2023 has been a big year for online image generation using generative pre-trained transformers linked to diffusion models . Here is a roundup of some of the most useful online tools offering this kind of advanced image generation. They are presented in rough order of usefulness. Each model was prompted with the prompt “medium closeup of a female office worker with short pink hair and purple eyes, shot against a professional looking boardroom background” to produce comparable images where possible.

Commercial Generators

Leonardo

• Location: https://leonardo.ai/
• Free images: 20-40 images, daily renewing
• Features: Negative prompting, image prompting, styles, models, pipelines, additional tools
• License: Commercial use allowed

Probably one of the best and most flexible of the online generators currently, it has a wide range of models and pipelines. it also has useful tools such as background removal. It also allows commercial use of images even without a paid plan (though of course the rights status of generated images are in a strange legal position at present)

Playground

• Location: https://playgroundai.com/
• Free images: 500 daily renewing
• Features: Negative prompting, image prompting, styles, models, pipelines, inpainting, additional tools
• License: Commercial use allowed

Playground has an image generator alongside other tools the generation is normally fast and of good quality but its interface can be a little fiddly at times.

Night Cafe

• Location: https://creator.nightcafe.studio/
• Free images: 5 per day
• Features: Negative prompting, image prompting, styles, models
• License: Commercial use allowed

One of the early image generators to become available Night Cafe is fairly easy to use and produces good results though it does not always follow the prompt in all details. You can earn some extra credits, but they no longer offer renewing free credits.

Midjourney

• Location: https://www.midjourney.com/
• Free images: 25 non-renewing
• Features: Negative prompting
• License: Commercial use allowed, but rights are also granted to other users also unless you pay extra for private image generation

Midjourney was an early entrant to the image generation gold rush. However it adopted a somewhat limiting discord based interface and while its generated images are often nice, it suffers from interface issues as a result with a rather clunky chat interface (see this helper document). Additionally its free credits no longer renew and it sometimes excludes new users in the newbie channels with the message “Due to extreme demand we can’t provide a free trial right now. Please /subscribe to create images with Midjourney.” As such the sample image could not be generated so here is an alternative image in a similar vein created by Midjourney to give a sense of its capabilities.

Krea

• Location: https://www.krea.ai/apps/image/realtime
• Free images: about 5 per day
• Features: illusions, draggable elements
• License: Not stated

On the border between a tech demo and a product, Krea allows you to move shapes about to customise where the major elements of a generated image should be. Currently the images are not great quality, but the interface allows easy image manipulation and this might turn into an interesting product in time. Better documentation would be welcome

Tech Demos

Microsoft Bing Image Creator

• Location: https://www.bing.com/images/create
• Free images: 15 per day
• Features: none
• License: Personal Use only

While only allowing 15 image generations per day this does generate 4 images per generation. While the results are reasonable they can look a little artificial, this generator is better for art than photorealism. This is a bare bones generator with nothing but a positive prompt. Also the license is nominally only for personal use, though of course the copyright enforceability around generations is fraught with controversy at present. Another oddity is that at present you do not seem to be able to pay directly for more credits. Mostly a tech demo at this stage, but a useful place to see DALLE-3 in action without upgrading to premium at OpenAI.

SDXL Turbo

• Location: https://sdxlturbo.ai/
• Free images: unlimited but only 512 x512
• Features: speed
• License:Non-commercial use only

A late but interesting entrant in 2023, is another tech demo. SDXL Turbo is a new, extremely rapid generator, which can literally generate images as fast as you can type. The demo site is bare bones and the images are only 512 x 512 and not currently for commercial use (with the usual enforceability caveats), but advances like this may well lead to very reasonable pricing for generative models in future.

Illusion Diffusion

• Location: https://huggingface.co/spaces/AP123/IllusionDiffusion
• Free images: unlimited but queued
• Features: can add patterns for optical illusion shots
• License: None stated

Another tech demo. Primarily designed to create optical illusions, this generator is free, it can develop a queue and be rather slow, but it is useful in a pinch especially for optical illusion shots. The quality of generation is not as good as many other models however

Conclusion

Leonardo and Playground lead the pack, while Night Cafe is still relevant. SDXL Turbo is an impressive tech demo, while Illusion Diffusion is an interesting demonstration of optical illusions. In their current formats Krea,  Midjourney and Bing Image Creator are less useable and useful than the leaders of the pack.

Any other image generators I should know about? Drop a comment.