DALL-E 3: what owners actually say

Multiple users note that while DALL-E 3 follows prompts better than MJ or SD, the actual output quality is not as visually impressive — one commenter summarized it as 'better at following prompts than MJ or SD. But the output quality not so much.'

Users carry skepticism from DALL-E 2, which they say used heavy cherry-picking in marketing while the actual product was 'fairly terrible and difficult to get any good results.' Several worry DALL-E 3 demos may follow the same pattern.

One tester observed that AI generators look great in cherry-picked showcases or with low expectations, but 'the more you have a specific image in mind or the more you want certain details the worse it gets. And you quickly realise the limitations.'

Users discovered OpenAI injects an internal prompt that forces diversity into depictions of people and removes requests for specific descent, which overrides user intent without disclosure.

One commenter found generative art 'really frustrating' and felt typing long prompts was 'not worth the effort,' while another noted the 'create text version of image' intermediate prompt matters enormously and requires extensive tweaking.

Users consistently note DALL-E puts the things you actually asked for in the image, unlike SD which 'sometimes no matter what you tell it to do, it just doesn't do it.' Even critics concede this advantage.

Multiple commenters were blown away by text rendering, calling the avocado example 'insane' and saying they had 'yet to see any image prompt AI get text and composition anywhere near this level.'

A software engineer with 'absolutely no eye for design' reports being able to write a prompt and get a 'good-enough image in 90% of cases' that was previously unreachable without hiring an artist.

Users appreciate that ChatGPT translates natural language into effective image prompts, making it significantly easier to use than crafting raw prompts for SD or MJ.