Your videos are always easy (and entertaining) to follow. I always thought I have several different personalities, will have to try this out. Did Runway not have the same tool?
This is part of the reason I wait for things to settle first before jumping in. It can save so much time and effort. Unless it's something I really, really want to use of course.
"Business up front, party in the back", hahaha, love it! That said, mullet wouldn't be totally wrong, would it? 😉 Thanks again for a great vid, I really dig the way you present information.
Haha, I'll say, I probably did have a mullet when I was a teenager. Doesn't every kid go through that weird mullet awkward phase? At least it didn't give me pimples and braces as well...
I mean, technically you CAN do it-- but, that's just video. I get what you're saying though. There are some interesting things you can do on that front, but it veers toward animation. I did one pretty cool workflow using Domo and Skyglass. I think that's under the thumbnail OP AI Filmmaking or something like that.
Hi Tim - this looks really cool. I have what I think should be a simpler problem but probably isn't 🤔 Is there a system that can produce two consistent characters in the same image? I just need images, not a video - yet 😊
Tim, your exploration of the Kling AI's face-swapping capabilities was really insightful, especially when you mentioned, 'This AI eggplant art doesn't seamlessly repeat, so let's make one that does.' Have you considered integrating these AI-generated visuals with virtual reality or augmented reality applications? Such an integration could offer interactive and immersive experiences using the patterns you've created, potentially opening up new markets and applications.
Awesome video!! That Guitar Face is the wildest I’ve seen, haha. Thanks so much for including my examples-some of them are truly hilarious! And yep, I brought the whole family along to cheer me on as I went for the pizza record! XD You’re the best!
All the face swappers I have seen have the same problem they can't handle it when something crosses the face, like a hand. They glitch out when the entire face is not visible, because it interprets it as "no face". Also extreme angles, like a character turning 180 degrees, back towards camera, the face glitches when they are almost turned. They also flatten emotion, for example if the source image has eyes open, it doesn't work well if the video should be about someone with eyes closed. It wants to keep the eyes open, like in the source image
Yeah, that's a general Text to Video problem. Oddly, back in 2021-- they pulled it off in that Deepfake Tom Cruise video. Granted, different method there (V2V), but still. We'll get there, I'm sure-- but the permanence issue is a valid one.
1. You playing the saxophone was incredibly good. 👍 2. As a PSA to anyone who reads this, Kling does _not_ make it easy to unsubscribe. Trust me. :) You're welcome. :) Thanks for the video, Tim.
I actually played sax in high school. Now, I didn't say I played it well! Haha, might be a good thing AI video doesn't yet have sound! Good to note on Kling unsubs. I haven't actually tried yet, I got the annual and haven't come around yet. Although, I likely will resub. I can try to reach out to them to see if the issue has been resolved. I heard some rumblings about it in their discord.
@@TheoreticallyMedia They are the least responsive of all of the AI platforms, runway, hailuo - also they are the only ones who don't offer unlimited plans, it's a shame cause in some cases i think their output is marginally better and more consistent than the competition, unless it's more cinematic stuff, hailuo always wins that race :)
No video to video yet with your custom model? So you can't insert characters into other existing scenes? Seems like it's just prompts from what you said at the end. I'm sure they'll have image to video and video to video soon enough with your character.
Is there anything stopping us from training it on AI generated 10sec clips (for our own characters)? Do you know if you can add multiple characters in one video? Finally, does it pick up the clothes you are wearing or just your face?
Just face from what I've seen thus far. Although, if I prompt for a black tee shirt (which is what I was wearing in the training footage) it seemed to do quite well. But, that's also an easy prompt. In the video I go over multiple characters. Spoliers: Sorta. And lastly on the synthetic characters....maybe? I mean- you still need an additional 10 to 30 clips of them in action...That's something I'll explore in a follow up.
Maybe Act-One would be good for making videos off diff angles/expressions for Kling? That way you can start with just one image of the character, and could re-use the driver recorded videos for diff characters.
Plot Twist: Tim IS THE MAN IN THE BLUE BUSINESS SUIT - tun, tun, tun! I paused on Guitar Face composite. Hilarious and awesome. Downside: The Chinese now have your face data. Your clone is being constructed as we speak. Thanks for the update and another fun tute and demo! Btw, I liked you better with a mullet ala Bowie Labyrinth era.
@@PixelsVerwisselaar its bizzare that some companies dont offer free version. ??? I really wanna try mangnific upscaler but its $30 ??? Say if I dont like it? I'm not paying 30 bucks to try something I find pointless?. Freepik own them and that is outstanding. It's off the chart what they offer. Flux, mystic (image generator). Sketch to img- re touch (now with flux model), expand, upscaler, plus millions of stock images and videos. My 2 goto's are MJ- freepik. Lol. Still trying to find the ultimate vid gen. I see pixverse 3.0 just dropped . Looks very consistent. Woo hoo.
I'm trying to keep the videos under 15 minutes-- I know the trend is to do 30min videos, but as my own editor...I mean, I just can't stand to look at myself that long! (haha)
Thanks for the videos.In runway there is already video to video and act one, but the back of the head is coherent without using Modelli that take too long like lora etc, thanks.
Haven't seen the video yet, but can't wait to try after Kling finishes generating videos after 3 days (and if you're lucky enough to not get a 'failed' message after those days)
Spend $10. My videos are 30 minutes to an hour Nice thing about Kling if the month runs out, your credits don't go away. Downside is you just have to pay another $10 if you don't want a watermark. You still get 66 credits a day free, so you get 1 10 second video a day for only 4 credits.
Haha, I noticed that too, but I figured it was too nerdy for gen-pop! I actually played one of those 8 strings the other day. Nope'd out super quick! I've got enough trouble with 6!
Impressive Tim! It looks so real... Sadly, this technique won't help me to create a virtual rockstar, that (for now) only exist in pictures. By the way, about eating pizza the wrong way... look up Trumps old Pizza Hut ad... is he an alien? Who eats pizza like that?
Haha. I talk about that a bit in the video. I don’t think I have it enough emotional range in the training data. I’ll probably re-do it at some point and just rename this one to “Angry Tim!”
look closely there is a small amount of bleeding onto women too, the eyes in particular, I would imagine this could be offset similar to how correct captioning works in Lora/Check point training, obviously that will be something they update themselves so I guess its just a matter of time. Next will be Voice cloning I reckon then arguably it will be length of video increase
Yeah, I don't think they released it yet. My guess? We'll see it tomorrow (after the election...that seems to be when everyone is finally releasing everything)
Personally, I can't tell the difference between you and that version...but since you know your face more than anyone else, how much do you feel it closely matched how you look?
I can finally make a movie of my favorite actress who retired early in her career and doesn't want to do movies even if they offered her more roles... Phoebe Cates.
Haha, I know, right? It's insane how accurate it was-- and I didn't even train it with my shirt off! haha...actually, those results would probably get me banned.
Haha, AI Me is cranky about being AI! To be honest, I probably didn't train enough "happy" and "smiling"-- I bring that up in the video-- So, if you do train yourself: make sure to smile a bit at the camera! (actually, smile a LOT at the camera!)
Are you paying for Kling? When I tried it out, it took many many hours to make a video from image and prompt. I also suspect that the video generation is being processed on my computer.
It seems like they're getting better with it. Did you know that original Kling social media launch wasn't even them? Haha-- Yeah, they started as a mess. It feels like they brought someone on to handle the X side of things at least.
@@TheoreticallyMedia I hope so! There was a lot of negative buzz around their practices which is a shame cause the tools are okay. So the channels that were removed weren’t run by them either?
It says something that you filmed yourself walking down the street while cosplaying as the man in a blue suit....yet, you use AI generation for the to camera, studio stuff. Not sure what it says but it's something. 😁
We talk about not sharing our private information, but BOY straight up uploading your identity willingly into something nobody knows how well it is protected and could absolutely be used against you, it's beyond crazy to me..
Plus, Kling is a Chinese platform. Wasn’t the so called issue with Chinese platforms that they store data from Americans? I guess it’s ok when you voluntarily just give them your face. 🤣
Now I can watch my favourite non Schwarzenegger movies with Schwarzenegger as the main.
or...hear me out: Twins, but both characters are Arnold.
😂🤣@@TheoreticallyMedia
Or Schwarzenegger movies with yourself as main. 🙂
@@derpopanz4502 Lol
Your videos are always easy (and entertaining) to follow. I always thought I have several different personalities, will have to try this out. Did Runway not have the same tool?
Two things.. Im waiting for video-to-video and an unlimited plan. If we get these two, I'm with Kling forever. PERIOD.
Couldn't agree more.
Yeah we need another solid video to video provider. Runway is waaay too expensive
@@shamz_ai im also waiting for Meta's Movie Gen. Looks interesting
If it brings greater benefits than the amount of money you spend, then boldly use it
I'd hate for Hollywood to take you away from RUclips (and AI videos) but you do have a face suitable for movies. Distinctive and easy to remember.
Haha ,worked there for a little over 15 years. They had their shot! RUclips is my home now!
It's honestly getting a little hard to keep up with these advances. I mean it's happening daily. What a time to be alive.
This is part of the reason I wait for things to settle first before jumping in. It can save so much time and effort. Unless it's something I really, really want to use of course.
@@Elwaves2925 Man, I'm not even sure when it will settle down. It's going to be quite a rush for the next year at least!
"Business up front, party in the back", hahaha, love it! That said, mullet wouldn't be totally wrong, would it? 😉 Thanks again for a great vid, I really dig the way you present information.
Haha, I'll say, I probably did have a mullet when I was a teenager. Doesn't every kid go through that weird mullet awkward phase?
At least it didn't give me pimples and braces as well...
I would never want to train AI with my own face. I’d much rather work with fictional consistent characters that I create.
still, video to video with consistent characters will be the game changer
I mean, technically you CAN do it-- but, that's just video. I get what you're saying though. There are some interesting things you can do on that front, but it veers toward animation. I did one pretty cool workflow using Domo and Skyglass. I think that's under the thumbnail OP AI Filmmaking or something like that.
Hi Tim - this looks really cool. I have what I think should be a simpler problem but probably isn't 🤔
Is there a system that can produce two consistent characters in the same image? I just need images, not a video - yet 😊
Just edit it yourself
Tim, your exploration of the Kling AI's face-swapping capabilities was really insightful, especially when you mentioned, 'This AI eggplant art doesn't seamlessly repeat, so let's make one that does.' Have you considered integrating these AI-generated visuals with virtual reality or augmented reality applications? Such an integration could offer interactive and immersive experiences using the patterns you've created, potentially opening up new markets and applications.
Awesome video!! That Guitar Face is the wildest I’ve seen, haha.
Thanks so much for including my examples-some of them are truly hilarious! And yep, I brought the whole family along to cheer me on as I went for the pizza record! XD
You’re the best!
Shame you coudn't make a 30 second video. Loved to have seen you eat it all
@@yoagcur😂😂😂
All the face swappers I have seen have the same problem they can't handle it when something crosses the face, like a hand. They glitch out when the entire face is not visible, because it interprets it as "no face". Also extreme angles, like a character turning 180 degrees, back towards camera, the face glitches when they are almost turned. They also flatten emotion, for example if the source image has eyes open, it doesn't work well if the video should be about someone with eyes closed. It wants to keep the eyes open, like in the source image
Yeah, that's a general Text to Video problem. Oddly, back in 2021-- they pulled it off in that Deepfake Tom Cruise video. Granted, different method there (V2V), but still.
We'll get there, I'm sure-- but the permanence issue is a valid one.
This is not a face swap, but a LORA weight applied to the general model. So obstructions should be handled better than a faceswap.
1:19
😆 🤣 😂 I love your hilarious commentaries in your vids........ don't ever stop that! 💜
1. You playing the saxophone was incredibly good. 👍
2. As a PSA to anyone who reads this, Kling does _not_ make it easy to unsubscribe. Trust me. :) You're welcome. :)
Thanks for the video, Tim.
I actually played sax in high school. Now, I didn't say I played it well! Haha, might be a good thing AI video doesn't yet have sound!
Good to note on Kling unsubs. I haven't actually tried yet, I got the annual and haven't come around yet. Although, I likely will resub. I can try to reach out to them to see if the issue has been resolved. I heard some rumblings about it in their discord.
@@TheoreticallyMedia They are the least responsive of all of the AI platforms, runway, hailuo - also they are the only ones who don't offer unlimited plans, it's a shame cause in some cases i think their output is marginally better and more consistent than the competition, unless it's more cinematic stuff, hailuo always wins that race :)
Once again cheers for the informative video 🎉
This technology is really evolving fast, Truly amazing.
Great selection of guitar faces, some of the best.
Da man in blue suit looking soooo serious 😁
Tx again amigo 🙃
Haha, AI Tim is all: "Why I gotta be the AI? Why can't YOU be the AI?"
Still, he's got some sweet dance moves.
@@TheoreticallyMedia
EXCELLENCE❤❤❤😍I can't wait to see this
It is sooooo much fun. A lot of work upfront with the training, but I've been laughing almost non-stop at AI me videos.
No video to video yet with your custom model? So you can't insert characters into other existing scenes? Seems like it's just prompts from what you said at the end. I'm sure they'll have image to video and video to video soon enough with your character.
This sounds great! But did Kling work out the freeze up issues that many of us had? I could never get Kling to work.
Is there anything stopping us from training it on AI generated 10sec clips (for our own characters)? Do you know if you can add multiple characters in one video? Finally, does it pick up the clothes you are wearing or just your face?
Just face from what I've seen thus far. Although, if I prompt for a black tee shirt (which is what I was wearing in the training footage) it seemed to do quite well. But, that's also an easy prompt.
In the video I go over multiple characters. Spoliers: Sorta.
And lastly on the synthetic characters....maybe? I mean- you still need an additional 10 to 30 clips of them in action...That's something I'll explore in a follow up.
Maybe Act-One would be good for making videos off diff angles/expressions for Kling? That way you can start with just one image of the character, and could re-use the driver recorded videos for diff characters.
I remember finding you when you have about 20k subscribers maybe less you’ve been growing slowly and steadily it’s been an interesting journey
Plot Twist: Tim IS THE MAN IN THE BLUE BUSINESS SUIT - tun, tun, tun! I paused on Guitar Face composite. Hilarious and awesome. Downside: The Chinese now have your face data. Your clone is being constructed as we speak. Thanks for the update and another fun tute and demo! Btw, I liked you better with a mullet ala Bowie Labyrinth era.
And they killed the free version
the good old days will vanish, soon openAI will let you play with chatGPT 3.5 if you were lucky...
There was no free version. I had 3 vids on 99% for 3 days 😂
@@armondtanz Good old times. I was lucky back then 😆
@@PixelsVerwisselaar its bizzare that some companies dont offer free version. ???
I really wanna try mangnific upscaler but its $30 ??? Say if I dont like it? I'm not paying 30 bucks to try something I find pointless?.
Freepik own them and that is outstanding. It's off the chart what they offer.
Flux, mystic (image generator). Sketch to img- re touch (now with flux model), expand, upscaler, plus millions of stock images and videos.
My 2 goto's are MJ- freepik.
Lol. Still trying to find the ultimate vid gen. I see pixverse 3.0 just dropped . Looks very consistent. Woo hoo.
nice work Tim!
Are Kling allowed for commercial use?
This is going to be tons of fun :D
This is impressive from Kling and I love how you always explain fast but in detail .
I'm trying to keep the videos under 15 minutes-- I know the trend is to do 30min videos, but as my own editor...I mean, I just can't stand to look at myself that long! (haha)
Thanks for the videos.In runway there is already video to video and act one, but the back of the head is coherent without using Modelli that take too long like lora etc, thanks.
@@TheoreticallyMedia hahaha
Happy to see you as the man in the suit
Dancing Tim is hilarious 😂
you should see REAL Tim dancing! It's just as terrible! Although, less morphing haha
5:00 You used a word I have never heard. Gimble. Is that like a stabilizer?
Yup! Exactly that!
Friends don’t let friends skip leg day 🏋️
We are way closer to fully AI Generated films than anyone can expect and I am super excited for that
Haven't seen the video yet, but can't wait to try after Kling finishes generating videos after 3 days (and if you're lucky enough to not get a 'failed' message after those days)
Spend $10. My videos are 30 minutes to an hour Nice thing about Kling if the month runs out, your credits don't go away. Downside is you just have to pay another $10 if you don't want a watermark. You still get 66 credits a day free, so you get 1 10 second video a day for only 4 credits.
Great video especially the walkthrough of the various hoops to jump through. I am curious to see how it handles a female subject.
I thought about wearing a wig to train for another character…
Ha!
@TheoreticallyMedia 1970s disco Tim intensifies!
5:56
😆 🤣 😂 But pray-tell, why is she crying....... what didn't you do???
🤪
Imma wait until you can just upload 1 picture, and they auto train all the steps for me...
guitar face 10/10, 7 string guitar with floating control knobs 10/10
Haha, I noticed that too, but I figured it was too nerdy for gen-pop!
I actually played one of those 8 strings the other day. Nope'd out super quick! I've got enough trouble with 6!
@@TheoreticallyMedia TBH I didn't know there were 8 strings, same, 6 is a lot already
The model bleed would be great for Clone Troopers.
Cody and Rex finally get their own movie! I'm into it!
Oh I am SO digging into all this! I have like 5000 credits w Pro so time to get busy!
Impressive Tim! It looks so real... Sadly, this technique won't help me to create a virtual rockstar, that (for now) only exist in pictures. By the way, about eating pizza the wrong way... look up Trumps old Pizza Hut ad... is he an alien? Who eats pizza like that?
what if you train the model using videos generated by kling based on those images?
It's the My Model feature will be also available on IMAGE to VIDEO?
I can’t wait for the day we’ll have multiple consistent characters in the same scene.
Hi, is it still now possible to access the Kling Beta? How did you avail it?
Should be releasing this week.
TM your Ai doppelgangerino looks so angry in all shots 😮😁 why is that?
Haha. I talk about that a bit in the video. I don’t think I have it enough emotional range in the training data. I’ll probably re-do it at some point and just rename this one to “Angry Tim!”
@@TheoreticallyMediahaha 😂💯👏
you move like an AI on the face video sample you uploaded
You are the man for this
The overuse of the word "actually" is mind bending.
I love Tim in the blue business suit
I give up. Which is the yeti?
HEY OHHHHH!!
Iron Man Vs The Intergalactic Space Hero
you're finally the man in the blue business suit! LOL
This is incredible!!!!!
Fun fact: If you click on a comment twice, you’ll like it, but if you click three times, you’ll reply to it.
Why I don't have this already?? I want it!
Cool!
It's pretty great!
Which one is best for image to video for Consistent Character Model I created in Midjourney thanks
I have Kling Pro lets go, i just need it to get 1 Minute outputs
08:35 Slapped her soul out _love your content, I just had to share my laugh_
Haha, I didn't even notice that! I think I was too busy looking at AI-Me's insane face! Thank so much! Great laughs all around!
@TheoreticallyMedia I mean, I shouldn't have noticed it either if I wasn't focusing on the women 😭🤦♂️
You put a Clint Eastwood face in all your faceswaps 😅🤣😂
Haha, AI me has the natural Eastwood blue steel gaze. Real me can't pull that off.
So lovely ❤❤❤❤
You gotta fold the pizza to eat it. But starting from the back unfolded is blasphemy.
+1 to the fold!! And yeah, eating from the back in insane-- that's where all that wonderful orange grease drips out! haha
look closely there is a small amount of bleeding onto women too, the eyes in particular, I would imagine this could be offset similar to how correct captioning works in Lora/Check point training, obviously that will be something they update themselves so I guess its just a matter of time. Next will be Voice cloning I reckon then arguably it will be length of video increase
can you please post the url for model training, maybe it works with my pro account, which is not yet unlocked 🙂
Yeah, I don't think they released it yet. My guess? We'll see it tomorrow (after the election...that seems to be when everyone is finally releasing everything)
@@TheoreticallyMedia thank you for your answer and your not-answer 😉
What's his name -- Tech Holla? Any more info on him. He looks strikingly like a guy I met in prison last year.
Techhalla-- yeah, you can find his Twitter profile (I think where he posts most) in the description
Awesome job this is dope.. but this isn't characters. I thought you meant creating characters. But thank you, great work
Is it possible to do face swap and then do lip sync to create a music video?
👋 Impressive results
Personally, I can't tell the difference between you and that version...but since you know your face more than anyone else, how much do you feel it closely matched how you look?
I love my Osmo too!
How many credits will we have to pay for 1 model?
990. He said it in the video.
Does anyone know if there are any AI animating programs that will do clips of up to 30 seconds in length?
why i dont have this feature
How is it on copyright? **cough** askin for a friend.
07:49 Stop the cap bro 🤣🧢
I literally thought the same thing but didn't want to be offensive
Kling gave you a seven-string. Eat your heart out, John Petrucci
give it 2 years how much things will change
I can finally make a movie of my favorite actress who retired early in her career and doesn't want to do movies even if they offered her more roles... Phoebe Cates.
cool
It's a LOT of stupid fun!
0:58 ah I see what you did here. You slipped in a real video of yourself to try and test us.
Haha, I know, right? It's insane how accurate it was-- and I didn't even train it with my shirt off!
haha...actually, those results would probably get me banned.
why you're lookind always grumpy in any of these ai shots..?
Haha, AI Me is cranky about being AI! To be honest, I probably didn't train enough "happy" and "smiling"-- I bring that up in the video-- So, if you do train yourself: make sure to smile a bit at the camera! (actually, smile a LOT at the camera!)
AI will replace lot of workforce.... and here it comes... The Mullet AI
Haha, AI's real threat: The return of the Mullet! As it turns out, John Connor was a barber!
Has anyone explored Kling's TOS? Do they own what we make with it? Also, is it censored, or can we make naughty videos with it?
Think around Pg-13. And yeah, I'm pretty sure it's commercial use if you have a paid plan.
I guess it's right time for you to jump into acting 😊
Your character looks kind of strained. Like struggling.
Are you paying for Kling? When I tried it out, it took many many hours to make a video from image and prompt. I also suspect that the video generation is being processed on my computer.
payed tiers are a lot faster!, like, when they reach 99% they actually are almost ready
Can you use a 3d Model of fictional characters? Otherwise, there isn't a major use for this in making consistent character animations.
Staggering.
How about creating 2 models 1 of woman and 1 of man and creating videos using 2 of em.
Only thing about Kling I don't like is that it's potentially Chinese spyware......
-NO Mullet
Oh Kling. The company that removed all their social media channels cause they can’t deal with the hate on their terrible customer care.
It seems like they're getting better with it. Did you know that original Kling social media launch wasn't even them? Haha-- Yeah, they started as a mess. It feels like they brought someone on to handle the X side of things at least.
@@TheoreticallyMedia I hope so! There was a lot of negative buzz around their practices which is a shame cause the tools are okay. So the channels that were removed weren’t run by them either?
Now we can gender swap all the comic characters back to their original designs!
It says something that you filmed yourself walking down the street while cosplaying as the man in a blue suit....yet, you use AI generation for the to camera, studio stuff. Not sure what it says but it's something. 😁
Haha, all this stuff can sometimes be the snake eating its own tail-- but that's also what makes it so fun!
api cost is just way too much high!!
We talk about not sharing our private information, but BOY straight up uploading your identity willingly into something nobody knows how well it is protected and could absolutely be used against you, it's beyond crazy to me..
Plus, Kling is a Chinese platform. Wasn’t the so called issue with Chinese platforms that they store data from Americans? I guess it’s ok when you voluntarily just give them your face. 🤣
Honestly results not too convincing comsidering amount of prep needed
You could be Ryan Reynolds' dad
LoRa, for IA video
New ai tool melies hollywood level video generate just