DALL-E 3 Is Now Free For Everyone!

Great news. The new image generator AI DALL E3 is now available for everyone for free. I've put a link to it in the video description. Make sure to check it out. And it already has some surprises for us. So we are going to look at six unexpected results. First, I read that supposedly it is excellent at visualizing proverbs. For instance, laughter is the best medicine. Hmm, what could that look like? Oh, do you remember when we tested Photoshop's image generator and I asked for an image with caustics, colorful patterns of reflected light that should look like this, and we got this instead. Well, similar phenomenon here. Hope it gets sorted out now. Luckily, just a week ago someone ran the same prompt successfully and this is truly excellent. Note that it did not get detailed instructions on how to do this. Just laughter is the best medicine. You figure out the rest. Little AI plus points for the old comic book style. So good. Midjourney cannot do this. Not even close. 2. What about character consistency? We got a taste of this with the Amazing Larry the Hedgehog. However, that was in the promo materials. Does it also work well in practice? Oh my, yes. Yes it does. This is the same family in four different settings. Once again, this requires the implementation of research papers or at the very least a series of tricks in most other text to image systems. I have to note that I tried it too, and in my case it was a little copy pasty. You know, it's easy to keep consistency if you just copy the same thing over and over. The images are also of somewhat lower quality. So my impression is that this can be done, but requires a little cherry picking. That is a huge step forward. Next step will be that we take it for granted, especially that 3. This consistency can be requested in even more advanced ways. For instance, the same character with different ages. This is one of the best text to image AI result I have ever seen. Goodness. And hold on to your papers, fellow scholars, because it gets even better. If you feed these images into Runway's interpolation technique, you get a fantastic video. I put a link to these results in the video description, many of which have prompts available which will help you in your own scholarly experiments. 4. I was surprised by how good it is at sketches. This one is incredible. And of course we will get sketches of you fellow scholars as well. Always. But you know what is missing here? Fellow scholars is great, but fellow scholars holding onto their papers is even better. Oh, look at that. 5. Text. Text is finally working. A historic moment. Glorious. And finally. 6. It is surprisingly imaginative. It can cook up an image of a relaxing parrot. Easy. There is room for the tail and the legs are up in the air. Really cool. Asking for a base on the moon, not a problem. And paintings are also incredible. This one just radiates with joy. Love it. Digital art. It cannot only do it, but it is super imaginative. And now, what you all have been waiting for. As is tradition, our legendary stern Fox librarian from Dall? E 2 can now be created again. Will it get even better? Oh, it is really good. And he has gotten a lot more stern this iteration. I can tell you that right now. So yes, Dolly 3 is really good. No, it is not perfect, but really good. Yes, it might require a little cherry picking. And one more. And one more. There we go. Thank you, little AI. And just one more thing. We are in the age of AI and I hear more and more rumors about myself being replaced by an AI who is now narrating these videos. So let it be known that every episode is being recorded by my own voice. We are sponsored by Weights and Biases, experiment tracking, model evaluation and production monitoring for your deep learning products and LLM apps. It is the the best. Everyone is using it. Try it out now at WNB Me Papers or click the link in the description below.