Which one is better? Are you surprised by the new DALL-E? Should it be integrated into Bing?

@austinlevine

Жыл бұрын

I loved the images it made, "totally in bing chat" 🫤🤣

@robaustin_

Жыл бұрын

It's closer "than" you think.

@camfuinrules

Жыл бұрын

bruh the new dalle img.... renders the downtown area of my small city.... by name....... downtown Kelowna..... and it does it well.... try asking it to render an area you know of.. it must have a controllnet or something... querry prompt for existing images on line.. use that img as template for render ... or something.. but this seems advanced to me

@kuko61

Жыл бұрын

I've been doing some experiments now. With Bing Image Creator, the quality of some images is close to MJ5. But for some types of content it lags behind. Figures and faces are very bad. I'm missing more control over the image, e.g. aspect ratio, negative prompt.

@jlyn8228

Жыл бұрын

I'm surprised you used the wrong than more than all.

Microsoft bing keeps coming out of nowhere to clap Google's cheeks.

@MattVidPro

Жыл бұрын

literally

@imanigordon6803

Жыл бұрын

@@MattVidPro google can’t breathe with all these updates 😂

@solaawodiya7360

Жыл бұрын

😂😂😂

@TheMediaLogoEditor928

Жыл бұрын

Google does have an image generator too, and a video generator

@solaawodiya7360

Жыл бұрын

@@TheMediaLogoEditor928 Where? On Bard?

I like how the different models all seem to be good at different things.

I wanna see Dalle 2 go open source 😢

@Relics

Жыл бұрын

LOL

@curioza1

Жыл бұрын

No because *open*ai wouldn't do that

@benoitcorvol7482

Жыл бұрын

Coming soon

@Truth_Unleashed

Жыл бұрын

If only Sam Altman wasnt in charge.... 😂

@7TheWhiteWolf

Жыл бұрын

@@curioza1 They were supposed to destroy the Sith, not join them.

My favourite use for DALL_E is for generating textures because of the amount of micro details that are in its generations compared to SD and MJ.

@nahiddotai

Жыл бұрын

Interesting. I've been able to generate awesome textures/backgrounds using MJ!

@apatsa_basiteni

Жыл бұрын

@@nahiddotai awesome

something I noticed is that midjourney generates really good images, but they are much more similar to each other. A great example of that is how all 4 Walter White generations were in the hazmat suit.

@miyegombodashdondog923

Жыл бұрын

You can change that with 'chaos' command

I can't wait to give them both a try Matt, thank you so much for explaining each tool so well! You made the whole process fun😄 What are your thoughts on BlueWillow? Reading through your comments, I am glad to see I am not the only one interested in this new tool and what it can do. Please share in your next clip? I am really looking forward to it😄

matt, you're so fast uploading videos, Well-done !

I love your disgruntled frog detective v3

I'm waiting for an update to the image editing features of Dall-E. The image editing is so easy to use to correct problems left by other art generators but the current version we have had for a while makes the images look a little blurry…

MJ quality is great, however the main downfall with MJ is that the results are most always similar. I'm getting good results starting a prompt in Bing Image Creator then using the output in MJ to get something even better that is otherwise unique from what MJ would have given me.

@limeflashlight4101

Жыл бұрын

Mj is basically adding its own set of what it thinks look good descriptors to every one of your prompts behind the scenes too

@DFortuna

Жыл бұрын

​@@limeflashlight4101yeah if you want to remove MJs influence use --stylize 0

big improvement over dalle 2, but i personally still think its behind stable diffusion (particulary the illuminati model) and yea, midjourney is still king, but the competition is fierce and im having a blast trying all these models

Thank you, sir Matt o7 I wasn't expecting this from Microsoft so soon, although it's not that surprising since they have been talking about image generation coming to their products for a while now. I shouldn't really be doing any comparisons, I've only played around with Stable Diffusion before, but Bing generated decent images with much less tinkering. For example I was happy when I finally got my "recursive tree folding in on itself" right off the bat (YMMV), even if it's not a very good or descriptive prompt 😅 Have fun generating, people!

Is Bing Create going to be free like Blue Willow? To what extent do the freebie services aim low to save money, vs. aim low to push people toward premium plans?

I love the way the credit system is made, even if i prefer daily credits like 5 credits per day rather than 30 per week, but its really good. What i would love to see is a model that can do pixel art, and like a model that is really focused on pixel art. As i could see this new Dalle is quite great at pixel art, but i would really want something like midjourney V4 at least but specialized with Pixel art, that's the AI iam waiting for ❤❤

@hartvenus

Жыл бұрын

There are already stable diffusion models fine tuned for pixel art, see: aseprite diffusion

It does work Matt but you have to be in creative mode, I just tried it. Impressive!

Your personality is a spitting image of "Bender" from the movie, "The Breakfast Club", dude,....Had to subscribe!!! 😁😁😁 (don't take it in a bad way, I think it's really cool!) Keep up the good work!!😊😊😊

Dude, Google is sooo f7cked... MS integrating image generation into Edge? Holy crap! MJ prob not too happy right now either.

@solaawodiya7360

Жыл бұрын

Yeah. Google feels so outdated right now. It's so weird that when I think of Google, the word "old" comes to my mind 😂😂😂

I've been playing with it all morning, and it kicks the crap out of Dall-e 2 but is maybe less "cinematic" than the latest mid-journey? But that lends itself to make it look more realistic honestly because mid-journey tends to favor perfection too much to fool anyone lol

Man, I feel like we're getting spoiled. Just imagine how much were going to take for granted 5 years from now.

Thanks for all this wonderful info. I've been wanting to find and try some ai image art and a free Bing option is a good fit. It pulled off most of my prompts very well, although anything related to people on a beach it kept telling me I was asking for inappropriate images...granted I did not specify what they should be wearing, but I did not expect it to assume that all humans are nude on the beach. I think this ai hit fed some adult content or something. Aside from that hiccup, I did get some great ai images of landscapes and silly things with Bichon Frise puppies which it did very well.

If you use Bing gpt 4 you can actually add on to images if you want to add a feature or change a location

So Bing just inherited the newest DallE model? Am I understanding this correctly? Bing Image creator is by far much better than original Dalle 2, but defs not better than MJ v5 (or even v4 for that matter). Still, it's great that it's free! When do you think they'll release DallE 3 on the website?

Nice video bruv

amazing explanations 👍👍

@MattVidPro

Жыл бұрын

Glad you think so!

this new model is pretty rad ;]

With regards to the 4 generated images of the disgruntled frog detective smoking a cigarette in a bar, I'd say that, although not any _one_ of them were totally perfect by themselves at achieving the whole idea, there _are_ some particular elements from each that could conceivably make for a better image were they to be combined into a new and separate image (IMHO, anyways). Like, if you use the face of the frog from #3 (top right), the hat from #4 (bottom right), with the cigarette and mystery beverage from #2 (bottom left), with the clothing and background scene/setting from #1 (top left). That's a good froggy! 🐸

Thanks for the review. Q: When you specify “8K,” does that give you an 8K res image that would still look good blown up on a 75” TV or large wall poster? Or is it limited to around A4 size?

@seanspencer208

Жыл бұрын

No matter what you put in the prompt the image will be 1024x1024. Putting thinks like “8k” and “highly detailed” generally makes the resulting image look better, even if the resolution is the same

@morpheus2573

Жыл бұрын

@@seanspencer208 Thanks for the clarification. 👍

We are in the ai renaissance so much new content

@therealOXOC

Жыл бұрын

renAIssance

Bing image creator works together with Dall-E, it’s not just Dall-E it combines with it’s own ai magic. Of the three Bing Chat modes, it’s only Creative that will make images for you. but don’t ask it for image prompt ideas because it can’t give you the text because it generates an image. Also you can’t really build on information because it forgets what you said three messages back. Bing images are limited in how many that keep in your history, so save the ones you like. Finally, from my experience, boosts are really needed. Even without them, it doesn’t seem to take much time at all. Great to see the comparisons with other popular image generation tools.

Can I download Bing into my computer to use my own pre-saved images as a base for AI creation?

For me it it was super slow / not generating anything. Still running the first generation that I started at the beginning of your video so I guess in 20+ min it was not able to generate anything. It seems to be stuck on "generating" and nothing can change its mind, even complete refresh of the page and re-query,. I think I broke it with the first prompt When asking bing chat for it, it was also not able to generate in chat but was able to point to the article with the all the relevant links.

It seems like just yesterday that people were fawning over DALL-E's images. Wow, look! Avocado chair! The pace of development is mind blowing.

Amazing video bro. Well, let's wait and see maybe the new DALLE 2 will get some more updates. But overall, Midjourney V5 is still better.

i've been using this Bing Dall E for the last month. the results are good, it can get confused with more than one subject (IE a man in a suit as a rat runs past.) but on the whole it produces more exciting compositions than stable diffusion.

Nice to get something to public, and maybe people realize that there is really advanced telepathic technology. It can take copies of people to metaverse that dont only create images, but can effectively simulate future years forward.

MJ5 is blowing my mind. How are we here only 8 months after Dall-e 2, can you image where we will be after all the images on the internet have been image labelled with GPT4, maybe we will have a "rank-pairs" program, that filters lower quality images, imagine mid journey v5 with control net. I can't wait to see where we will be in another year.

by far the best ai image generator i got my hands on

I was comparing Bing with "Stable Diffusion Online" playground (both free). Really interesting. SDO was more creative, painterly, and had much more variety with each role/attempt (+faster). While Bing could get more realistic and seemed to understand the prompts better (higher res). For more surreal subjects SDO was better I think... But I'll look a bit more.

I kind of like Bing's image generation, but they have no idea what they're doing with their content flagging. I got two of my prompts flagged for nothing, the first one was literally the first prompt I tried: "two gorillas on a street". The second one was "spiralling echoes of the big bang, artistic photo". Am I missing something? These don't seem to me like they would any chance of being flagged, but apparently I have a different idea of offensive or inappropriate content than Microsoft has. I reported both to Microsoft, but I'm afraid to try out more things since they warn about losing my access - not a nice thing to read when you try out their shiny new service 🙄

I think new DALL-E is showing more intelligence than MJv5. It's like MJ has spent more time learning technique and has better gear, but DALL-E has more raw talent.

This was so enjoyable, you have really given me insight to both tools. I can't wait to try them out😃 What are your thoughts on BlueWillow? I have been looking into this new tool. I have also heard quite a number of people talk about it and this has only made me curiious to learn more. Please please share

This will be a fun video

@MattVidPro

Жыл бұрын

def!

You LOOOOOOVE lemon characters 😂

I can't quite get my head around the commercial use of these images.... Do you know?

8:20 Dall-E’s first words💀💀💀

As Midjourney now only have paid version (free version is "busy and not available due to extreme demands" for almost 4 days) this Bing AI creator with new Adobe Firefly are the best for creating images for newbies or anyone just for fun.

crazy how we used to consider Dall-E amazing but now we have so many models to compare it to it just doesn't look as good.

Awesome!

I just tried bing chat and it says it can't create images.

At 18:28 bottom left generation, I find it interesting to see the watermark/signature of the photo. This means the AI looked at a lot of copyrighted photos and learnt to draw the signature. I think this will be a big discussion in the future, because maybe a photographer might get upset to loose his job over an AI that learnt from his photos

@Unbridled-Whimsy

Жыл бұрын

Midjourney generates watermarks too, it's been used as a talking point against it by the anti-ai art crowd. Dall•E 3 doing the same thing will def will add to that discussion

@Samuel-wl4fw

Жыл бұрын

Lose*

Fascinating.

Fun fact: now that when you actually put in a video game character as the prompt it will actually give you what you asked instead of a bootleg Version if it.

Boost or no boost, just keep refreshing the page, usually generates images in 10 seconds

Microsoft is working so hard. I love it.

"cat eating sushi in a cinematic style" hahahahahaha.

Those Microsoft rewards add up quite nicely. I get myself a gift card every now and then just by using edge

The first AI that makes a good picture with this promt: "clean, modern logo for a weather app"

"He's also missing an eye". Well, he is a pirate dog! 🙂

And when people started complaining about Dalle-2 lowering quality, some said it was "personal bias" and there was no change... Sure.

closer than you think

Love it

Bing Points..."what are those!" LOL, I have 175,876 bing points! :-)

@tukanhamen

Жыл бұрын

Use them. Can redeem them for all sorts of stuff and if you have that many you can get something good.

Bing might be dumping the big image archives then. The end of Shutterstock and Getty Images?

@fynnjackson2298

Жыл бұрын

Just my thoughts, drop a Getty image in and say, generate alternatives, book, upscale, book, another whole industry replaced and upgraded.

@revod35ytb37

Жыл бұрын

​@@fynnjackson2298 how can i get it to generate alternatives

The Bing image generator is working in chat and did a much better job with a detailed description than midj

Junk text means its lifting the image from somewhere else. I am favouring adobe atm as they are not ‘stealing’ the basis of their generative art. Shame on you MS! Good video. Thanks

you need to ask for it to generate or make a image not draw or ask it to make an animal it can not do that... next level bro

i love bing now

We need to be able to start with our own image. This is due to copyright policies

if i create a "monkey in a bathtub in front of a castle" for example. i dont understand how the software gets a 3d model from billions of 2d photos. how is that possible?

The end of Stock Photos lol...

I've already had Image Creator before Bing Chat lol

Dal E will rise again Microsoft won't let it fall forever

I'm a huge fan of the new dalle model..... best of both worlds for me.

You have to be in the creative mode in Bing chat

Make a video: How to use DALLE 2 for free?

I don't understand why Dall E2 can't produce better text to images

I don't know.. It doesn't really look like a way beyond to be honest)) I mean they are very comparable

Midjourney V5..hold my beer (yet again!!!)

you have to be in creative mode in the chat for it to make pictures.

Midjourney images are much more professional looking, artistic, especially the one of the tabby cat. The other AI's look like an amateur took the photo or the cat was cropped out of a larger photo, the lighting isn't dynamic. But...I don't want to pay to use Midjourney and be only able to use it inside the Discord server. Currently, I'm trying Leonardo. Which is also free for the most part.

DALL-E 3 basically

Is Bing absolutely free or not

Guess who's getting my 30 dollars this month? Leonardo, playground .... they're nice (and free) but Midjourney keeps the spot at the top.

Midjourney has little control and are gonna get left behind without options that all the others are getting.

I wanna be a GlitchPunk artist

Cool (you all are welcome for me not typing "FIRST COMMENT")

@MattVidPro

Жыл бұрын

thank god

Midjourney is way above any other image generators. It’s not even close

Everyone, you can skip to 4:35 right ahead.

Our world is about to change...fast

I obtained an Open AI API, and had chatGPT build me a web application that uses the DALL-E API to generate images from text. Problem solved.

Google, what are you doing man ? Are you sleeping. Give some competition.

KZreadrs need to stop trying to compare Midjourney to XYZ.

YOU NEED TO BE IN CREATIVE MODE TO DRAW IMAGES

I asked to create a picture of Bowser and Dr. Eggman and Dr. Eggman was way Off because it will take me fishing world, literally instead of thinking of sonic

dude really you need to look at your question litteraly, you are skippin the "create an image of" part

As a German 5:49 is very sus (ai generated muschi)

Ur so rude to dalle 1 generation for no reason😭😭😭

@MattVidPro

Жыл бұрын

haha, sorry

Do you remember when Rocket League was invated by trained bots at the beginning of this year?

Can't even ask it to generate an asian girl. Got blocked that prompt