Looks good, but I agree, it's no Midjourney "killer" for now. And I 100% agree. I see dozens of uploads like "Midjourney killer?" or "ChatGPT killer is here!" Annoying, I wish content creators would have the same integrity you have. You've done masterfully to reach 100k subs with 0 click bait tactics.
Ah, thanks! I mean, I've done some "Everything Just Changed" titles in the past, but these days I kind of keep that in my pocket for when I see something that REALLY blows me away! But yeah, the whole this vs that thing is so played out. I keep saying, the last thing I want is a "One to Rule Them All" situation, because that's when everything looks exactly the same. Bland.
I just wanna say I love your videos. Always to the point, good use of humor here and there, good chapters structure, fresh info on AI Tools. You don't seem to have any bias and you give honest and experienced insights. I also love your plain wood guitar on the (viewer's) left. I have a very similar-looking custom Telecaster. Finally remembered to subscribe despite the fact that my feed was doing a good job of showing me your vids anyway.
Excellent to hear! Thank you so much! And yeah, I’m a sucker for natural wood guitars! I think it’s such a classic style that never looks dated. Funny that you mention a tele! That’s been on my wishlist for awhile now. I saw there were some knockoffs of George Harrison’s Rosewood Tele (dark wood) floating around, and I might have to pick that up at some point. Can’t afford the real thing (and the real/real thing I could NEVER afford) but it might be fun to take a knock off and mod it up to a playable guitar! Natural Wood forever! (Although I also still want a green meany at some point. I have this theory that there’s always something a little special about a green guitar!)
@@TheoreticallyMedia definitely! Green (and Mint Green especially) is another great option. I am generally a sucker for vintage pastel color looks, especially on a Tele. I ordered mine from Haar. He is a Dutch Custom Guitars manufacturer. After digging around quite a bit I saw many great things about his guitars and he is definitely more affordable than buying a custom one from Fender. Got mine for around 1.2k euro if I’m not mistaken. It’s been a lot of years though so I don’t remember! We didn’t even use an e-shop for it back then. I ordered it via email and sent him lots of pics of how I wanted it to look like and he sent me photos of the manufacturing process and I paid him in 2 installments (50% upfront 50% when the guitar was ready to ship) via bank transfer. Thinking about it now…. Well let’s call it “Living life dangerously…” 😂
@@TheoreticallyMedia by the way I just saw a fender partscaster with George Harrison’s rooftop rosewood look and it looks amazing indeed. Plus I love the Beatles so it gets extra points.
Yeah, that's the downside of being so small, and likely without a ton of fincancal backing-- compute is going to cost a lot, and they don't have the deep pockets to offset it-- I presume at least. Still, if they can get a good enough user base going, I'm sure someone will come along and throw some money at them, and then the cost can come down.
@@TheoreticallyMedia Hotshot was opensourced before this model. Hopefully they release an open version maybe not the full one thats very understandable
@@TheoreticallyMedia lol I bought a $95 subscription for Runway ai because after tinkering around with it with the free credits, I ran out and wanted to make more video clips and use Lip Synch. I'm in the process of creating my first ai short film with a script by ChatGPT. Dialogue heavy of sorts. A short film tests to see what can be done and I'm loving it!
Matt Wolfe done some live prompts. Everyone was terrible. I've been using luma and it's never produced bad video after bad video... it's usually an even split of 50/50.
As far as the 5 sec or 10 sec question, I am satisfied with 5 sec and the ability to extend an additional 5 sec if desired. Most shorter videos like commercials, music videos, trailers, etc.. don't go past 3 second clips which is what I use it for. But if you plan on making movies then I understand 10 sec.
Yeah, I think now that we've gotten into the era of being able to lip sync and provide "acting" to our AI characters, the need to 10 second shots (and possibly longer) is going to be more of a thing. That said, most films and TV shows still won't hold a 10 second shot-- if it's a dialogue sequence, you'll often cut to a reaction shot. So...maybe 10 seconds is the sweet spot.
Good to have you back Tim. 👍 Completely agree with x is the new y killer case issue. Different use cases, different workflows, more options...Look at how many men in blue suits we have on the streets! With FreePik and Maginific together, I really cannot see how they could possibly mess up photorealistic images at least. Considering Magnific is pretty expensive by itself, $99 may seem on the steeper side, but we'd be getting two very capable platforms. Stock photos are going to take an even bigger hit. That's bittersweet, but technology beats humans. Speaking of steep, you should end your videos with LTX studio's pricing. 🤯 10-second joke - can't believe you 'didn't' go there. 🤣
Freepik are really working on some amazing tools. I'm primarily working on pitch documents, treatments and concept art and it's been amazingly useful. Their existing upscaler is amazing, with magnific it's only gonna get better.
I actually feel bad about not checking in on Freepik. Since posting this video, a lot of comments about how many people love it. Sounds like it might be the best kept secret in AI!
The almost weekly introduction of multiple AI platforms is just insane, freepik is pretty good. You get the 2 magnific upscaled so I am sure they are busy integrating it. With so many new AI platforms appearing weekly what do you think about doing a video on the remains of the failed platforms and why they might have failed? On the clip time anything over 4 seconds works for me. If you can extend even better.
I'd love to do a "whatever happened to" type thing-- but if there's one thing I've learned: Don't count anyone out. I've seen a few platforms that were proclaimed dead, only to see them come roaring back with one update. Rumor has it, we'll all be talking about Pika again in a few months. They aren't dead...they're cooking.
13:13 👍 agreed, I do praise and intellectually support people developing AI tools for creative people in spite of the fact that my bar for usage and financial support is relatively high.
Thanks Tim, good to have you back... seems like the new arms race is the somewhat 'tricky' price tags for these things... $99 per month... ouch! It's like the current LTX Studio prices... to get out of the 'personal use only' tier, you need to stump up $175 per month. It seems that the AI world is going to be taking a leaf out of the cable TV subscription model where you end up with a dozen subscriptions to a variety of services, end up watching/using only a couple and lamenting the cost but never quite capable of cutting the cord for good. The future will needlessly be a two tier society of those who can afford generative AI tools and those who will be mercilessly trolled by them. What a wonderful world.
Haha, I legit thought about training up a clone avatar-- BUT, I'd still have to type in all the news, and I'm pretty sure the family would have gotten mad at me for hanging at the beach with my laptop! haha
Mystic looks really promising. I love the simple interface. ($12/Month for the annual plan - includes all of their various tools. I'm glad other companies are finally competing against Adobe.)
Its bizarrely the same as freepik. I'm quite shocked a lit more ppl havent got onto it. $8 a month ($100 a year) . You get flux, magnific ... 1 million tokens per year. I'd be lost without it. I sound like a bot who's promoting it but it is sweet and offers something a bit different to MJ.
Saaaaame. I think most AI generators are getting better with UI-- and hey, we aren't on Discord anymore-- but those early versions of Leonardo? Man-- that was like getting whapped in the face with sliders and options. (To note: They've gotten a LOT better about it!)
@@TheoreticallyMedia looks like my comment got removed. Curse u YT Yes it's incredible. I'll let u into a secret I signed up blk friday and got it 50% off 🤣.
For the next 12 months, we will see now weekly a new stunning image generator, and video generator. every week again. Every week mind blowing again. Every week groundbreaking again. Every week next level again.
For sure we're in a growth period. It ebbs and flows. With Flux being open source, we're going to be seeing A LOT of image generators pop up for sure. And I'm cool with that, as long as they all bring something unique to the table. Not just: Look, it's Midjourney. So tired of that...so tired.
Oh it was the best! NGL, around the last day I was itching to get back! Apparently I get itchy to jump back in after 6 days! That first day was AWESOME though! haha, glad to be back!!
Since AI has hit us, I've been able to express a lot in music and video combined, it has been a great source of exploration. A lot of people I send my stuff to just don't get it, but there are those that really do, it's another tool really, to use, just like a brush and paint, if you're creative you will make creative stuff just like any other medium. It really isn't like you can just push a button and out comes a hit music video, you still have to mix it, find the right sites to animate to music etc. I use SUNO as it really is exceptional at getting you're vibe if you input correctly it almost seems to understand what you are after and it becomes a very useful tool. Thanks for keeping us all informed, I take your videos seriously as I know you are right up there with the latest AI. I must admit I was sceptical of AI at first like a lot of people, but now see it as a useful creative tool.
Something like that for sure- but maybe without the upscale? I still think the upscale side sprinkles some magic on it as well. It’s certainly interesting. And I do love the idea of being able to choose Loras for fine tune control. It looks promising!
It's in Alpha-- so only a handful of people have access to it right now. That's why I was kind of posing this as a "preview"-- it's coming soon though!
I know it's not your usual area but I'd love to see some alternate non-photography tests on some of these platforms. You mentioned it but having a few more illustrative generations in your back pocket like our favorite business clad man or pirate lady would be helpful. I've found that new image generators like Flux are focusing heavily on photorealistic images but have a lot less depth than Midjourney. So while I can get close to Midjourney on Flux for a photo, as soon as I try, say, a fantasy illustration, it quickly falls waaaay behind what Midjourney can do. I think having a small handful of alternate styles to throw in your tests could show some differences people are missing between these image generators. Anyway, love the video as usual! P.S. Also really love your point about different Gens doing different things. I'd love if they specialized even more. Ideogram v2 is out today and it does the text Gen/Design stuff soo much better than anyone else.
Yup! I was playing around with Ideogram today-- it's still the king of Text and Design! As far as more illustrative stuff: I've got one coming up pretty soon-- Great note as well, I'll try to put more tests and examples of that into future videos!
Currently only 2 generations per day, so sadly it might take you a bit to make a full project! But, for sure- check it out and let me know what you think! And go weird!’ Weird rules!!!
It's always wonderful to see new AI tools! Think about looking at alternative solutions that provide special benefits for your artistic endeavors. #AIImageGenerator #Technology
question: i use paid version of freepik but i can´t see MYSTIC as a model. i have 3 times different FLUX and two time CLASSIC. where do you find MYSTIC?
So, it's currently in Alpha-- meaning that only a handful of people (I was lucky) have access to it as we're testing and stressing it out. I think it should be opening up to a wider release soon. So, I mentioned in the video, think of this as a preview of something coming soon. I know, I know-- I hate doing these videos where I'm like: "Look at this thing you don't have"-- but, it IS coming-- and I really loved the look, so I wanted to share it with you all!
@@TheoreticallyMedia no problem at all. i am a viewer from germany and i like your videos. i did not realize that this was ALPHA, sorry. and as you certainly noticed, i also liked the comments here which stated that you are not doing videos with untrue spoilers. you are akind of "real" and not hyping everything that is new in AI. I really enjoy your videos . And to be honest, i'm also a little envious of her music skills... ;-)
Freepik, is one of the stars of AI image generators I always use in my production. But lately, Flex module starts to enforce its natterator to the prompt. Especially on color grading. Using some words like mediaval. The only AI can give details to background. In contrast, Leonardo AI, always fail to give details in characters in background, it only provide render to foreground character and the background characters are murfed.
Is that with Leo’s new Phoenix model? To be honest, I haven’t played around with it too much, but in their defense, it’s a brand new model that (I believe) they built from scratch- so, it’s a bit like a v.1. So, I can see that with the background characters. For now at least- I’m sure it’ll get better as time goes on and they continue to improve. On the Flux color grade side- that’s where those Loras are going to come into play! As stunning as Flux is, it’s still a baby model. As the community starts building workflows, it’s going to grow into a real powerhouse! It’s funny, we took this big step forward lately, but that also required taking a half step back’
I mean, I hate to say Topaz, because it costs so much...but, Topaz. I'm actually surprised no one else has picked up Creative Video Upscaling like Krea has. It seems like such a slam dunk idea, but to my knowledge, they're the only ones doing it!
I think an extend feature and a mini prompt to direct the next 5-10 seconds of action is the most useful thing any text2video ai can add for users. I am interested in making full length (90 minute) movies and therefore I need to be able to make videos up to 3-4 minutes. in moonvalley I had to generate hundreds of clips and use a linear blur to match up similar scenes, very labour intensive but it has it's own unique art style I guess.
There's Krea which is fremium And Leornado AI universal upscale fremium. Clarity upscaler on their website and replicate which freemium. And tones of upscaler models on huggingface and replicate to try
not from the author: i am very satisfied with TOPAZ PHOTO AI since over 1 year. this must be installed locally but you have full control and they have updates at least all 4 weeks.
Awesome video as always. Can someone help me figure out if after the freepik and magnific merge, the actual upscale on freepik is the same as on magnific? After testing it, it feels kinda inferior to magnific, but I might be mistaken.
2:45 okay, I'm guessing that these places paid models to pose and paid a flat usage fee in perpetuity so that the AI could use that to create these photorealistic images just like people are paying big money for voice actors to read like 10,000 words and phrases so that they can have one more voice model to draw from in creating a voice over from text. cuz 2:45 does not look like a 3D model or anything generated by a computer to me. It looks like a literal photograph of that guy but greenscreened with a blurred background. How soon before people begin recognizing particular faces and images from these AI programs? Just curious.
Good call. Very promising model for sure. I’m sure someone on the open source side will be working on reverse engineering it by the end of the week as well!
They keep increasing the quality of these things without adressing the most pressing issues. I can make great images with stable diffusion right now. The diffrence between that and "the best generator there is" wouldnt be huge. Usually the better models just have better prompt understanding. What i would really like to see is a generator that understands the passing of time and can portray that in a series of pictures that are in chronological and or logical order. Basically the ability to create a visual story from a written one. And if that is too much to ask for, give me a manual one where i can just promt for picture 1 to 10 myself and the ai then generates me 10 images based on my prompts that folle the outlines story or progressen. Bit of a rant right thereXD
For the most part all these Midjourney and Sora killers are superfluous. I’m finding that all I need is Midjourney and RunwayML Hedra if I need some lip syncing. The rest Dream Machine, and Kling aren’t necessary. The next model that can “kill” would be a video model that can do dynamic action coherently.
Man, dynamic action is such a holy grail right now. I think we’ll see it soon via Video to Video referencing. Kind of like Viggle, but one step up. We’ll likely see it this year.
10 seconds is... somewhat useful. If consistency was more consistent, I would not care as long as we could extend to 30/60/90 with multiple generations from the first prompt. But mine generally tend to look great for those first handful of seconds then warp into some catastrophe or completely different scene
I'd really like to see AI reviewers test consistent characters capabilities as well as if the AI can generate multiple distinct people if you describe them independently. Instead of "3 guys standing in the middle of a busy street", describe each guy and see what it does.
I'm hoping it gets to the point of just a simple single downloadable app to install on the pc. Not 1111 or anything like that but for people who don't want to install a million things. Guess in time :)
Someone will crack it at some point. Right now, there are options, but like you said: There's a lot of hoops to jump through to get it working. I had played around with doing a Comfy tutorial-- but, man, that ramps up so quick. Plus, I know those types of videos just turn into people asking me for tech support in the comments! haha...I do enough of that in real life! Speaking of which, I have to fix my neighbor's bluetooth connection to their printer. Sigh. Sometimes it sucks knowing how things work.
10 seconds is generally acceptable but is still a it limiting if you want to make a proper film where characters are going to have any amount of dialog without having to try and look the video in an inconspicuous way. 2 minute long continuous shots are only necessary in certain scenarios but 20 or 30 seconds would open a lot of doors and 5 isn't enough for anything other than trailers, especially when you need to edit around weirdness so maybe only 2 or 3 of those seconds are usable.
"Built by four people" It just goes on to show that people who are really dedicated and are into it are far more capable of producing amazing stuff compared to all these 'companies' trying to capitalize on the AI hype with sub-par workforce of dozens or more lol of course there's some internal politics and investment issues related to it, but whatevs. Thanks as always for the latest update :) Also considering whether 10 secs is enough, could be the sweetspot because all these generations need to be edited anyways to make the video more immersive and engaging to look at (while also hiding AI flaws as best as possible)
totally agree on that smaller team front. Quick and nimble for sure! On the 10 second front: I had a bit that I cut from the video talking about doing a one'er, like the attack in Children Of Men-- the thing is, I don't trust any current model to maintain the level of consistency needed for a shot that long. For sure someone in that scene would sprout an extra head...although, that might be cool. But a different movie for sure!
Light mode in the streets, Dark Mode in the...haha...yeah, I know-- people always yell at me for the actual need for Light Mode-- but still, it's like staring into the sun for 8 hours! Too much!
Honestly Dark mode is the best UI invention and I believe Apple iPhones were among the first to make this popular. I never can use light mode on anything anymore. PERIOD. 😂
that hotshot thing is not sora quality, its pretty wonky and early. the gens you see might be cherry picked from 20 -30 gens I tried it. though sometimes when it works. it gives better quality then the others. (assuming humans wont have 3 arms.
Haha, I'm telling you, once the whole 3-arm thing is totally stamped out of AI Video, I think I'm going to miss it a bit. There have been so many gens that went wonky that really made me laugh. Good to know about Hotshot...Yeah, I think the cooler story in the whole release is that it was built by 4 people. Let's see what they can do in the next 13 months!
Flux is still version 1 and it is killing it you can use replicate to use it easily, it is almost free with its pricing But as you mentioned, it is great because we options now, you just summed it up Great video thank you
I got really confused with this vid. Im sat there watching this new mystic website and im thinking 'whoa this looks exactly the same as freepik'. Musta missed that bit 😂. Im always shocked why freepik never gets a shoutout . $90 a year while MJ is $30 a month. You get 1 million tolkens now. I'd be lost without it
Freepik helpfully deleted the images I created before and apparently their ability to create images using a doodling space is no longer accessible... Downgrade as far as I'm concerned.
Personally I'm fed up with Leonardo because of the censorship and limitations of options now. They started off great and sank like a stone . Lately I'm using CGdream and it's much better.
Hey bro, I just released a free and open source app to make interactive videos and films. Would love to see you make an interactive Ai short like Bandersnatch.
If anything, I guess it’s a compliment to MJ, as it is sort of the de facto king of the hill with the most name recognition. Still, at this point, I think it’s just lazy content. We gotta start branching out and finding new angles. At least, that’s what I’m doing on this channel.
the video model that was built by 4 people was reviewed by Matt Wolf, and it SUCK, not only suck, SUCK so bad, the demos they had is completely can never be replicated.
Well, like I said in the video-- the real story to me is that it was built by 4 people. And yeah, it might suck today-- but let's give them some time to scale up and develop it out a bit. I'm NOT suggestion to pay for it...but let's give them a round of clapping for releasing and see where they take this next.
@@TheoreticallyMedia I have no problem with that, but the issue is that they are scamming people into thinking they can generate like the demo and charging $100 for totally useless generations, that's a bad bad sign to startup your company with..
Eh, like I said: It's like day 2. Let's give them a minute to find their footing. The coolest part of the story to me is the fact that 4 guys pulled this off. That's the super cool part. Fingers crossed they'll improve and maybe even surpass....well, I was going to say Sora-- but, y'know...
Yeah, same UI. I suspect it’ll be ported over to Magnific as well (I mean, why wouldn’t it? They own it) and maybe we’ll see some UI changes there. But then again, if it ain’t broke, why fix it?
Good Info
Looks good, but I agree, it's no Midjourney "killer" for now. And I 100% agree. I see dozens of uploads like "Midjourney killer?" or "ChatGPT killer is here!" Annoying, I wish content creators would have the same integrity you have. You've done masterfully to reach 100k subs with 0 click bait tactics.
Ah, thanks! I mean, I've done some "Everything Just Changed" titles in the past, but these days I kind of keep that in my pocket for when I see something that REALLY blows me away!
But yeah, the whole this vs that thing is so played out. I keep saying, the last thing I want is a "One to Rule Them All" situation, because that's when everything looks exactly the same. Bland.
Agreed! Great comment
I just wanna say I love your videos.
Always to the point, good use of humor here and there, good chapters structure, fresh info on AI Tools.
You don't seem to have any bias and you give honest and experienced insights.
I also love your plain wood guitar on the (viewer's) left. I have a very similar-looking custom Telecaster.
Finally remembered to subscribe despite the fact that my feed was doing a good job of showing me your vids anyway.
Excellent to hear! Thank you so much! And yeah, I’m a sucker for natural wood guitars! I think it’s such a classic style that never looks dated. Funny that you mention a tele! That’s been on my wishlist for awhile now. I saw there were some knockoffs of George Harrison’s Rosewood Tele (dark wood) floating around, and I might have to pick that up at some point. Can’t afford the real thing (and the real/real thing I could NEVER afford) but it might be fun to take a knock off and mod it up to a playable guitar!
Natural Wood forever!
(Although I also still want a green meany at some point. I have this theory that there’s always something a little special about a green guitar!)
@@TheoreticallyMedia definitely! Green (and Mint Green especially) is another great option. I am generally a sucker for vintage pastel color looks, especially on a Tele.
I ordered mine from Haar. He is a Dutch Custom Guitars manufacturer. After digging around quite a bit I saw many great things about his guitars and he is definitely more affordable than buying a custom one from Fender. Got mine for around 1.2k euro if I’m not mistaken. It’s been a lot of years though so I don’t remember! We didn’t even use an e-shop for it back then. I ordered it via email and sent him lots of pics of how I wanted it to look like and he sent me photos of the manufacturing process and I paid him in 2 installments (50% upfront 50% when the guitar was ready to ship) via bank transfer. Thinking about it now…. Well let’s call it “Living life dangerously…” 😂
@@TheoreticallyMedia by the way I just saw a fender partscaster with George Harrison’s rooftop rosewood look and it looks amazing indeed. Plus I love the Beatles so it gets extra points.
99$ a month is insane.
Yeah, that's the downside of being so small, and likely without a ton of fincancal backing-- compute is going to cost a lot, and they don't have the deep pockets to offset it-- I presume at least.
Still, if they can get a good enough user base going, I'm sure someone will come along and throw some money at them, and then the cost can come down.
@@TheoreticallyMedia Hotshot was opensourced before this model. Hopefully they release an open version maybe not the full one thats very understandable
@@TheoreticallyMedia lol I bought a $95 subscription for Runway ai because after tinkering around with it with the free credits, I ran out and wanted to make more video clips and use Lip Synch. I'm in the process of creating my first ai short film with a script by ChatGPT. Dialogue heavy of sorts. A short film tests to see what can be done and I'm loving it!
Matt Wolfe done some live prompts. Everyone was terrible.
I've been using luma and it's never produced bad video after bad video... it's usually an even split of 50/50.
@@armondtanz Thanks for that info ill check that out!!
Thank you for the efforts you are making. This is an honor for us. We are proud of you
As far as the 5 sec or 10 sec question, I am satisfied with 5 sec and the ability to extend an additional 5 sec if desired.
Most shorter videos like commercials, music videos, trailers, etc.. don't go past 3 second clips which is what I use it for.
But if you plan on making movies then I understand 10 sec.
Yeah, I think now that we've gotten into the era of being able to lip sync and provide "acting" to our AI characters, the need to 10 second shots (and possibly longer) is going to be more of a thing.
That said, most films and TV shows still won't hold a 10 second shot-- if it's a dialogue sequence, you'll often cut to a reaction shot. So...maybe 10 seconds is the sweet spot.
Good to have you back Tim. 👍 Completely agree with x is the new y killer case issue. Different use cases, different workflows, more options...Look at how many men in blue suits we have on the streets! With FreePik and Maginific together, I really cannot see how they could possibly mess up photorealistic images at least. Considering Magnific is pretty expensive by itself, $99 may seem on the steeper side, but we'd be getting two very capable platforms. Stock photos are going to take an even bigger hit. That's bittersweet, but technology beats humans. Speaking of steep, you should end your videos with LTX studio's pricing. 🤯 10-second joke - can't believe you 'didn't' go there. 🤣
Freepik are really working on some amazing tools. I'm primarily working on pitch documents, treatments and concept art and it's been amazingly useful. Their existing upscaler is amazing, with magnific it's only gonna get better.
I actually feel bad about not checking in on Freepik. Since posting this video, a lot of comments about how many people love it. Sounds like it might be the best kept secret in AI!
The almost weekly introduction of multiple AI platforms is just insane, freepik is pretty good. You get the 2 magnific upscaled so I am sure they are busy integrating it.
With so many new AI platforms appearing weekly what do you think about doing a video on the remains of the failed platforms and why they might have failed?
On the clip time anything over 4 seconds works for me. If you can extend even better.
I'd love to do a "whatever happened to" type thing-- but if there's one thing I've learned: Don't count anyone out. I've seen a few platforms that were proclaimed dead, only to see them come roaring back with one update.
Rumor has it, we'll all be talking about Pika again in a few months. They aren't dead...they're cooking.
this AI image generator looks absolutely stunning! I'm really excited about the potential of AI in creativity.
Always the best Tim! Love your videos everyday. Keep going!
13:13
👍 agreed, I do praise and intellectually support people developing AI tools for creative people in spite of the fact that my bar for usage and financial support is relatively high.
Many Thanks, Tim, you're a Legend, my Brother!! 🏆
Thanks Tim, good to have you back... seems like the new arms race is the somewhat 'tricky' price tags for these things... $99 per month... ouch!
It's like the current LTX Studio prices... to get out of the 'personal use only' tier, you need to stump up $175 per month.
It seems that the AI world is going to be taking a leaf out of the cable TV subscription model where you end up with a dozen subscriptions to a variety of services, end up watching/using only a couple and lamenting the cost but never quite capable of cutting the cord for good.
The future will needlessly be a two tier society of those who can afford generative AI tools and those who will be mercilessly trolled by them.
What a wonderful world.
So much happened during that one week you were away -- next vacation, I expect Digital Scorsese to launch! Keep up the good work, Tim.
Haha, I legit thought about training up a clone avatar-- BUT, I'd still have to type in all the news, and I'm pretty sure the family would have gotten mad at me for hanging at the beach with my laptop! haha
Totally know what you mean with the Mr Robot reference! They did such a good job with their visual styling. Loved the subject in a weird corner vibe
Great info as always Tim!
Thank you Tim for avoiding the click-bait headline generation. Great to have all these options to pick and choose.
Nice one time, love the ideogram update overview… most impressed with the mock catwalk promo t shirts render .
Mystic looks really promising. I love the simple interface. ($12/Month for the annual plan - includes all of their various tools. I'm glad other companies are finally competing against Adobe.)
Its bizarrely the same as freepik.
I'm quite shocked a lit more ppl havent got onto it. $8 a month ($100 a year) . You get flux, magnific ... 1 million tokens per year. I'd be lost without it.
I sound like a bot who's promoting it but it is sweet and offers something a bit different to MJ.
Saaaaame. I think most AI generators are getting better with UI-- and hey, we aren't on Discord anymore-- but those early versions of Leonardo? Man-- that was like getting whapped in the face with sliders and options. (To note: They've gotten a LOT better about it!)
Is it that cheap? I didn't even realize that it was that cheap! Man, they should hire you for PR! I'm signing up now!
@@TheoreticallyMedia looks like my comment got removed. Curse u YT
Yes it's incredible.
I'll let u into a secret I signed up blk friday and got it 50% off 🤣.
For the next 12 months, we will see now weekly a new stunning image generator, and video generator. every week again. Every week mind blowing again. Every week groundbreaking again. Every week next level again.
For sure we're in a growth period. It ebbs and flows. With Flux being open source, we're going to be seeing A LOT of image generators pop up for sure. And I'm cool with that, as long as they all bring something unique to the table. Not just: Look, it's Midjourney.
So tired of that...so tired.
Wooho glad ur back hope u had great Vaca and was looking forward to your vibes, input and jokes
Oh it was the best! NGL, around the last day I was itching to get back! Apparently I get itchy to jump back in after 6 days! That first day was AWESOME though!
haha, glad to be back!!
Since AI has hit us, I've been able to express a lot in music and video combined, it has been a great source of exploration. A lot of people I send my stuff to just don't get it, but there are those that really do, it's another tool really, to use, just like a brush and paint, if you're creative you will make creative stuff just like any other medium. It really isn't like you can just push a button and out comes a hit music video, you still have to mix it, find the right sites to animate to music etc. I use SUNO as it really is exceptional at getting you're vibe if you input correctly it almost seems to understand what you are after and it becomes a very useful tool. Thanks for keeping us all informed, I take your videos seriously as I know you are right up there with the latest AI. I must admit I was sceptical of AI at first like a lot of people, but now see it as a useful creative tool.
Funnest ai news bringer 🙃
Haha, thank you! My kids roll their eyes at my jokes, so I’m glad you appreciate them!! Haha
Thank you! Great work.
So happy you are back posting again. Please never go on vacation again.
Haha, can't promise that! but only once a year, sadly...Already dreaming of the next one!
SWEET Peavey Horizon!
I LOVE THAT GUITAR!! It's such an oddball, but man-- LOVE IT! The stupid amount of tones you can get out of it is unreal. Little Peavey Wonder!
Ha, that was a great Mr. Robot shout out!
Love that show! Might be time for a rewatch!
oooh, best looking AI spy EVER!!!
ahhh-- makin' me blush over here! Haha, AI Spy me is much more dashing!
@03:38 I assume it's a combo of flux that is then run through their Magnific setup before you see the image
Something like that for sure- but maybe without the upscale? I still think the upscale side sprinkles some magic on it as well.
It’s certainly interesting. And I do love the idea of being able to choose Loras for fine tune control.
It looks promising!
Has the model gone ? It’s not an option from the model choices on their website any longer ?
It's in Alpha-- so only a handful of people have access to it right now. That's why I was kind of posing this as a "preview"-- it's coming soon though!
Honored ❤
I know it's not your usual area but I'd love to see some alternate non-photography tests on some of these platforms. You mentioned it but having a few more illustrative generations in your back pocket like our favorite business clad man or pirate lady would be helpful. I've found that new image generators like Flux are focusing heavily on photorealistic images but have a lot less depth than Midjourney. So while I can get close to Midjourney on Flux for a photo, as soon as I try, say, a fantasy illustration, it quickly falls waaaay behind what Midjourney can do. I think having a small handful of alternate styles to throw in your tests could show some differences people are missing between these image generators.
Anyway, love the video as usual!
P.S. Also really love your point about different Gens doing different things. I'd love if they specialized even more. Ideogram v2 is out today and it does the text Gen/Design stuff soo much better than anyone else.
Yup! I was playing around with Ideogram today-- it's still the king of Text and Design! As far as more illustrative stuff: I've got one coming up pretty soon-- Great note as well, I'll try to put more tests and examples of that into future videos!
Thanks for the info. I was hoping for a free to use generator like hotshot. I'm gonna try and make some weird stuff!
Currently only 2 generations per day, so sadly it might take you a bit to make a full project! But, for sure- check it out and let me know what you think! And go weird!’ Weird rules!!!
Best image to video programs?
i actually started using klingai which you recommended previously. i am open to experiment on more ai in future.
excellent to hear! Kling is really killing it!
@@TheoreticallyMedia I can't wait for the KlingAI Video Editor which is greyed out at the moment. I think Kling are the one to watch.
You can get a grasp of what 'structure' entails in Freepik, by changing the model to 'classic'.
Ah good deal! But, that Adobe FF generation was pretty funny-- oh, Firefly. It's like everyone's little kid brother.
It's always wonderful to see new AI tools! Think about looking at alternative solutions that provide special benefits for your artistic endeavors. #AIImageGenerator #Technology
Bravooooo 🎉😁
Thank you thank you!!
question: i use paid version of freepik but i can´t see MYSTIC as a model. i have 3 times different FLUX and two time CLASSIC. where do you find MYSTIC?
So, it's currently in Alpha-- meaning that only a handful of people (I was lucky) have access to it as we're testing and stressing it out. I think it should be opening up to a wider release soon. So, I mentioned in the video, think of this as a preview of something coming soon.
I know, I know-- I hate doing these videos where I'm like: "Look at this thing you don't have"-- but, it IS coming-- and I really loved the look, so I wanted to share it with you all!
@@TheoreticallyMedia no problem at all. i am a viewer from germany and i like your videos. i did not realize that this was ALPHA, sorry. and as you certainly noticed, i also liked the comments here which stated that you are not doing videos with untrue spoilers. you are akind of "real" and not hyping everything that is new in AI. I really enjoy your videos . And to be honest, i'm also a little envious of her music skills... ;-)
not her but YOUR music skills....
When will we get real-time ai filters for videogames?
Freepik, is one of the stars of AI image generators I always use in my production. But lately, Flex module starts to enforce its natterator to the prompt. Especially on color grading. Using some words like mediaval. The only AI can give details to background. In contrast, Leonardo AI, always fail to give details in characters in background, it only provide render to foreground character and the background characters are murfed.
Is that with Leo’s new Phoenix model? To be honest, I haven’t played around with it too much, but in their defense, it’s a brand new model that (I believe) they built from scratch- so, it’s a bit like a v.1. So, I can see that with the background characters. For now at least- I’m sure it’ll get better as time goes on and they continue to improve.
On the Flux color grade side- that’s where those Loras are going to come into play! As stunning as Flux is, it’s still a baby model. As the community starts building workflows, it’s going to grow into a real powerhouse!
It’s funny, we took this big step forward lately, but that also required taking a half step back’
is mistic can use for commercial?
Since it's in Freepik, I'd imagine so?
Could you share the link to this platform?
It's Freepik, but the Mystic model hasn't been publicly released yet. I'll let you know once it goes into open Beta!
Check now. The structure button is working for me. Will be posting some comparisons.
Ahhh, sweet! Hopefully the results are a little more striking than Firefly! (Although, again: I’ll give FF its due! That was a tough ask!)
Thanks for these videos dude. Any insight on the best video upscalers available currently besides KREA?
I mean, I hate to say Topaz, because it costs so much...but, Topaz.
I'm actually surprised no one else has picked up Creative Video Upscaling like Krea has. It seems like such a slam dunk idea, but to my knowledge, they're the only ones doing it!
Both seem very impressive.
whats the website for mystic my freepik doesnt look like your ui
I think an extend feature and a mini prompt to direct the next 5-10 seconds of action is the most useful thing any text2video ai can add for users. I am interested in making full length (90 minute) movies and therefore I need to be able to make videos up to 3-4 minutes. in moonvalley I had to generate hundreds of clips and use a linear blur to match up similar scenes, very labour intensive but it has it's own unique art style I guess.
where is the link for Mystic AI ?
Do you know of any upscalers as good as magnific ?
There's Krea which is fremium
And Leornado AI universal upscale fremium. Clarity upscaler on their website and replicate which freemium. And tones of upscaler models on huggingface and replicate to try
not from the author: i am very satisfied with TOPAZ PHOTO AI since over 1 year. this must be installed locally but you have full control and they have updates at least all 4 weeks.
What’s the link for the new tool?
12:59 😂 Made my day
Haha, I’m not making that joke, but we all know what it is! Too low hanging a fruit for me! I’m better than that!
Awesome video as always.
Can someone help me figure out if after the freepik and magnific merge, the actual upscale on freepik is the same as on magnific?
After testing it, it feels kinda inferior to magnific, but I might be mistaken.
2:45 okay, I'm guessing that these places paid models to pose and paid a flat usage fee in perpetuity so that the AI could use that to create these photorealistic images just like people are paying big money for voice actors to read like 10,000 words and phrases so that they can have one more voice model to draw from in creating a voice over from text. cuz 2:45 does not look like a 3D model or anything generated by a computer to me. It looks like a literal photograph of that guy but greenscreened with a blurred background. How soon before people begin recognizing particular faces and images from these AI programs? Just curious.
Perfect 👍👍👍
HotShot looks .... err... Cool!
; )
Mystic seems like Flux.1 dev with 4x upscaller and lux. Very impressive
Good call. Very promising model for sure. I’m sure someone on the open source side will be working on reverse engineering it by the end of the week as well!
They keep increasing the quality of these things without adressing the most pressing issues. I can make great images with stable diffusion right now. The diffrence between that and "the best generator there is" wouldnt be huge. Usually the better models just have better prompt understanding. What i would really like to see is a generator that understands the passing of time and can portray that in a series of pictures that are in chronological and or logical order. Basically the ability to create a visual story from a written one. And if that is too much to ask for, give me a manual one where i can just promt for picture 1 to 10 myself and the ai then generates me 10 images based on my prompts that folle the outlines story or progressen. Bit of a rant right thereXD
will mystic be open source? seems natural with freepik hosting so many open source models and having "free" in their name. either way it looks nice
For the most part all these Midjourney and Sora killers are superfluous. I’m finding that all I need is Midjourney and RunwayML Hedra if I need some lip syncing. The rest Dream Machine, and Kling aren’t necessary. The next model that can “kill” would be a video model that can do dynamic action coherently.
Man, dynamic action is such a holy grail right now. I think we’ll see it soon via Video to Video referencing. Kind of like Viggle, but one step up.
We’ll likely see it this year.
@@TheoreticallyMedia Yes please! I have so many projects that need it! Or simply, “someone walks into a room and sits down at a table.”
My eyes like light mode WAY better than dark mode.
Haha. I’m basically Gollum in the cave. The light! It burns!!!
5:38
😆 me too!
There will be generation models under our Christmas tree this year.
10 seconds is... somewhat useful. If consistency was more consistent, I would not care as long as we could extend to 30/60/90 with multiple generations from the first prompt. But mine generally tend to look great for those first handful of seconds then warp into some catastrophe or completely different scene
I'd really like to see AI reviewers test consistent characters capabilities as well as if the AI can generate multiple distinct people if you describe them independently.
Instead of "3 guys standing in the middle of a busy street", describe each guy and see what it does.
Why you dont share the tool link ? Your videos do not contain link always
I'm hoping it gets to the point of just a simple single downloadable app to install on the pc. Not 1111 or anything like that but for people who don't want to install a million things. Guess in time :)
Someone will crack it at some point. Right now, there are options, but like you said: There's a lot of hoops to jump through to get it working. I had played around with doing a Comfy tutorial-- but, man, that ramps up so quick. Plus, I know those types of videos just turn into people asking me for tech support in the comments! haha...I do enough of that in real life!
Speaking of which, I have to fix my neighbor's bluetooth connection to their printer.
Sigh. Sometimes it sucks knowing how things work.
Is this supposed to be available with a free account? Mine is grayed out.
10 seconds is generally acceptable but is still a it limiting if you want to make a proper film where characters are going to have any amount of dialog without having to try and look the video in an inconspicuous way. 2 minute long continuous shots are only necessary in certain scenarios but 20 or 30 seconds would open a lot of doors and 5 isn't enough for anything other than trailers, especially when you need to edit around weirdness so maybe only 2 or 3 of those seconds are usable.
Hotshot seems better than Luma to me so far, less morphing and weirdness in my first couple of generations.
"Built by four people"
It just goes on to show that people who are really dedicated and are into it are far more capable of producing amazing stuff compared to all these 'companies' trying to capitalize on the AI hype with sub-par workforce of dozens or more lol of course there's some internal politics and investment issues related to it, but whatevs. Thanks as always for the latest update :)
Also considering whether 10 secs is enough, could be the sweetspot because all these generations need to be edited anyways to make the video more immersive and engaging to look at (while also hiding AI flaws as best as possible)
totally agree on that smaller team front. Quick and nimble for sure! On the 10 second front: I had a bit that I cut from the video talking about doing a one'er, like the attack in Children Of Men-- the thing is, I don't trust any current model to maintain the level of consistency needed for a shot that long.
For sure someone in that scene would sprout an extra head...although, that might be cool. But a different movie for sure!
Light mode = lightroom. Dark mode = dark room.
Light mode in the streets, Dark Mode in the...haha...yeah, I know-- people always yell at me for the actual need for Light Mode-- but still, it's like staring into the sun for 8 hours! Too much!
Honestly Dark mode is the best UI invention and I believe Apple iPhones were among the first to make this popular. I never can use light mode on anything anymore. PERIOD. 😂
Ideogram 2.0 is now available.
Gonna try to cover it tomorrow. It’s pretty great!
"Sora-level image generator". So deep in the cult.
that hotshot thing is not sora quality, its pretty wonky and early. the gens you see might be cherry picked from 20 -30 gens
I tried it. though sometimes when it works. it gives better quality then the others. (assuming humans wont have 3 arms.
Haha, I'm telling you, once the whole 3-arm thing is totally stamped out of AI Video, I think I'm going to miss it a bit. There have been so many gens that went wonky that really made me laugh.
Good to know about Hotshot...Yeah, I think the cooler story in the whole release is that it was built by 4 people. Let's see what they can do in the next 13 months!
i learnt that creating video with pre made image is simpler, generate a starting scene in image,
Flux is still version 1 and it is killing it
you can use replicate to use it easily, it is almost free with its pricing
But as you mentioned, it is great because we options now, you just summed it up
Great video
thank you
I got really confused with this vid. Im sat there watching this new mystic website and im thinking 'whoa this looks exactly the same as freepik'. Musta missed that bit 😂.
Im always shocked why freepik never gets a shoutout . $90 a year while MJ is $30 a month.
You get 1 million tolkens now.
I'd be lost without it
I hadn't really looked into Freepik until now-- someone else mentioned that price, that's NUTS! I'm all over it now!
Freepik helpfully deleted the images I created before and apparently their ability to create images using a doodling space is no longer accessible... Downgrade as far as I'm concerned.
Personally I'm fed up with Leonardo because of the censorship and limitations of options now. They started off great and sank like a stone . Lately I'm using CGdream and it's much better.
Hey bro, I just released a free and open source app to make interactive videos and films. Would love to see you make an interactive Ai short like Bandersnatch.
Oh that's awesome! Send it over to Theoreticallymedia@gmail dot com-- I'd love to check it out!
Totally agree, this "Midjourney" killer headline is becoming boring as hell...
If anything, I guess it’s a compliment to MJ, as it is sort of the de facto king of the hill with the most name recognition.
Still, at this point, I think it’s just lazy content. We gotta start branching out and finding new angles.
At least, that’s what I’m doing on this channel.
Still waiting my 2 clips are still in the queue 24+ hours. LOL!!
its not mystick. i was able to check some codes in console in older version luckily. its just FLUX.. they fine tuned it
justspent the last 20 minutes tring to create a username, the UI seems broken atm.
I subscribed bc you dont have clickbait, referrals, etc., hidden into the content or on the page.
That video AI was BS. another youtuber bought the subscription and couldn't get anything close to the examples
12:50
10 second video generation should be the minimum........ with the caveat that it can't be slow-mo. If slow-mo, then minimum must be 20 seconds.
I like it
the video model that was built by 4 people was reviewed by Matt Wolf, and it SUCK, not only suck, SUCK so bad, the demos they had is completely can never be replicated.
Well, like I said in the video-- the real story to me is that it was built by 4 people. And yeah, it might suck today-- but let's give them some time to scale up and develop it out a bit. I'm NOT suggestion to pay for it...but let's give them a round of clapping for releasing and see where they take this next.
@@TheoreticallyMedia I have no problem with that, but the issue is that they are scamming people into thinking they can generate like the demo and charging $100 for totally useless generations, that's a bad bad sign to startup your company with..
Are you doing a video about Canva buying leonardo ?
Did one! th-cam.com/video/HBkyiQwbm6E/w-d-xo.html
@@TheoreticallyMedia thank you for your reply
I love Magnific and I use is just about everyday. There are ni rumors of if going away correct? I paid for a sub.
Nope! Magnific is going to stay independently operating! They aren't going anywhere! Near daily user here as well!
This is all very clever but what are you supposed to do with these images?
I finally get it, I've been a psychopath all my life then lol
Haha, so you also enjoy staring directly into the sun? I see you!!
@@TheoreticallyMedia lol
Hotshot examples were super cherry picked. Most generations sort of suck tbh.
Eh, like I said: It's like day 2. Let's give them a minute to find their footing. The coolest part of the story to me is the fact that 4 guys pulled this off. That's the super cool part.
Fingers crossed they'll improve and maybe even surpass....well, I was going to say Sora-- but, y'know...
Just dropped this AI Gucci Video Commercial using the new Runway 3.0 : th-cam.com/video/vy08ivkUai8/w-d-xo.html
“Only psychopaths use light mode” 😂
mystic is same functionality like freepik
Yeah, same UI. I suspect it’ll be ported over to Magnific as well (I mean, why wouldn’t it? They own it) and maybe we’ll see some UI changes there. But then again, if it ain’t broke, why fix it?