Skip to Main Content

ChatGPT and other Generative AI tools

Image Generation Tools | This information is adopted from Ithaka S+R’s Generative AI Product Tracker

Name

Purchasing Model

Description

Key Features

Pros

Limitations

Midjourney

Subscription plans, Basic $10/month

Text-to-image generation.

Prof Ethan Mollick says here that Midjourney has lowest learning curve and best quality images.

DALL-E 2, DALL-E 3 OpenAI

DALL-E 2: Use it through purchased credits, after creating an account.

DALL-E 3 built natively on ChatGPT.

Text-to-image generation.

Image generation, outpainting (expanding images), inpainting (making edits to existing images), variation (creating variations of an original image).

Stable Diffusion XL, Stability AI

Download or use on the web via DreamStudio or Clipdrop. DreamStudio and Clipdrop have some free credits, after that purchase credits.

Text-to-image generator.

Image to video generator

Music and sound effect generation

Open source.

Adobe Firefly (web app here)

Pricing for web app here: Free plan with 25/month credits, Premium plan with fee for 100/month credits.

Text-to-image generator.

Access via Firefly web app or Firefly-powered features in Adobe apps (Photoshop, Illustrator, Adobe Express, Adobe Stock, more possible in future).

Image generation, generative fill, add styles/textures to text, generative recolor, sketch to image, generate image from 3D scene.

Trained on Adobe Stock images + openly licensed and public domain content (see FAQ here).

User inputs not used to train models. See Jisc evaluation.

Runway ML

Basic version free, then a few tiers with fees + an enterprise version. Pricing here

Number of different tools that include text to image and text to video.

Craiyon

Basic free version or a few tiers with fees. Pricing

Text to image generation

3 primary modes of image: art, drawing, photo

Intended to be user friendly and accessible

Imagine with Meta

Free with account

Text to image generation. Uses Emu, Meta’s image foundation model.

Concerns over historical inaccuracy and bias in outputs