Same issue. Why change up? It’s silly to make users have to go to another add on or a separate app…
April 6: I can’t even download an image I’ve created. The download arrow no longer works.
Dall-E isn’t the only image generator…OpenAI is playing with fire here…
Just install Imagemagick, then run the following command in the folder where the images are saved: magick mogrify -format png -define png:compression-level=0 -define png:compression-filter=0 -define png:compression-strategy=0 *.webp
After you create the image ask Chatgpt to convert the image to png or jpeg.
It is simply barefaced from openai to use this webp, only to push google products. It is not even using the lossless variant. And this for a paid service! Give the customers the choice! i want the ability to export a lossless PNG format (not webp or jpg). For this little bit traffic more, set up some more servers if needed. And if people want webp or jpg, give them the options.
Even the tools form google not allow simple things like rotating a image lossless (Dalle sometimes exports pictures in wrong rotation…), or conversion in PNG. Export in jpg is now with double loss, this is unprofessional.
In addition, the meta data include no useful information’s. the Meta data should include at least:
The text prompt (in original and translated language if needed, or what ever Dalle used to create the image.)
The seed, so it is possible to regenerate the exact picture again. this makes it possible to regenerate the same image in a later version with higher resolution and quality.
(And as a extra, GPT give 3 times wrong advises how to rotate the images lossless, cwebp.exe for example, all tools not have such options, to then found out, there is no way, and the webp is used in lossy format, and rotate it is again double loss…)
Pleas, give us the option for png back!
(and put in the meta data all important infos, or let us download a log file with the infos)
Tanks!
You had the power the whole time, Dorothy.
- Generate a full-body-length image of a winged pixie in a detailed forest with dalle.
- Use your built-in python tool with Jupyter notebook to write and execute code with PIL (pillow) to:
- load mount point image,
- then rotate mount point image 90 degrees clockwise;
- resize width to 1920px and maintain aspect ratio;
- save to PNG with metadata you sent about the prompt employed as description;
- provide download link to image as notebook output
- display revised image from the code
—>
Sorry, but i not want mess around with python, or add prompts and seed numbers manually to a image. Exactly for such repetitive and boring word computers are made. The system should but all information’s like prompts and the really used Text, seed number date etc in the meta data.
It would even be a good feature, if a future Dalle version can recreate a image exactly, with higher resolution or details. this meta data would come with 0 costs for the system.
I think everybody has some favorites and would like them in 4K. Now this is not possible, because of missing meta data.
And Dalle 3 should be able to rotate pictures correctly…
(It is a common behavior of some developers to let millions of users do work around’s, instead of simply fix the issue. The most worse is if developers see a feature request or bug report as a personal offence or criticism. If sometimes users load down all there frustrations, this is surly a bad attitude of humans. For this, professionals should filter all aggression out and just deliver pure information’s to the developers. Now AI can do this automatically…)
(Would a feature in GPT it self be useful, for feature requests or bug reports? I think yes.)
