Use it today?: Yes. Open source.

Cost: Free, as it is open source. You have to run it on your own or rented hardware. If you want to use the official web client (dreamstudio.ai/dream), it is about $10 for 1000 images or $0.01 per image.

API: Yes. Anyone can run this on any computer that supports it.

Creator: Stability AI, with collaborators. https://stability.ai/