Same problem. When I asked why, I got this response:
“The WEBP format is used because it’s a modern image format that provides superior lossless and lossy compression for images on the web. It allows for smaller file sizes while maintaining a high level of image quality. This means faster loading times and less bandwidth consumption, which can be especially beneficial when dealing with high-quality images like the ones generated here.”
So I told the GPT: “in future create the image in .png format”. Response was: “Understood! For future images, I’ll ensure that they’re generated in the .png format”.
Asked it to generate a new image, but again in WEBP. My response: “It is still in WEBP format. Are you not to be trusted?”
Reply from GPT: “I apologize for the confusion and any inconvenience caused. It seems there was a misunderstanding. While my intent was to ensure images are suitable for your needs, the format of the images generated by the tool I use (DALL-E) is currently set to WEBP by default, and I don’t have the capability to change the output format directly.”
Unfortunately no closer to knowing why DALL-E’s default output is now set to WEBP.
Same problem, everything is in WEBP fortmat!!. Also It looks like something chaged, it is creating very different images compared to the last days, with low quality and less details
You can also ask the AI what its knowledge cutoff is. It will say that it can’t answer about things that OpenAI changed under 24 hours ago.
ChatGPT only writes the image prompt to send for you. It doesn’t have any control of the image format.
What was a PNG in your chat history from a few days ago is now downloaded again as a webp.
I did a subtraction of the prior image from the current one, and the difference is black - the image data is nearly identical.
The problem is that there are so few applications (besides a web browser) that understand the webp format. Its like Opus audio: better, but not going to work in your car’s mp3 player.
I use CS3, and I have yet to find a plugin which allows me to use WebP. I also use Win7, so all the programs which convert WebP to PNG won’t run for me. This means I now have to take any image produced by DALL-E and convert it on an external website, adding an extra step to my workflow. Absolutely haram.
Opus audio, that’s a pretty cool analogy, huh?
But I can’t tell the difference, so I’d prefer the more common PNG or JPG format.
After all, PNG is almost lossless, right?
Hey everybody, well it seems the webp issue is an annoying problem for all of us. I think it is very disrespectful of OpenAI for not communicating this obviously big change to their subscribers and developers.
In the mean time I found somewhat a solution for converting the webp images to png with a free software program.
It is called XnConvert and it can convert images on the batch within seconds. It is freeware but they also have a version for donators. I don’t know what is exactly the difference as the free version worked fine for me.
I’m not sure why this change was made, we should have been informed in advance to determine what we want to do with our subscriptions.
Conjectures are currently (a) saves disk space into the future reducing storage requirements and reducing costs associated with that and media replacement costs etc. (b) saves disk space into the past reducing storage requirements and reducing costs associated with that much like (a). B saves immediately on maintenance requirements by freeing up a ton of disk space reducing cloud fees etc. Are you going to give that back to the community or pocket the saving and reinvest, or are you going to take that to the bank in the form of increasing profitability in the valuation race of tech company culture?
I have a solution, insert a WEBP > PNG converter in the download pipeline which most people will use anyways, but only as long as the WEBP is lossless. Also give people the option to set the output format. Be transparent in communicating that the WEBP change is lossless front to end, because for those of us using scalers for professional print production, this is critical.
We are the platform’s champions and some of us are its influencer’s in a community. Support us please!
I found a way to save the file as a png until they are able to fix things. Populate your image as usual, click the image to view the bigger version, then save the image using the snipping tool. So far it has been working for me and no loss of quality.
Is it just me of has the quality of created images went down as well? It seems like DALL-E has forgot how animals look like all of a sudden and keeps creating these weird mutant blobs all day
We’re probably being throttled to give more service to their enterprise clients. We were used to build hype, but now the billions have started to roll in from their well-heeled corporate clients, we’re just an inconvenience for them. If they didn’t need us to keep feeding their alogorithms, they’d probably just kick us off. People who have access to the enterprise version say it’s like night and day. We’re digital plebs.
Starting today, February 4th, GPT has been downloading images created in DALL-E in .webp format, whereas until yesterday, the format was .png. Photoshop struggles with .webp format and I prefer not to rely on third-party plugins. As a subscribed GPT user, I hope this is just a minor bug because the .png format is essential for us! Let’s bring attention to this issue/thread so at least someone will get us more information on this issue.
UPDATE:
I’ve decided to use this plugin for Adobe Photoshop, directly from the Google WebM project . I think it’s the best solution for me right now, as I want to avoid uploading images to third-party sites for conversion. If anyone needs a temporary workaround, here it is.