ALREADY?! Ideogram AI Cleans House - IMO the BEST Image Generator

[@MattVidPro]

Idk WHAT ya’ll are trying to generate but to ME this is pretty unrestricted 😅😅 you guys must be more “creative” than I am… Nevertheless I have decided to change the title and thumbnail to reflect the viewer’s opinions.

[@hipjoeroflmto4764]

They just want to free the nipple😂

[@LouisGedo]

👋 Did you know that now (for a couple of weeks) inside Night Cafe, we can now generate Dall-E 3 images in 3 different aspect ratios? Amazing!! 😲

[@NicVandEmZ]

Your lemon prompt might be better if say Disney Pixar 3D render

[@LouisGedo]

12:30 The fidelity is remarkable

[@LouisGedo]

14:39 😆 🤣 😂 awesome images!

[@DiceDecides]

ok the text capabilites blew my mind, it's perfect almost every time... the fact that it's free is just amazing too

[@IceMetalPunk]

Holy shit, dude. The 1.0 model is *so much better* at text than everything I've tried before, even the previous Ideogram models! My go-to test is to try generating album art for my Synthia Nova AI songwriting/Suno Chirp framework. The title is "Synthia Nova: Concert in Silicon", which *no* models have been able to spell right *ever.* Something about the three vertical lines near each other in the "ili" of "silicon" messes them up. I generated three times with this 1.0 model (so, 12 generations), and it spelled everything perfectly more than 3/4 of the time. I actually got a perfect album art for whenever I compile Synthia's songs into an album 😁 (Well, mostly perfect. It shows the singer holding a mic while also in front of a mic stand... I'll keep trialling.)

[@aleassimarro]

Cant state this enough. This guy brings news to us without putting fcking ads in the middle of the video. Started to hate some ytbers because of ads in the mid of the video because sometimes i cant skip it. But this guy, no ads at all just a smooth and nice experience. Props to you mate, you deserve all success

[@keklord5817]

Mattvid and mattwolfe are both good, both recommended

[@MattVidPro]

Ad integrations will be in some of my videos - but only products tested and personally advocated for by me!!!

[@UniversalManifest]

i dont mind ads add more please

[@Sid_this_side]

​Yeah why people putting ads, I want to skip it!! they shouldn't earn to manage their finances😢 only my time matters....lmao😂

[@brexitgreens]

I do get ads here though. The last one was from Topaz AI.

[@Sparrow9612]

I've used Ideogram many times to create team/faction logos for my Fire Pro Wrestling World roster.

[@Sweettooth1231]

I love how ideogram stole Dalle 3's thunder before it was released when it came to generating text, and now it did the same to Stable Diffusion 3 about prompt coherence lol

[@jerrybobs7378]

I only made it 7 minutes into your video when I went to test it out and after 5 generations purchased a years pro subscription. I also have MJ and Dalle but I think its text generation is the best so far. Thanks for making me spend more money!

[@runvnc208]

Lol.. the funny thing is that Stable Diffusion 3 will likely be about as good and coming out within a month or two. So it's nice that you are supporting Ideogram but you will probably regret buying a year.

[@Enu_Vibe]

I see You are a man that likes to live on the edge lol. Subscriping to Annual plan in this AI area is wild.

[@fast_harmonic_psychedelic]

if you make big decisions based on 7 mintues and 5 generations , dont complain and Karen it up when you get disappointed. do not pass go, do not raise hell, do not collect $200 lol.

[@Airbender131090]

@@runvnc2081) not everyone has 4090 ( yes you will need 24 gm vram to use the best sd 3.0 model ) 2) Ideogram always improving and could do amazing text before sd xl was released. Sd will not catch it text wise. 3) sd 3.0 can actually be coming as soon as 1-2 weeks

[@apache937]

buying a years subscription is not a good idea in the AI space

[@robroufla]

thanks! Note that the free plan is 100 pix per day but only 25 prompts. Dall-E through bing is 50 pics/day now (edit : 15 per day)

[@puja1985]

its still 15 per day for me for dalle 3 through bing

[@Av-uv6xu]

note bing DALLE doesnt give you commercial use notice in tos

[@TheBeatfox]

DALL-E 3 through Bing is actually unlimited. You just get 15 instant prompt generations per day; after that, it introduces a wait time that varies depending on current usage (usually a minute or less, from my recent experience). Occasionally it may prevent generation entirely during periods of especially high traffic, but that's not as common an occurrence as when DALLE-3 first arrived.

[@unhingedninja8565]

bing dalle 3 not refreshing coins for me :(

[@MissMolly3377]

@@TheBeatfoxYes, and it doesn’t even take that long after the 15 are done.

[@gonkdroid8279]

I've been looking for a good AI image with text site for a month. This is perfect!

[@chariots8x230]

I hope ‘ideogram’ also masters consistent characters soon.

[@amykpop1]

That would be huge! MJ is also working on it, heard in the office hours that they wanted to release it with the definitive version for v6.

[@FariaKaisar-yi7js]

Like bro true. They cant do celebrities. Only messi, ronaldo, beyonce, taylor, biden, obama, trump, spongebob, doraemon etc. It cant do all celebs like justin bieber, will smith, mrbeast etc

[@chariots8x230]

@@amykpop1 Yes, I heard about that too.

[@chariots8x230]

@@FariaKaisar-yi7js Maybe they’re worried about liability when it comes to celebrities. There’s a lot of problems that can come from people freely being able to use someone else’s image. Personally, I would stay away from creating images with celebrities in them. I would like to create unique characters, and then have AI consistently depict those characters in every image. That’s the type of character consistency I’m looking for.

[@freemasonry666]

​@@FariaKaisar-yi7jsi think only paid user. With upload example photo

[@youngbroadway]

Take ideogram images and upscale them with krea or magnific etc. MAGIC

[@FusionDeveloper]

Wow, I tested my super long prompt that I used to test Dalle-3 and SDXL and others and it included everything in Ideogram accurately. Some of the small dog faces were a bit messed up (typical for small faces in an image), but it didn't ignore any details, which was shocking to me. Better than Dalle-3 with listening to prompts... this is insane. I wonder how Ideogram will compare to Stable Diffusion 3, I have a feeling Ideogram will be the best for now. If you only want to do a simple image like "a dog on a white background" then you can probably use any AI and get a great image, but it's when you do really specific and really unusual that they usually fail.

[@yoyo-jc5qg]

Personalized birthday cards is a good idea, it makes it seem like u spent lots of time on the card 😀

[@louielouie9502]

My Boy Matt you look like a young mad scientist in your thumbnails. It adds flavor to your posts in a cool way. Your videos have actually gotten better over time also. Thanks for creating on this platform. One question, could we do consistent characters with ideogram ?

[@MattVidPro]

I don’t believe character consistency is a strong suit for it

[@OriBengal]

Got excited by your video, so I signed up for their $8 version, so I can have my experiments be private, but to use Image upload (reference, img2img) you need the $20 version. So, upgraded. Many of the results are pretty amateur. The text as well-- It's cute, gets the spelling right, but is grainy and jagged. I'd say that it's a cute toy- maybe a good starting point as part of a more intricate workflow with other tools.... but not a professional tool on its own.

[@mounirbousli4323]

Thank your comment it helped me to understand better

[@sergio2494]

Results are amateur compared to what? Midjourney/any others? also, hows the reference working so far? is it like the --sref from MJ?

[@DalaluddeenHalala]

​@@sergio2494 Better img 2 img can be achieved in SDXL

[@Sandro_g99]

its the first time i can generate corectly a " blue cat with 4 wings "

[@MemePopperz]

I am honestly surprised by Ideogram 1.0's coherence to prompt understanding and text accuracy! AI is getting so much better in 2024..can't wait for Sora and SD3 release

[@dahozabich]

Matt, every time I see the DSOTM logo I have to acknowledge it. And you earned a subscription for it also. This ideogram thing looks AWESOME!

[@itsallgoodaversa]

For the posters of the slap movie, it looked like some of them were of Chris Rock, who is the person Will Smith slapped

[@alexgreene9399]

I saw that too 😂😂😂😂😂😂😂 I was like that’s Pookie!

[@calumscott.1442]

It’s probably linked the two names due to the news story about them recently

[@AtrumVox-DE]

Ideogram's precision in handling complex prompts is notable. Curious to see how it evolves.

[@johnny1tap]

If you think about it it actually got the prompt kinda wrong with the cube. It's "A blue cube wrapped as a christmas present" which means it should be open to interpretation what color the actual wrapping is, since the blue cube underneath can't be seen.

[@grahamatkinson9851]

Thanks for a nother great vid Matt. Finding that Ideogram's prompt coherence is very good, but I'm finding text accuracy only moderate, and feel it lacks the photorealism we have become accustomed to. But can see this being great for posters, art, t-shirt design etc.

[@NYCsantiagoMusic]

So helpful. Ideogram was collecting a little dust as of late. But this... this is insane.

[@AnthonyCook78]

Ive tested it. Generated people sometimes look melted. Its like Midjourney V2 but with nicer hands.

[@GabrieloTomorrow]

Now I know which AI Image Generator I will use for my new company. This is insane!

[@Av-uv6xu]

first try, than cry

[@GabrieloTomorrow]

@xu What do you mean by that?

[@nickgirdwood3082]

I tried it and it's better than it used to be, getting clothes and hair colour right, but it made some derpy 3D models and paintings of my characters. Also, unrelated, the Grammarly that's built into my keyboard is locked to American English even though my account is set to Canadian English.

[@Russman]

Finally something not restricted!

[@BoOgieAndMelodysVarietyShow]

Oh. This should be exciting because I've been using Leonardo. Aloha, Matt 🌴😃🤙

[@_Mach3]

@15:52 is that a bean in Blackface? Lol

[@JamesWebb-f2r]

this ai can do fight scenes well not many others have been able to do a fight scene

[@aidanblah9646]

Hey Matt, love the channel. After watching this video, I though I’d give it a go with a prompt that no image gen has ever gotten right. Ideogram fail gloriously as well. Oh well. 😢 This was the prompt. a guy on the left aiming a bow and arrow at a guitar up against the wall painted like a red and white target, in a large loft with wood floors and brick walls.

[@XNY_Music]

A man armed with a bow and arrows, is practicing archery in his studio loft apartment. He has hung a guitar on his wall and is using it for target practice. The man is standing to the left of the image and is firing arrows at the guitar, which is on the right of the image. His studio loft apartment has wooden floors and the walls are brick. Stylish image, influenced by music videos, masculine, archery, cinematic, atmospheric distortions. This might get you close, but obviously results will vary depending on the AI generator you are using.

[@fast_harmonic_psychedelic]

stop using commas, commas do no one any favors. everything ive ever seen has done better just doing one long run on sentence with no punctuation

[@sususim76]

@@fast_harmonic_psychedelic that depends...it's not a definitive

[@fast_harmonic_psychedelic]

well ive seen its true through my thousands of genrations @@sususim76

[@fast_harmonic_psychedelic]

A man armed with a bow and arrows practicing archery in his studio loft apartment he has hung a guitar on his wall and is using it for target practice the man is standing to the left of the image and is firing arrows at the guitar which is on the right of the image his studio loft apartment has wooden floors and the walls are brick stylish image influenced by music videos masculine archery cinematic atmospheric distortions@@XNY_Music

[@Politicalhomeless]

If you add Dalle 3’s illustration ability with midjourneys realistic photograph ability, and this text generator ability and you got a really rock solid image generator. Too bad they can’t work together to generate the ultimate image haha. That would be one sick ai team. Adobe’s firefly is honestly only worth using inside photoshop to edit images and I think is the best at it, so if you add in paint ability you can throw firefly on the team. Add the uncensored feature using Medusa to unlock a world of morbid funny possibilities. But no this isn’t the best image generator but it is the best least censored one that I’ve personally used so far. We all know that could change by the time I’m finished writing this comment.

[@mycloudvip]

Great VIDEO and awesome updates for today! I am doing my share of letting colleagues and friends know about your channel! KUDOS!

[@undergroundo]

VIDEO IDEA: Lemon images progression, from the first video to the last.

[@frazy4487]

cant wait for the next banger vid where we compare all 4 models including SD3

[@42ndMoose]

havent they trained ideo since release? im surprised it's still in its first itiration.

[@DaleTrevors]

Thank you Matt for showcasing this new amazing tool. I'm going to use it.

[@inmyfreetyme]

I just tried Ideogram out and love it so far. Thanks 4 the heads up.😎😎

[@FariaKaisar-yi7js]

Ideogram is my fav. Midjourney is bad 😡 midjourney used to give free 40 prompts but now its over. Midjourney is 100% paid now, no free trial. Midjourney has 17 million users but they still dont give any free trial, how evil! While ideogram has few users like 100k but they still give 25 free prompts everyday. Ideogram is the best

[@thatsthewaythecookiecrumbles]

Matt - great info - i have a paid Ideogram subscription - I am attempting to use uploaded images of my products and incorporate them in the image. Example - box of latex gloves which is my brand and upload and then prompts of let's say and extremely happy dentist holding the box of gloves etc. just not working. What is the proper method to get the most out of an uploaded image? Thanks.

[@carlkim2577]

Thanks, I've been waiting for this. Good accurate text generation! This is the first actually useful AI image generator.

[@allenmitchell3596]

Good video Matt, good job!

[@BlackMita]

“A giant can of Sanpellegrino soda crushing an elephant into the mud.” checks out.

[@fabiano8888]

Great video Matt!

[@AnnaRosaEverythingEnergy]

The graphics are great. How to you get the text to work properly? How do you designate text vs not?

[@mopolo]

Just goes to show how nothing is set in stone in the AI space and how you can still innovate and make a name for yourself

[@sybervisions]

Looks like an interesting new tool. I'll give it a shot on my next project! 😃😃

[@Skeeva007]

Man the new and improved AI is getting insane! At this rate the world is going to be unrecognizable in 5 years! Great video as always 🤘

[@ronilevarez901]

I assure you, with a 100% confidence, that they world will be exactly the same in 5 or even 10 years. Yes, maybe there will be some robots walking around the streets, and less jobs, but in general the world will be the same. Only the richer will experience fragments of the new lifestyle first hand. The rest, wich is most of the humanity, will keep living as we have for the last 10 thousand years. Wake up. Go out to get food. Come back home. Spend time with the family. Have fun in whatever way you like. Sleep. Repeat . Or you think that changing the device in wich we watch TV or how programs are made are world-changing events? Nah. Until we enter the matrix-like, metaverse age, the world will still be the same. With different make up, but the same after all.

[@brexitgreens]

The current AI train took off in 2020. The world is still sadly very familiar.

[@nebojsamatusijevic7886]

How can I import my JPG image into Ideogram and make my image better? For example, the same image in oil on canvas and after download in better resolution and quality. Thanks for the reply .

[@SardiPax]

Interesting developments. As an aside, when you say 'coherence' I think you generally mean 'adherence'. They have a somewhat different meaning.

[@amj2048]

I kept using Ideogram and then stopped because Dall-E 3 came out, so I missed when this change happened, but wow, it's amazing now.

[@tke-nationupsilonpi9544]

Can you make a video on how to get consistent characters and which platform is the best for it ? Thx

[@WhatLurksBeneath]

Dude... the text generations 😀

[@DWSP101]

Those glasses make them look like you slowly transforming into a cyberpunk hippie

[@TroyCollier]

I got lost with midjourney 6 the prompting got confusing. But this was so simple and pumped out some of the best images I've ever seen. Plus your Trump prompts are sweet!

[@Revex08]

Hey Matt! Been watching a while, this image generator is good, the text is very impressive. But the images have a lot of flaws that I don't believe you will have missed so maybe glossed over? Eyes for a lot of animals and people were weird again. The hands and legs sometimes had multiple like when using S.D. the celeb references were off as we saw. The "Obama" sign was misspelt in the image you said could cause issues. The quality of midjourney and dalle still looked aesthetically very nice compared to some of these ones. Sometimes it feels like you give these new generators or models too easy of a time and don't call them out enough on some of their flaws. All in all great video again though and looking forward to the next couple of months!

[@iammaxwellcole]

I agree, the new model is insane! I also appreciate their price points and rendering speeds so far. Excellent. I like the way they are laid out and the options they provide. Great post as always! Also, Dall-e-3 and the stupid token system despite being awesome at generation is what made me go pro with Ideogram. I want to be able to make as many images as I want without all of the caps and waiting forever due to throttling as a paid member.

[@koen.mortier_fitchen]

I like it a lot. Great for text styles.

[@ventonthorn3455]

That rooster is really creepy. 😳

[@brexitgreens]

Straight out of vegetarian propaganda.

[@samitsinha]

You are always so full of surprises,

[@guzu672]

Use this prompt if you want to break any new model: View from below of a circle of friends smiling.

[@jvoidhuey3088]

Can't sign up to Ideogram as when I hit the Continue with Google button, a pop up comes up for a half a second then disappears. I turned off all pop up blocker and ad removal extensions, reloaded the page, and it does the same thing no matter what.

[@WayOfTheZombie]

Restart your computer. Update your browser

[@fast_harmonic_psychedelic]

what happens when you open the developer console with control+shift+i and look in the console errors? Do you have some experimental flags or settings enabled? did you try clearing the cache and deleting cookies? Are your Service workers on and not being bypassed? is a key on your keyboard pressing itself and closing out the window prematurely? Do you have cookie editor extension which can sometimes cause certain interfaces to fail and return exceptions?

[@caramell5841]

This might be weird to say but I feel like it has more detail and better lighting than Dalle-3 does. Imo ideogram doesn't do image generation and prompt following as well(in certain styles) as Dalle-3. However it seems to show more emotion, more correct, and look more....complete? I guess?

[@fuzzydragon]

Is it less sensored into Oblivion than Dalle 3?

[@bladechild2449]

the 1.0 is censored in the sense that it obviously just isn't trained on anything copyrighted whatsover. It doesn't understand any artist style or any real characters.

[@DUDEiTHINKIMNUDE]

How can you use this text prompter for Tshirt designs!? Thank You!

[@FRareDom]

damn this was unexpected, ig this should be expected tho knowing how fast ai has developed over the past year

[@AkashVan-fq9mt]

please make a video on ideogram where we can upload images of product and the AI of ideogram creates a image of a model wearing our products

[@MODDINGPRIME]

Thanks so much Matt this is truly amazing

[@GamersBlogX]

I've been messing with it, and while the understanding is mostly amazing the quality of the images themselves is a little... mixed. Some of them end up looking great, while others look like something I could make on SD 1.5 without a negative prompt.

[@GaryJr530]

Dont forget about the image editor!

[@vainezaiven6677]

From what I've experienced, DALL-E has almost no comprehension of right or left.

[@brexitgreens]

Because they wanted to make it centrist.

[@vainezaiven6677]

@@brexitgreens I gotta give you a thumbs-up for that one.

[@brexitgreens]

@@vainezaiven6677 😁

[@AmandaFessler]

Still needs some work, but DAMN when it gets my logo prompts right, it goes HARD. Now I just wish it could be converted into vectors, so I can edit and scale it in Adobe Illustrator. Heck, some kind of inpainting would be nice. THEN I'll convert that to vector. I basically just ask GPT4 to come up witth a prompt for an idea I have. Then I rewrite it to fit my ideas better before pasting it in Ideogram. It's crazy.

[@PseudoProphet]

And this is just 1.0 Their standards are too high. 🤯💥🧠

[@The_Questionaut]

Matt, what do you imagine we will have by the end of the year?

[@amkire65]

Any idea how to delete images generated on Ideogram... I get the occasional scary bad one that I just want to banish forever but don't know how to.

[@sususim76]

You have to pay for an account to delete

[@Geen-jv6ck]

It appears to be a very capable model. It does text better than any image generator I've used before. You could make memes with it.

[@smaugthewyrm]

Please demonstrate the image upload feature of the paid subscription plan and its advantages / uses.

[@michaelandremovies]

Wow, Ideogram is and has been for a while my goto image generator!

[@thewebstylist]

Amazing! Just made some $ick logos for my new brand! 🎉

[@djnoj3371]

So I just tried this new version, it's definitely really good, but my prompts just mess up these image generators, even this version. Thanks for the info

[@fast_harmonic_psychedelic]

you just need to get better at prompting. avoid commas, spell accurately, ask bing or chatgpt to help improve the prompt

[@GaryJr530]

In 100 years i feel sorry for the people looking back at all these images, not knowing that they're fake

[@CHARIOTangler]

If it ain't opensource, it ain't the king.

[@nindu6913]

even still for free this stuff is insane

[@DreaMagnifier]

Is that any chance ideogram is using sd3 ??

[@sususim76]

No it's their own model they built

[@piteshbhanushali1140]

Where sd 3 release date ?

[@sajinscartoonvine]

Why do you create much longer video....

[@kuromiLayfe]

finally a announcement with a handson release which lets you actually test their examples.. makes me wonder if they are using early access of the SD3 model 🤔

[@sususim76]

nope. It's proprietary.

[@dwainmorris7854]

This is perect for making Memes and no censorship almost perfect for a free price , now they need to add animation

[@okolenmi7511]

It's nice, but I will wait for Stable Diffusion 3.

[@cacogenicist]

It has some odd holes in it's training data. For example, it simply can't generate an image of Gollum -- unlike Dall-E-3

[@IlRincreTeam]

I was waiting for an update of theirs for a while, v0.2 was a little bummer compared to v0.1, I'll try v1.0 immediately

[@TheSopk]

I'm so SHOCKED and STUNNED ! it's UNBELIEVALE, BREAKTHTAKING, AMAZING

[@MrDutch1968]

I take it you are also MIND BLOWN?

[@BusinessButlers]

I enjoy Ideogram but heck they still really cannot do hands. I asked for a pic of a lady holding a jam jar that was a terrarium. Her left hand had 5 fingers and one thumb. Her right hand had 3 fingers and one odd looking thumb. It WILL do text accurately eventually but heck, it is forever putting extra letters or leaving other letters out and you do one run, choose the best and then all 4 new ones are $h1te. It will however yield you some great images if you persist and have enough prompts and maybe I do some fairly complex typography. It can only get better but in the meantime, PS helps to make things perfect.

[@TwinTailsRyu]

dud those steve jobs posters were sick

[@theconcreteshamans]

UT is truly amazing

[@jaredgreen2363]

6:12 hey at least they are drizzled with ketchup…

[@GOATEDFOOTBALLEREDITZ]

Stopping kids from saying "stopping kids from saying first"

[@tmendoza6]

What is going on with Quantum computing and AI. it should be interesting

[@Xtremevibes-nd7gm]

suno V3 is unbelievably know all music style especially the instrumental options. tried math rock, funk, trance and vaporwave. it will really replace the musician virtuosos out there but ofcors it cannot replace the actual people’s talent (for now)

[@fast_harmonic_psychedelic]

i bet it cant do Hitech Psytrance or Dark Psychedelic Trance you gave it easy styles

[@lylek8933]

DeepDreamGenerator's AIVision is pretty good. :)

[@Yipper64]

my yoga zoroark test has terrible results sad to day. Bodies look contorted and broken, and it doesnt look anything like zoroark. My melting spongebob comic test is rather interesting, it not only works but incorporates other spongebob characters without prompting. (likely due to the magic prompt) Of course there's not very much consistency but it sometimes kind of works. And of course it sometimes does not generate a comic at all. it is crazy good at text though, when not trying to make a comic

[@apatsa_basiteni]

Not that bad - good the image generators now have microdetails like, prompt understanding, hands and text as their next area of improvement.