Thanks again to today's sponsor, Hostinger! If you'd like to build a website with AI Tools in just a few minutes, head on over: hostinger.com/theoretically
On the free Recraft plan, all generated images are not yours: "Images are owned by Recraft. Upgrade to own all your next generations." So probably can't use them anywhere commercially.
That whole commercial license stuff is complete nonsense. They trained their data on stolen art so they have no legal standing to be able to grant you a commercial license whether or not you buy one. And equally if you don’t upgrade and get access to this fake license who is going to ever find out. It’s all smoke and mirrors.
Legally- no ai generated art from any software including open ai is copyright protected. This means you can use any images made by any software publicly and commercially without recourse.
for any model to even be placed in the sentence " _____ the flux killer" it needs to be free, OPEN SOURCE, locally runnable, training enabled....if its missing any of those components, then all your doing is making a clickbait title. compare apples to apples...."red panda the MIDJOURNEY killer" would be a maybe statement....
@@TheoreticallyMedia it will never be a flux killer unless it meet all 4 of the things flux already does....period. it would be more accurate to thumbnail these closed source models as "midjourney killers" becuase then its apples to apples.....please be considerate about that, and consider that flux is in the category with sd3, sdxl, pixart, ect open sourced models...and adobe firefly, midjourney, ect is a completely different category of model...closed source, no training, not really free, not locally runnable....your information is spot on most the time but i feel like whoever you are getting to do your thumbnails cant tell apples from oranges.....
Wow, that Blendbox layer feature is wild. Midjorney needs that ASAP! You got great taste in movies and tv-series. Dark was amazingly complex, and earlier you also mentioned Mr. Robot.
I want that Blendbox workflow in ALL of them! I’d also love to see it with real time drawing as well! (Plus importing your own images, locking etc)- it’ll get there! Oh man, Dark and Robot are so good! When Severance returns, I’ll be making a lot of references to that as well!
@ take a look at Moho - enables you to add bones to vector art (rigging) and then animate from there. I created over 70 television cartoons for Aussie kids group ‘The Wiggles’ using Moho (20 years ago - omg) - the illustration work took forever. I wish I had this tech back then.
Honestly Bro... either we watched the same shows as kids, we went to school together or we were separated at birth... But your Humor and oneliners are GOLDEN! I subscribed during the Pika prompts days and I had no idea what I was getting into! Keep doing what you doing bro and keep filling the AI/media need. If you build it ... they will come! (Remember when SNL was actually funny? Then you mix that with the Nightly News but just AI Tech stuff and a dose of Seinfeld) that's how I describe Theoretically Media. I'm going to be in bed tonight hearing: "The calls were coming from inside the house" and laughing myself to sleep. Great job Tim!
@@TheoreticallyMedia True... and LOL Last night I watched the scene from "When a Stranger Calls" 1. That was a really bad idea! 2. Let's just say I DIDN'T laugh myself to sleep. Thank goodness I had Alexa play my AI Lullaby at full blast. 🤣
When you made the ELO joke I had flashbacks to the 70s. Cyberpunk lady really shows what it can do. I judge image gens with my standard character and the recent generators like flux and this are pretty impressive.
I absolutely HATE canvases. Can anyone explain the appeal to me, please? 😢 I keep accidentally shifting elements around, there is no way to sort/align elements automatically, mouse wheel zooms instead of scrolling, etc., etc. What is the appeal and the convenience?! Loved Kaiber so much, exactly because of impressionist look, ability to set camera movement, flexible time approach, and audio reactivity - and they are becoming just another API aggregator, with the stupid nodes of semi-realistic-yet-crappy and unusable other generators 😢 Sadness. Will have to wait and see what kind of video gen Midjourney will make.
WOW @7:47 Being able to generate vector images is def gonna change the game. At least as far as graphic design and animation are concerned that is a gonna be a ginormous time saver.
I'm not a power Illustrator user, but offhand, I'd say: Adobe? It's just that industry standard thing. On the photo side, I'll mess around with Affinity (which I do like), but ultimately, most of my work is done in PS...just because everyone uses it.
(Re-commenting now that I watched the video more carefully) Interesting video! Some points: - Vector generation is the big story here about the model. Hopefully that is the model, not just a "Live Trace" of a posterized raster image. Based on the video it seems legit because of the layers, but hard to say as the layers it created are a bit non-sense - The ELO score thingie sounds easily spoofed if you can recognize the telltales of base models or read meta data from a picture so I wouldn't trust it much as a highly motivated maker can likely buy votes. - The interface is also cool, and it's an attempt at monetizing basic things a power user can do in Comfy with any model (except vector). Since those basic things and ComfyUI itself are not so basic for 90% of users out there, it makes sense to package those workflows in a good UX and make it more acessible to casual users.
great points here! Totally agree that the Vector/SVG thing is the real hidden feature here-- but, I'll say that most of the viewership won't realize how big a deal it is. (I mean, I see it-- but, as mentioned: I suck with Illustrator!) 100% on the Comfy front. I use Comfy here and there, but-- yeah, 10% is probably a pretty accurate assessment of the overall user base in GenAI. Fairly steep learning curve and a prereq of needing a decent machine. I try to focus on the stuff that is available to the wider public, as I want the tech to be in the hands of everyone. And-- not everyone has the time or resources to go local/open source.
Tried 'rock concert, full body portrait, a beautiful female rock star singing with a mic on stage, black nose ring, trendy short silver hair with red streaks, she is screaming into the mic, she is pointing to the crowd, dynamic pose, black leather pants, delicate shoulders, tattoos on her arms, big black industrial shoes, big sign behind her with text "NIN", text "I am not a panda" on her arm, detailed hazel eyes, dramatic lighting, intricate details, depth of field' - It got EVERY one of those elements correct in photorealistic mode on the 2nd roll. Fingers fine. Anatomy fine. Pose perfect. Thankfully, not a standard expected composition. All in 12 seconds. With a decent creative upscaler and some Photoshop work, it's a pretty darned good generation. I'm an MJ fan, and I tried it there. Nope (although the photorealism was a little better in MJ). Tried an acrylic abstract painting and MJ killed it with the textures. The moderation seems just as ridiculously stringent as MJ, though. In the prompt above I had 'big black boots', and that was moderated! 😆 Not a bad start at all. Thanks Tim. 👍
Awesome to hear! Yeah, the prompt adherence is kind of stunning. That opening shot of the lab coats with the blue business suit guy was first roll. If you look, it kind of failed on the background, but other than that, all the prompted “asks” were there. I likely could have rolled a few more, or just regenerated the background. It’s impressive. Although, agreed: it needs a better upscaler.
It's PRETTY good! Actually, you'll have a TON of fun with the Blockade labs thing as well! You can totally get lost in that one! Also: Watch DARK on Netflix!! SOOO GOOD!
Midjourney still no.1 , for their aesthetic images and custom character reference and Style reference and other stuff , We don't need random images, we can control images that can be used in consistent character and film making 😅
Thanks Tim, I'm checking this one out, and I really enjoyed your film analysis for the Curious Refuge Film Contest, which I recently went to re-watch! Cheers, brother
How is the censorship on red panda? One of the key features of flux (imo) is that it lets you generate whatever without some moderation message saying "errmm you cant do that ☝️🤓"
@AG_before yea it can be moderated via the api, there is a "safe" api version. Running locally should be uncensored. There are some places online that seem to be running the uncensored version but theyll ip ban you if you make enough non pc images lol. Go with locally run if you can
❌, wrong, still can’t do MOST copyrighted characters, still only does the most popular ones. 7/10 (cause it still gets some aspects right with proper stylization aspects)
To be honest, I'm actually not that interested in creating images with IP. I mean, I get it from fanfiction and Memes and all-- but, with the technology so early in, I'd really rather all of us focus on creating our OWN properties. Y'know? Why keep building on rented land. We've got a whole world to explore.
Thanks for all the videos, love it. Question, what tool would you use to make a person hold a magazine with a custom cover, one you can design yourself, like a Vogue magazine with your own face and title?
Honestly, give this one a shot. That “scientists wearing lab coats and looking at a computer with a man in a blue business suit” was generated with Red Panda. It’s pretty good. If you have a specific face, you might need to face swap on the magazine, but I think it can be done.
@@TheoreticallyMediathanks, I will give that a try. Haven’t thought about face swapping, been busy for days making a music video with Midjourney and Runway, really cool stuff. Thanks for sharing all your content, try to watch asap after getting the notification 🙏
Recraft is good but still nowhere near flux or midjourney levels. I found most generations are the same, not enough variation, different angles etc so I would assume they using similar seeds every time. Faces are quite distorted too. Not sure why Recraft was at the top of the list. Should be about 5th at most.
You’re best ai UA-cam maker ever we love you from iraq 🇮🇶 me and my all classroom students,,,and your easy way 🎉 we love the ai tools news and upgrades just in your UA-cam channel so thank you 🙏
Hm! Reflex is quite interesting, but didn't follow any of my prompts well and the faces look a bit distorted IMO. Not to knock Reflex as there is definitely potential with this, but introducing Reflex as the image generator that "beat Midjourney and Flux" is going to set high expectations that it may not meet yet IMO.
It's pretty neat! And check out that Blockade Labs thing as well. Not 100% there yet-- still needs some things like locking layers and importing your own images-- but still, it's a TON of fun!
I didn't get a chance to dig into it too deeply, but I appreciate the editing tools available-- that's a super important part of the AI image generation process!
I like this ^^ An AI for in-game textures changer could be nice, I know my comment have nothing to do with your vid but I imagine games with real life gameplay graphics. ^^
If it ain't open sourced it ain't actually "free". The costs for this red panda model outside of the very limited free gens are actually fairly high.. So sick of calling companies giving like 25 images a day "free".
@@TPCDAZ If its free and on another companies service YOU are the product. Most reputable companies release a toned down open sourced version for people to use locally. Your car analogy makes no sense to the situation? My saying if its not open sourced its not truly free is not asking to go through the finances of the Ai company that released it. Dont be silly now.
@@TPCDAZ except with the comparison is made with flux...."red panda the flux killer" is clickbait....flux is free, open source, locally runnable, and trainable.....
@@omegablast2002 and not everybody can run Flux. Don't assume all because you can that everyone can. Also it's not clickbait since it finished above Flux. so the title is correct.
@@JustArtsCreations The point I was making is all because something is free doesn't mean you get full access to everything. I didn't realize I had to dumb it down for you and actually no. Most companies do not release a toned down open course version at all.
Think Kaiber is going to end up adding this Recraft model into their new platform as well like the rest? Also I am trying to get on it and it does not want to load at all.
Just a heads-up Tim... NYC just passed a new law legalizing all jay walking! So man in a blue suit can seek redemption from woman walking along the beach!
hmm .. but what is the success/fail ratio with the credits..having every 6 generations be a fail is something that needs to be reviewed also especially on paid services.
It’s actually not that bad. I got roughly what I was looking for in a shot or two. Granted, I was not being super specific about poses or items. But hey, did you check out the blockade labs thing toward the end of the video? That seems like a really interesting workflow that will (eventually) have a stellar level of control
Haha, gotta keep the bills paid! That said, I always try to do stuff that is actually usable to you all. You'd be amazed at how many offers I get for stupid mobile games, but I always turn them down. I'm probably stupid and leaving a stupid amount of money on the table-- but, at the same time, I mean...those games suck?
It's the best. You'll end up grabbing it far more often than dragging it out of a case. Just a quick noodle here and there, but it keeps the fire burning.
1:50 I bet half your audience is not old enough to get 'Jill, we traced the call. It's coming from inside the house.' (When a Stranger Calls - 1979).l 7:27 Boop.
@@TheoreticallyMedia Flux is a distilled model but it’s their own model. SD3.5 is a base model. I think all base models struggle with human anatomy. I haven’t seen any exception. After watching your video, I tried it out and just like I thoughts. Hand, and toes are just terrible
You think? Compositionally, I was impressed w/ what it's doing beyond the standard "character looks at camera" but, there are some problem spots here and there that I'm noticing as well. To be honest, I'm not one for "this is better than that"-- but, y'know: the headline story was the whole Red Panda rocking the charts, so I had to follow that narrative. I think they all do pretty great stuff...just gotta dig in to see where each one shines.
I think in combo with some of the other workflows over on the platform, it's got something unique going on. I'll say, the one thing I've found if pretty good at is with composition. Like, it's not doing a lot of "character staring at camera"-- which is refreshing.
The red panda images are clearly squished. Maybe they're compositionally different from what we tend to get out of models these days, but that's it. Different, not better. In fact, clearly worse. Lots of moshed details. Overfit on blur. There doesn't seem to be a negative prompt option. Not trainable. Not local. Trash taste, trash ideas.
recraft is garbage compared to flux. just tested it out. very low quality by comparison. the images look ok, sometimes, but the prompt adherence is super limited. i couldnt even get it to do a nice graphic novel style image of a bear fighting a ninja. the realism images looked ok but it could only do so much. everything ultimately was quite bland and it could only pick up a few elements of the promt. i got a ninja in a blue outfit with a red scarf, but the bear simply wouldnt have samurai armor at all. it got grass and a sunset, but it couldnt get the style even when telling it marvel comics, graphic novel cover art, etc. it wouldnt take angle cues really, couldnt make anything dynamic, couldnt really take motion prompts very well. flux hears you and tries its best to give it to you.
I was right. Where MJ shines that these others don't is going further out on a limb in creative areas, not just sticking to cinematic realism or established illustrational or cartoon styles. And the "Free" image generator is only limited free generations per day, and there's a paid subscription.
well, yeah-- besides open source, there is no free lunch in AI. But, 25 gens a day is fairly generous. To note, I know it's a sensational title, but the metrics from the leaderboards actually DO say that users preferred the Red Panda outputs more than MJ. That said, I'm a die-hard MJ fan as well. They've been the one consistent AI Image generator in my toolbelt since day one.
Not really some advance, it doesnt follow prompt as needed, faces are ugly, if you ask for beautiful woman, it generates a huge "tank" looking creatures.... Also it doesnt allow some simple words categorizing it as NSFW
Thanks again to today's sponsor, Hostinger! If you'd like to build a website with AI Tools in just a few minutes, head on over: hostinger.com/theoretically
Hi 👋
On the free Recraft plan, all generated images are not yours: "Images are owned by Recraft. Upgrade to own all your next generations." So probably can't use them anywhere commercially.
Sad...you are right...
That whole commercial license stuff is complete nonsense. They trained their data on stolen art so they have no legal standing to be able to grant you a commercial license whether or not you buy one. And equally if you don’t upgrade and get access to this fake license who is going to ever find out. It’s all smoke and mirrors.
@@lukewilliams7020compute isn't free though. training and inference
Well it's not like they can just do this for free, it does cost money to have these servers do all this processing.
Legally- no ai generated art from any software including open ai is copyright protected. This means you can use any images made by any software publicly and commercially without recourse.
for any model to even be placed in the sentence " _____ the flux killer" it needs to be free, OPEN SOURCE, locally runnable, training enabled....if its missing any of those components, then all your doing is making a clickbait title. compare apples to apples...."red panda the MIDJOURNEY killer" would be a maybe statement....
I mean, ELO scores would disagree, but you have a valid point.
@@TheoreticallyMedia it will never be a flux killer unless it meet all 4 of the things flux already does....period. it would be more accurate to thumbnail these closed source models as "midjourney killers" becuase then its apples to apples.....please be considerate about that, and consider that flux is in the category with sd3, sdxl, pixart, ect open sourced models...and adobe firefly, midjourney, ect is a completely different category of model...closed source, no training, not really free, not locally runnable....your information is spot on most the time but i feel like whoever you are getting to do your thumbnails cant tell apples from oranges.....
@@omegablast2002 100%
Damn dude guess you better get on that.
Go ahead, pay for the development and release it. I'll wait.
@@michaelwoodby5261 I'm not in UA-cam comparing apples to oranges
Wow, that Blendbox layer feature is wild. Midjorney needs that ASAP! You got great taste in movies and tv-series. Dark was amazingly complex, and earlier you also mentioned Mr. Robot.
I want that Blendbox workflow in ALL of them! I’d also love to see it with real time drawing as well! (Plus importing your own images, locking etc)- it’ll get there!
Oh man, Dark and Robot are so good! When Severance returns, I’ll be making a lot of references to that as well!
Vector output is huge. As a 2d animator this is a game changer!
I know right! I gotta start not sucking at illustrator!
@ take a look at Moho - enables you to add bones to vector art (rigging) and then animate from there. I created over 70 television cartoons for Aussie kids group ‘The Wiggles’ using Moho (20 years ago - omg) - the illustration work took forever. I wish I had this tech back then.
Honestly Bro... either we watched the same shows as kids, we went to school together or we were separated at birth... But your Humor and oneliners are GOLDEN! I subscribed during the Pika prompts days and I had no idea what I was getting into! Keep doing what you doing bro and keep filling the AI/media need. If you build it ... they will come! (Remember when SNL was actually funny? Then you mix that with the Nightly News but just AI Tech stuff and a dose of Seinfeld) that's how I describe Theoretically Media. I'm going to be in bed tonight hearing: "The calls were coming from inside the house" and laughing myself to sleep. Great job Tim!
Haha, we need to grab a beer and stay up until 4am gabbing!
Eh, who am I kidding, we're old now and would probably be asleep by 11! haha
@@TheoreticallyMedia True... and LOL Last night I watched the scene from "When a Stranger Calls" 1. That was a really bad idea! 2. Let's just say I DIDN'T laugh myself to sleep. Thank goodness I had Alexa play my AI Lullaby at full blast. 🤣
When you made the ELO joke I had flashbacks to the 70s.
Cyberpunk lady really shows what it can do. I judge image gens with my standard character and the recent generators like flux and this are pretty impressive.
I absolutely HATE canvases. Can anyone explain the appeal to me, please? 😢 I keep accidentally shifting elements around, there is no way to sort/align elements automatically, mouse wheel zooms instead of scrolling, etc., etc. What is the appeal and the convenience?!
Loved Kaiber so much, exactly because of impressionist look, ability to set camera movement, flexible time approach, and audio reactivity - and they are becoming just another API aggregator, with the stupid nodes of semi-realistic-yet-crappy and unusable other generators 😢
Sadness.
Will have to wait and see what kind of video gen Midjourney will make.
WOW @7:47 Being able to generate vector images is def gonna change the game. At least as far as graphic design and animation are concerned that is a gonna be a ginormous time saver.
As a graphic designer, the ability to get SVG's I can use in Illustrator is kind of sick.
I'm in inkscape guy. What are the benefits of illustrator over inkscape do you know?
It's a much bigger deal than I think is being noticed! Time to start boning up on my Illustrator skills!
I'm not a power Illustrator user, but offhand, I'd say: Adobe?
It's just that industry standard thing. On the photo side, I'll mess around with Affinity (which I do like), but ultimately, most of my work is done in PS...just because everyone uses it.
Free images generation?...I'm in...
Favorite Four Letter F word!
(Re-commenting now that I watched the video more carefully) Interesting video! Some points:
- Vector generation is the big story here about the model. Hopefully that is the model, not just a "Live Trace" of a posterized raster image. Based on the video it seems legit because of the layers, but hard to say as the layers it created are a bit non-sense
- The ELO score thingie sounds easily spoofed if you can recognize the telltales of base models or read meta data from a picture so I wouldn't trust it much as a highly motivated maker can likely buy votes.
- The interface is also cool, and it's an attempt at monetizing basic things a power user can do in Comfy with any model (except vector). Since those basic things and ComfyUI itself are not so basic for 90% of users out there, it makes sense to package those workflows in a good UX and make it more acessible to casual users.
great points here! Totally agree that the Vector/SVG thing is the real hidden feature here-- but, I'll say that most of the viewership won't realize how big a deal it is. (I mean, I see it-- but, as mentioned: I suck with Illustrator!)
100% on the Comfy front. I use Comfy here and there, but-- yeah, 10% is probably a pretty accurate assessment of the overall user base in GenAI. Fairly steep learning curve and a prereq of needing a decent machine.
I try to focus on the stuff that is available to the wider public, as I want the tech to be in the hands of everyone. And-- not everyone has the time or resources to go local/open source.
Recraft is amazing, always was, glad you are exploring it, deep dive would be awesome!!
Will do!! So much to explore on a deep dive that I think I'll be certified in SCUBA by the end!
aw, i was hoping red panda to be open source 😔😔guess we have to wait for flux 1.2
Sadly not-- but still, it looks like it's popping up on Replicate and other spots. It IS super cheap as well. So, there is that!
Tried 'rock concert, full body portrait, a beautiful female rock star singing with a mic on stage, black nose ring, trendy short silver hair with red streaks, she is screaming into the mic, she is pointing to the crowd, dynamic pose, black leather pants, delicate shoulders, tattoos on her arms, big black industrial shoes, big sign behind her with text "NIN", text "I am not a panda" on her arm, detailed hazel eyes, dramatic lighting, intricate details, depth of field' - It got EVERY one of those elements correct in photorealistic mode on the 2nd roll. Fingers fine. Anatomy fine. Pose perfect. Thankfully, not a standard expected composition. All in 12 seconds. With a decent creative upscaler and some Photoshop work, it's a pretty darned good generation. I'm an MJ fan, and I tried it there. Nope (although the photorealism was a little better in MJ). Tried an acrylic abstract painting and MJ killed it with the textures. The moderation seems just as ridiculously stringent as MJ, though. In the prompt above I had 'big black boots', and that was moderated! 😆 Not a bad start at all. Thanks Tim. 👍
Awesome to hear! Yeah, the prompt adherence is kind of stunning. That opening shot of the lab coats with the blue business suit guy was first roll.
If you look, it kind of failed on the background, but other than that, all the prompted “asks” were there. I likely could have rolled a few more, or just regenerated the background.
It’s impressive.
Although, agreed: it needs a better upscaler.
SWEET SWEET RED PANDA ALGO BOOSTING! I do agree, Recraft should go all in on a REDPANDA rebrand.
Red Panda is SUCH a better name!!
oooh was waiting for this video -- i wanna try digging into it this weekend
It's PRETTY good! Actually, you'll have a TON of fun with the Blockade labs thing as well! You can totally get lost in that one!
Also: Watch DARK on Netflix!! SOOO GOOD!
Nice content, mate! Thank you for sharing it!
Appreciate it!!
Nice!!, I just test it right now. Works very well with text in images
The text gen is going to give Ideogram a run for their money, for sure!
Midjourney still no.1 , for their aesthetic images and custom character reference and Style reference and other stuff ,
We don't need random images, we can control images that can be used in consistent character and film making 😅
Thanks Tim, I'm checking this one out, and I really enjoyed your film analysis for the Curious Refuge Film Contest, which I recently went to re-watch! Cheers, brother
Source available? no thanks, flux ftw
+1 for that deep dive of recraft!
Filling up my Scuba Tank now!
The Terminator box was great, I think the AI was merging Bruce Lee with Arnold, looked pretty good.
i can confirm recraft is really good for generating realistic images.
I wonder if flux has the capacity to stay ahead, they may be outatime.
; )
They need to pick up the pace, put the peddle to the metal and hit 88MPH!
this new model free and open source, and locally runnable? didnt think so....
They DO have Video on the way!
How is the censorship on red panda? One of the key features of flux (imo) is that it lets you generate whatever without some moderation message saying "errmm you cant do that ☝️🤓"
What, really? Where and how? I thought it was moderated too. Assume I don't know what API and all that means. Thanks. 😅
@AG_before yea it can be moderated via the api, there is a "safe" api version. Running locally should be uncensored. There are some places online that seem to be running the uncensored version but theyll ip ban you if you make enough non pc images lol. Go with locally run if you can
❌, wrong, still can’t do MOST copyrighted characters, still only does the most popular ones. 7/10 (cause it still gets some aspects right with proper stylization aspects)
To be honest, I'm actually not that interested in creating images with IP. I mean, I get it from fanfiction and Memes and all-- but, with the technology so early in, I'd really rather all of us focus on creating our OWN properties.
Y'know? Why keep building on rented land. We've got a whole world to explore.
Great reporting Tim, I love getting these nuggets fresh out of your creative oven!!! 😎👍✨
Thanks for all the videos, love it.
Question, what tool would you use to make a person hold a magazine with a custom cover, one you can design yourself, like a Vogue magazine with your own face and title?
Honestly, give this one a shot. That “scientists wearing lab coats and looking at a computer with a man in a blue business suit” was generated with Red Panda. It’s pretty good.
If you have a specific face, you might need to face swap on the magazine, but I think it can be done.
@@TheoreticallyMediathanks, I will give that a try. Haven’t thought about face swapping, been busy for days making a music video with Midjourney and Runway, really cool stuff. Thanks for sharing all your content, try to watch asap after getting the notification 🙏
Recraft is good but still nowhere near flux or midjourney levels. I found most generations are the same, not enough variation, different angles etc so I would assume they using similar seeds every time. Faces are quite distorted too. Not sure why Recraft was at the top of the list. Should be about 5th at most.
You’re best ai UA-cam maker ever we love you from iraq 🇮🇶 me and my all classroom students,,,and your easy way 🎉 we love the ai tools news and upgrades just in your UA-cam channel so thank you 🙏
Thumbs up for the Dark reference ▲
SOOOO good. I've actually been itching for a rewatch!
Hm! Reflex is quite interesting, but didn't follow any of my prompts well and the faces look a bit distorted IMO. Not to knock Reflex as there is definitely potential with this, but introducing Reflex as the image generator that "beat Midjourney and Flux" is going to set high expectations that it may not meet yet IMO.
Can’t wait to play tx mate 😻
It's pretty neat! And check out that Blockade Labs thing as well. Not 100% there yet-- still needs some things like locking layers and importing your own images-- but still, it's a TON of fun!
Great job awesome! This is why I pay monthly for all of my services. There's always something new coming out that's better than the next thing.
Very nice! Please make a deep dive!
Will do! And it DOES look like it goes DEEP! hmmm, gonna need my snorkle for this one!
So far all the AI image generators only satisfied 50% of my prompt expectations. Thanks for this upgrade. 😂
I didn't get a chance to dig into it too deeply, but I appreciate the editing tools available-- that's a super important part of the AI image generation process!
I have hostinger, very happy , already 3 years.
I like this ^^ An AI for in-game textures changer could be nice, I know my comment have nothing to do with your vid but I imagine games with real life gameplay graphics. ^^
There's a few things that can...kinda do it-- but, not in real time-- yet.
Why is this Flux killer? What does it do better than Flux? Does it have LoRa's? Does it have consistent character?
Mostly from the Image Arena Leaderboards. I hear you though-- actually, I DO have something coming up w/ LoRas! It's pretty cool!
If it ain't open sourced it ain't actually "free".
The costs for this red panda model outside of the very limited free gens are actually fairly high..
So sick of calling companies giving like 25 images a day "free".
Open source and free are two different things. You can win a free car, doesn't mean you get to roam the factory and go through their finances.
@@TPCDAZ If its free and on another companies service YOU are the product.
Most reputable companies release a toned down open sourced version for people to use locally.
Your car analogy makes no sense to the situation? My saying if its not open sourced its not truly free is not asking to go through the finances of the Ai company that released it. Dont be silly now.
@@TPCDAZ except with the comparison is made with flux...."red panda the flux killer" is clickbait....flux is free, open source, locally runnable, and trainable.....
@@omegablast2002 and not everybody can run Flux. Don't assume all because you can that everyone can. Also it's not clickbait since it finished above Flux. so the title is correct.
@@JustArtsCreations The point I was making is all because something is free doesn't mean you get full access to everything. I didn't realize I had to dumb it down for you and actually no. Most companies do not release a toned down open course version at all.
Think Kaiber is going to end up adding this Recraft model into their new platform as well like the rest?
Also I am trying to get on it and it does not want to load at all.
Just a heads-up Tim... NYC just passed a new law legalizing all jay walking! So man in a blue suit can seek redemption from woman walking along the beach!
hmm .. but what is the success/fail ratio with the credits..having every 6 generations be a fail is something that needs to be reviewed also especially on paid services.
It’s actually not that bad. I got roughly what I was looking for in a shot or two. Granted, I was not being super specific about poses or items.
But hey, did you check out the blockade labs thing toward the end of the video? That seems like a really interesting workflow that will (eventually) have a stellar level of control
Great a.i news within this hostinger ad vid.....
Haha, gotta keep the bills paid! That said, I always try to do stuff that is actually usable to you all. You'd be amazed at how many offers I get for stupid mobile games, but I always turn them down. I'm probably stupid and leaving a stupid amount of money on the table-- but, at the same time, I mean...those games suck?
Awesome let me try that Recraft imagen.
InvokeAI have had a canvas based UI long before any other platform, strange how they never get a mention.
I was a big fan of recraft even before Red Panda release.
Unrelated question, sir, did you watch Agatha all along? Since you had said "all along' quite a few times.
Haha, I shot this on Halloween and DID watch the final two episodes last night...so, maybe it snuck in there subconsciously?
That reminds me...I've gotta hang my guitars.
It's the best. You'll end up grabbing it far more often than dragging it out of a case. Just a quick noodle here and there, but it keeps the fire burning.
😂😂Nice!!👏🏻👊🏻
1:50 I bet half your audience is not old enough to get 'Jill, we traced the call. It's coming from inside the house.' (When a Stranger Calls - 1979).l 7:27 Boop.
"not E.L.O." ... Moments later says E.L.O. 😉
Haha. I need to hire you to hit me with a stick everytime I do that!
How dare you expose this😭 This was a huge gatekeep
Don't see a free version. Bad panda. Bad.
Then look harder. Very first tier says Free.
@@TPCDAZ they offer 10 credits..one generation costing 5 credits. How is that free
is this a distilled or true base model?
I think (stress think) it's their own model. I'm sure the gang on the leaderboards would have sniffed out if it was Flux or SD3.5 under the hood.
@@TheoreticallyMedia Flux is a distilled model but it’s their own model. SD3.5 is a base model. I think all base models struggle with human anatomy. I haven’t seen any exception. After watching your video, I tried it out and just like I thoughts. Hand, and toes are just terrible
Not better than Flux from my testing. Plus it's not open. I would say this is even a tier down from Midjourney.
You think? Compositionally, I was impressed w/ what it's doing beyond the standard "character looks at camera" but, there are some problem spots here and there that I'm noticing as well.
To be honest, I'm not one for "this is better than that"-- but, y'know: the headline story was the whole Red Panda rocking the charts, so I had to follow that narrative.
I think they all do pretty great stuff...just gotta dig in to see where each one shines.
I AM SHOCKED, IT JUST KILLED FLUX THE DOLL MAKER.
How can you use 360 environments?
I think I want something new technology not just upgrading things!!!
Mr Blue Sky reference! 👍🏼😂
I was going back and forth between that one and "Don't Bring Me Down"
Open source and on my pc on pc. Is better than someone else doing it
Nice!
It's good, but I don't think it's anything special.
I think in combo with some of the other workflows over on the platform, it's got something unique going on. I'll say, the one thing I've found if pretty good at is with composition. Like, it's not doing a lot of "character staring at camera"-- which is refreshing.
Tested it on Replicate. Prefer Flux1.1 but V3 has it's use cases.
Unless it's open source i'm not interested :0 Still vector export is a cool feature
Long Live Firefox😁
link for ai?
Interesting
for sure-- check it out and get those free credits!
How we can make money from these tools?
The red panda images are clearly squished. Maybe they're compositionally different from what we tend to get out of models these days, but that's it. Different, not better. In fact, clearly worse. Lots of moshed details. Overfit on blur. There doesn't seem to be a negative prompt option. Not trainable. Not local. Trash taste, trash ideas.
recraft is garbage compared to flux. just tested it out. very low quality by comparison. the images look ok, sometimes, but the prompt adherence is super limited. i couldnt even get it to do a nice graphic novel style image of a bear fighting a ninja. the realism images looked ok but it could only do so much. everything ultimately was quite bland and it could only pick up a few elements of the promt. i got a ninja in a blue outfit with a red scarf, but the bear simply wouldnt have samurai armor at all. it got grass and a sunset, but it couldnt get the style even when telling it marvel comics, graphic novel cover art, etc. it wouldnt take angle cues really, couldnt make anything dynamic, couldnt really take motion prompts very well. flux hears you and tries its best to give it to you.
Recraft is not free!
I'll say AI Platform Free. 25 gens a day is actually pretty good. So, no- it isn't open source, but free enough to continuously play with.
Let me guess. It's not as good as MJ by a longshot, and it's only free temporarily. Now let me watch the video and I will admit if I was wrong.
I was right. Where MJ shines that these others don't is going further out on a limb in creative areas, not just sticking to cinematic realism or established illustrational or cartoon styles. And the "Free" image generator is only limited free generations per day, and there's a paid subscription.
well, yeah-- besides open source, there is no free lunch in AI. But, 25 gens a day is fairly generous. To note, I know it's a sensational title, but the metrics from the leaderboards actually DO say that users preferred the Red Panda outputs more than MJ.
That said, I'm a die-hard MJ fan as well. They've been the one consistent AI Image generator in my toolbelt since day one.
This isn't free
free...beats nothing
No thank you, anythings as a AI as a service sucks. No Lora's and I imagine its extremely censored and political correct like the others?
Not really some advance, it doesnt follow prompt as needed, faces are ugly, if you ask for beautiful woman, it generates a huge "tank" looking creatures.... Also it doesnt allow some simple words categorizing it as NSFW