I appreciate your channel more and more. Most other channels about AIs are rushing to bash Sam Altman / OpenAI for clicks and drama, while you produce far more useful and informative videos. Truly, kudos to you, Sir.
appreciate that so much!! Yeah, I'm trying to stay off the drama side and concentrate on the cool artistic and tech uses of this stuff. There's enough Sam commentary-- no need for another "hot" take!
Thank you, Tim. For helping us laymen keep up to date on developments in this technology. Let's say it will be here soon enough. In the near future, I can already see independent filmmakers making marvelous and exciting films with this.
I love following the white paper stuff-- it really is like having a little telescope into the near future. I ended up cutting it from the video, but I had a little bit in there about how we first saw Video Inpainting as a whitepaper like, 6 months before we saw the release. I even made a video about it---somewhere in the archives of this channel. At some point, three to six months from now, we'll see one or more of these features in an Adobe product I'm sure!
Ah, bless man!! Yeah-- the current rage is OpenAI/Sam Drama-- I just don't want to go down that road! Rather focus on the cool stuff coming up in the future and focus on the positive work these teams are doing to get it into the hands of creative folks!
Summarizing new research in AI dealing with methods for media generation and editing is very useful! I am not aware of anybody else on TH-cam doing it. Many thanks !!
The devopers of AI software need to integrate into pre-existing CG pipelines. ie a previz animation in Blender has objects where the shaders imitate the look and motion of materials and forms. I have seen Mandala etc...but that was a year ago. . Control of tiny elements is essential in vfx. Harnessing Maya and Blender as the primary interface through a Comfyui plugin is IMO the best way forward for professional use.
Agreed on the pro side. I think there seems to be a bit of waffling right now, as everyone is trying to figure out where to put their eggs. For GenPop use, 3D UIs are too confusing-- BUT, obviously you get a huge userbase. Ultimate control requires a more complex UI, with preexisting knowledge of 3D toolsets. More power, but a smaller group of users. From what I've seen, the attempts to hybrid the two result in a layout that is still confusing to the average user, and just frustrates the power users. That said, I'm sure someone will crack it soon enough.
How about animating the character looking in a specific direction? Is there a tool that lets us use arrows to choose the direction of the character’s gaze? Can we move the eyes without moving the head? Sometimes, I might want to keep the character’s head still, or maybe only give it a subtle movement, while making more noticeable changes to the eyes & facial expression. I just wish I had more control to accurately modify the little details that I want portrayed in the animation.
I see that one as being the "it" thing fairly soon. Pandora looks fantastic, but only if you have control over the timing of the actions. Point Drag seems to allow for that.
No more rapid quantum leaps in GAI like last year where it seemed like the way things were moving monthly we'd be much further along by now. In relation to sites like Pika and Runway in particular, they haven't done a whole lot in the last 6 months, or even announced anything forthcoming. I'm assuming they are waiting until the end of the year at the earliest
It does! While SORA/VEO and all of the other next generation AI Video models look really exciting, this is the first one I've seen that is looking at the control we'll also soon have!
@@TheoreticallyMedia I think there's also a lot of feature expansion potential to this tech as well........such as revealing intentionally hidden elements behind the subjects that are given this motion that we don't initially see but are revealed with a special command as just 1 of many possibilities.
@TheoreticallyMedia Ah, okay, that makes sense........ that reminds me of my long standing biggest criticism of Midjourney's web interface is how archaicly Anti-intuitive and Anti-User Friendly it is. Like there's no reason why directly below the prompt bar there isn't an intuitive button menu in which you can quickly and easily on the fly set various parameters that currently we're forced to either prompt into the prompt text box or access a cumbersome menu off to the side. 🙄
Ah man, thanks so much!! Lots of really cool stuff on the horizon! We had a bit of a slump at the start of the year, but like the kids say: "We're so back"
I know, there’s so many of these- like Boximator as well, that just kind of vanish. My thought is that they get bought up by someone like Adobe, and then just get lost in a warehouse/dev hell…
Autodesk and the Foundry could merge and buy Stable Diffusion to create a complete professional development platform orimary based in synthetic training from Arnold, Nuke and video ?
Good use of the back picture frame behind you as a "classroom board" to show examples, references, b-roll etc. I know it's probably more work on your end, but its worth it. So we should call you Professor Tim now 😂
Yeah, I stupidly tried to fix it manually-- I really should have just leaned on AI...haha. Literally my dumb mistake for not pushing the Record button. As good as all our tech gets, we still have human errors!
I know right? That's something that I think we'll be seeing a LOT more of this year. There has been a lot of talk about creating 3D scenes with Midjourney, Leonardo, SD etc-- I think this is the first time I've seen that idea put into practice. And what's cool is, this is just v.1!
I'm waiting on a camera control model for generative ai with an actual virtual camera like they have 3D modelers. Something where you can set the motion with stick figures then apply that motion to the generated video.
I think if you look at 3dSceneIt and the Camera Control feature, and kind of mash them together with a few version iterations, you've got exactly that. And...I think we'll be seeing that within the year! (fingers crossed!)
Since VASA-1 is not open sourced, yesterday I was eager to try V-Express on a 3090TI. I've used ComfyUI instead of the script. Installation took a while very probably I'm not that familiar with extending ComfyUI. It asked me to install sd-vae and wav2vec. I was able to get the same result as the demo. My main concern is in the speed. Either I've misconfigured or it takes 20 minutes for 15 seconds of audio. Consumed 10GB of VRAM and ~450W. GPU went to 91C. MS's VASA-1 demoed to be real-time like a tiny efficient app running in the corner. I wish V-Express was similar.
That’s awesome to hear! I haven’t seen benchmarks on anyone else running it locally, so I can’t say if that long run time (and jeez!) is normal- but, it does sound like something is off. Might be a misconfigure, might be sloppy code. On the plus side: it works! On the downside, it just needs to be optimized now!
I have been working on a AI music video for a few days now. and for free users krea ai is beyond supiror then runway and haiper and so on... Faces are really stable and turn and stay consistant.
Thanks so much! And if you hadn't seen some of them, I consider it a real success! That's always the tough part with TH-cam-- slight lag from the newest stuff! On the V-Express front-- not sure, I looked through the documentation, but didn't see any requirements. Hopefully it isn't TOO weighty!
Sammmmmmme. The insulation/drywall is finishing up. Next they have painting and some lights, and a new ceiling fan...so...maybe another week? I can't wait to get back to Studio A!
I appreciate your channel more and more. Most other channels about AIs are rushing to bash Sam Altman / OpenAI for clicks and drama, while you produce far more useful and informative videos. Truly, kudos to you, Sir.
appreciate that so much!! Yeah, I'm trying to stay off the drama side and concentrate on the cool artistic and tech uses of this stuff. There's enough Sam commentary-- no need for another "hot" take!
Thank you, Tim. For helping us laymen keep up to date on developments in this technology.
Let's say it will be here soon enough.
In the near future, I can already see independent filmmakers making marvelous and exciting films with this.
I love following the white paper stuff-- it really is like having a little telescope into the near future. I ended up cutting it from the video, but I had a little bit in there about how we first saw Video Inpainting as a whitepaper like, 6 months before we saw the release. I even made a video about it---somewhere in the archives of this channel.
At some point, three to six months from now, we'll see one or more of these features in an Adobe product I'm sure!
Great video bro thanks for not piling in with the AI parrot influencers on the same stuff and always grabbing the best new stuff out. You are seen 🙏🏾
10000% agree with this comment
Ah, bless man!! Yeah-- the current rage is OpenAI/Sam Drama-- I just don't want to go down that road! Rather focus on the cool stuff coming up in the future and focus on the positive work these teams are doing to get it into the hands of creative folks!
Summarizing new research in AI dealing with methods for media generation and editing is very useful! I am not aware of anybody else on TH-cam doing it. Many thanks !!
The devopers of AI software need to integrate into pre-existing CG pipelines.
ie
a previz animation in Blender has objects where the shaders imitate the look and motion of materials and forms.
I have seen Mandala etc...but that was a year ago.
.
Control of tiny elements is essential in vfx.
Harnessing Maya and Blender as the primary interface through a Comfyui plugin is IMO the best way forward for professional use.
Agreed on the pro side. I think there seems to be a bit of waffling right now, as everyone is trying to figure out where to put their eggs. For GenPop use, 3D UIs are too confusing-- BUT, obviously you get a huge userbase.
Ultimate control requires a more complex UI, with preexisting knowledge of 3D toolsets. More power, but a smaller group of users. From what I've seen, the attempts to hybrid the two result in a layout that is still confusing to the average user, and just frustrates the power users.
That said, I'm sure someone will crack it soon enough.
How about animating the character looking in a specific direction? Is there a tool that lets us use arrows to choose the direction of the character’s gaze? Can we move the eyes without moving the head? Sometimes, I might want to keep the character’s head still, or maybe only give it a subtle movement, while making more noticeable changes to the eyes & facial expression. I just wish I had more control to accurately modify the little details that I want portrayed in the animation.
Ha ha. "Release the hounds!" Wow!
Haha. It was a toss up between “Hounds”‘and “Kraken”
Point drag is a very interesting stuff 👀
I see that one as being the "it" thing fairly soon. Pandora looks fantastic, but only if you have control over the timing of the actions. Point Drag seems to allow for that.
No more rapid quantum leaps in GAI like last year where it seemed like the way things were moving monthly we'd be much further along by now. In relation to sites like Pika and Runway in particular, they haven't done a whole lot in the last 6 months, or even announced anything forthcoming. I'm assuming they are waiting until the end of the year at the earliest
Needed this! I’m creating some stuff today! Wish me luck
Pandora looks promising 🐱
It does! While SORA/VEO and all of the other next generation AI Video models look really exciting, this is the first one I've seen that is looking at the control we'll also soon have!
5:43
My favorite technology of this entire video 👍 👍 👍
That one is SO cool. and I think we'll be seeing that integrated into MJ/Leo/etc pretty soon. By the end of the year, for sure...but hopefully sooner!
@@TheoreticallyMedia
I think there's also a lot of feature expansion potential to this tech as well........such as revealing intentionally hidden elements behind the subjects that are given this motion that we don't initially see but are revealed with a special command as just 1 of many possibilities.
9:13
A skilled prompter using Dall-E 3, MJ, or even SD to some extent can already achieve most or all of those camera settings attributes.
oh, totally. I just want the sliders in an advanced setting menu!!
@TheoreticallyMedia
Ah, okay, that makes sense........ that reminds me of my long standing biggest criticism of Midjourney's web interface is how archaicly Anti-intuitive and Anti-User Friendly it is.
Like there's no reason why directly below the prompt bar there isn't an intuitive button menu in which you can quickly and easily on the fly set various parameters that currently we're forced to either prompt into the prompt text box or access a cumbersome menu off to the side.
🙄
HE'S BACK!!!
And I don't know if you saw: But it's a DOUBLE today!
Still the best AI video monitoring!
Ah man, thanks so much!! Lots of really cool stuff on the horizon! We had a bit of a slump at the start of the year, but like the kids say: "We're so back"
This made me wonder what happened to dragaan.
I know, there’s so many of these- like Boximator as well, that just kind of vanish.
My thought is that they get bought up by someone like Adobe, and then just get lost in a warehouse/dev hell…
@@TheoreticallyMedia Would not surprise me at all tech giants like Google, Microsoft etc. Have been doing it for years.
its close but its not better than emo or vasa
Yeah, I’d say competitive. But, I appreciate the fact they went right after Vasa, since MS was like “look how cool this is, but you can’t have it!”
Great, as always, quality information, great delivery, not overstated, thanks!
Thank you so much! I know...no SHOCKING in the title!
Autodesk and the Foundry could merge and buy Stable Diffusion to create a complete professional development platform orimary based in synthetic training from Arnold, Nuke and video ?
Where is chatgpt new voice model, when available
Haven't seen it yet. My guess is that it debuts at WWDC as the voice of new Siri.
@@TheoreticallyMedia but i am Android users 😭
What if I email Microsoft and say that if they don't give me access to VASA, I will use a cracked version of Windows!
Will it work!!? 🤔
Haha, check with Bill Gates! I'm sure he's still got some pull there!
Good use of the back picture frame behind you as a "classroom board" to show examples, references, b-roll etc. I know it's probably more work on your end, but its worth it.
So we should call you Professor Tim now 😂
For some reason your audio broke in the middle. Put it into adobe audio and enhancer and it fixes everything
Yeah, I stupidly tried to fix it manually-- I really should have just leaned on AI...haha. Literally my dumb mistake for not pushing the Record button. As good as all our tech gets, we still have human errors!
😮
By any chance can I get VASA to use!!?
Nope. Microsoft has decided that one is under lock and key sadly...
Gaussian Splat video = wow!
I know right? That's something that I think we'll be seeing a LOT more of this year. There has been a lot of talk about creating 3D scenes with Midjourney, Leonardo, SD etc-- I think this is the first time I've seen that idea put into practice. And what's cool is, this is just v.1!
I'm waiting on a camera control model for generative ai with an actual virtual camera like they have 3D modelers. Something where you can set the motion with stick figures then apply that motion to the generated video.
I think if you look at 3dSceneIt and the Camera Control feature, and kind of mash them together with a few version iterations, you've got exactly that. And...I think we'll be seeing that within the year! (fingers crossed!)
@@TheoreticallyMedia I can’t wait! It seems we’re in a holding pattern waiting on Sora to drop.
Since VASA-1 is not open sourced, yesterday I was eager to try V-Express on a 3090TI. I've used ComfyUI instead of the script. Installation took a while very probably I'm not that familiar with extending ComfyUI. It asked me to install sd-vae and wav2vec. I was able to get the same result as the demo.
My main concern is in the speed. Either I've misconfigured or it takes 20 minutes for 15 seconds of audio. Consumed 10GB of VRAM and ~450W. GPU went to 91C. MS's VASA-1 demoed to be real-time like a tiny efficient app running in the corner. I wish V-Express was similar.
That’s awesome to hear! I haven’t seen benchmarks on anyone else running it locally, so I can’t say if that long run time (and jeez!) is normal- but, it does sound like something is off.
Might be a misconfigure, might be sloppy code. On the plus side: it works! On the downside, it just needs to be optimized now!
I have been working on a AI music video for a few days now. and for free users krea ai is beyond supiror then runway and haiper and so on... Faces are really stable and turn and stay consistant.
That's awesome! Yeah, I'm really liking Krea as well! I have a few others in my back pocket to get to-- some cool stuff on the horizon!
The LipSync is getting pretty good.
Tim, not Hepburn, Princes Grace of Monaco
Is it?! man, here I've been thinking it was Audrey...I mean, apparently I have beauty blindness!
Is v express available on Google collab!!?
Really great video today, some of these I hadn't seen. Wondering on VRAM requirements for V Express. Going to need to start saving for that 5090 😑
Thanks so much! And if you hadn't seen some of them, I consider it a real success! That's always the tough part with TH-cam-- slight lag from the newest stuff!
On the V-Express front-- not sure, I looked through the documentation, but didn't see any requirements. Hopefully it isn't TOO weighty!
Fantastic video Tim... Great overview of new stuff... :)
Thanks so much!! This one was a lot of fun to put together! LOTS of cool stuff on the horizon!
Once again, thanks for all your work and exploration of the creAItive landscape. We are, in fact, almost there. Not too much longer.
The 35mm looks longer than the 85mm. Not sure many photographers will be impressed.
Tim. Thank you for making your videos. It is always very interesting and helpful.
Awesome this gonna be a good one thanks in advance good sir!
I gotta say, this one is pretty packed! Lots of good stuff in here! Enjoy!!
👋
Heya Louis!! I'm working backwards through your comments...haha...It's like Memento in comment form.
@@TheoreticallyMedia
😆 🤣
Thanks for the news Tim! I miss the guitars. ;)
Sammmmmmme. The insulation/drywall is finishing up. Next they have painting and some lights, and a new ceiling fan...so...maybe another week? I can't wait to get back to Studio A!
watch 11.10 IS ALL IS SAY
Grazie great info and great sense of humour 😁
Love the backgrounds👌🏻
Thanks so much! Had to so SOMETHING with that stupid blank wall!
Shouldn't be too much longer before I'm back in Studio A!
Always informative. Always hilarious. Thanks. :)
Much appreciated! And yeah man, THAT WAS NOT A WALLET!!
@@TheoreticallyMedia 😆
Thanks. May I ask how you write your video scripts? They're really enjoyable.