Definitely want to use it in the future. Ever since I saw these videos come out, I've been wondering if you could use it with a green screen? Meaning, have a green screen behind the subject (potential reasons maybe locations not available or to cut travel costs etc) and then use this AI tool to fill the rest. Let's say you want to have a scene with the Eiffel Tower in the background and use a green screen for that, but then use AI in PS to fill the rest. Any thoughts on that? Great video, thanks for the super simple explanation for a newbie like myself
I've been using this at work a little bit recently. A couple of tips for anyone who is going to try this: • Add a little noise or grain to the png when you put it on your timeline to help it blend in with your video footage a little more. All footage will have a little bit of chatter no mater how clean and well lit it is so this will help sell the effect if people start looking closely! you could even do noise reduction on your video segment and then add grain over the whole thing to make the noise/grain match even better. • Avoid trying to get it to create anything with words or symbols in it like a sign post or something. AI is still terrible at that and will give you some hilariously wonky results! • Be careful if you are shooting in natural environments. If there is a breeze moving branches or leaves around you, of water running or rippling, trying to expand that stuff will be a bit of a giveaway when the generated material isn't moving the same way.
Another tip: As of now, the resolution of generative fill is limited to 1024x1024px. If you select an area wider than this, photoshop will just upscale a 1024x1024px image to your selection, losing quality. To gain the best quality, select an area no more than 1024x1024px at a time, building the scene square by square.
@@JDRos Fstoppers and lots of other sources claim that the area that is being generated is 1024x1024 pixels. But I don't have any first hand information from Adobe, so if you have some information from Adobe directly, then you should probably trust them.
Absolutely incredible video brother! I love the idea of adding real estate above your subject for interviews if they aren't quite cropped properly and need to make adjustments. That right there could save you an entire re shoot!
It's neat and all. But why not film in 4k horizontal, and crop it to 1080 vertical for verticals? ;) do you gen fill in batches of ~500px, or don't you mind the upscaled fills?
I just had a great AI experience! I can see the AI in the thumbnail the way I learned to see wind turbulence (hang gliding) or see lens lengths from seeing a picture🙌 As a rule, I'd shoot horizontal and AI the vertical version when necessary. Even if we just crop a vertical out of an 1080 image we can also use AI to blow it up to 1920 vertical. There's a lot less area to invent, and the invented bits are less relevant, ie. a small patch of ground and sky, for example. Granted, inventing most of the frame is more impressive as a demo.
@@Mrlloydvideostrue, but the point is that with this method you have more control of your background...now you're not in a park, you're in a vast remote vista.
Great video - learnt alot - this is a game changer for me as I've moved countries and don't have gear or a studio anymore - this may allow me to create some "studio" to film in - cheers mate! BTW, loved your relaxed style on camera, really enjoyed watching you.
1) You don't need photoshop beta anymore, it's in the base app 2) no movement in the generated content ruins the illusion 3) it's better to just shoot square video at 4-8K and frame it so you're in the center for both crops, then you can just crop
Would be cool to see an old movie scene filmed in a narrow aspect ratio filled in with AI to become widescreen without cropping from the top and bottom
bro i knew somewhat of this concept already but underestimated how easy it is. you gave me such a breakthrough idea for my youtube video because of this, bless you
Great idea! I think the only thing that makes it look not so realistic is the too shallow depth of field. When you shoot vertically, you're closer to the camera and then when you widen the frame, the depth of field is much smaller than normal, and I think that's why it looks a bit unnatural.
Awesome tip about shooting vertical and then using content fill in post to create a 16:9 video. I'm definitely going to start using this in my videos. Thanks again!
Firstly, thanks for the video! Awesome results. One thing I would caution anyone about to try this out though - if you are shooting outside your lighting conditions will be constantly changing. The new 'mask' environment around your shot won't change though so you will see it straight away. Still, it's a great tool for setups when you can control the lighting and for short shots.
I don't know how I stumbled upon this video, but I am very glad that I did. NOTHING is real anymore. I had to subscribe and now going back to see what else you got for me. I'm hopeful that I can learn many things to help my social media presence along. Thanks!
You would be throwing away a lot of pixels that way, leaving a lower resolution product. Pixar did something similar with A Bug's Life in which they extended the frame in some scenes instead of cropping it for the then-still-prevelant 4:3 video release.
@@alphzoup Yet your way in order to get QUALITY pixels you’d need to only generate 1024x1024 since that’s all generative fill can do at a time! So AGAIN my way is the LOGICAL and QUALITY way to do it…sorry!
This could be good for inside background but for outside it doesn't feel reel because there's no movement in the background. Even though it's not a windy day trees would be moving.
you shouldn't even need to take you out of the png, with resolve, the black would be transparent either side of the clip, you could just pull in the png under the clip, only the edges will be visible around the clip, and the vertical will still be on top. Just a thought. Like the use of this though!
I was experimenting with this concept. But my still had no hole. I put the movie as a layer over the picture. You have nice tips and instructions. Thanks for that. BTW I was wondering how you blended in the cat at 2:49 ;)
It "might" be because you're trying to gen off a layer. You need to open the image rather than add it to an existing comp. Also, make sure your selection overlaps the image a bit.
Actually you can have movement...just need to incorporate After Effects to your workflow...tracking...done. But this whole thing still in development and people like you sharing their workflow really helps to shape it for the others.
If you are reading this comment, it is no coincidence. You have an immense power that lives within you. Everything is vibration, when it changes everything is transformed. I wish you that everything you dream makes it come true.
i like how you talk to us with clarity and complete transparency whereas others are too busy trying to sell us anything and everything. thanks a lot for being here and doing this, Cheers!
Pro tip- if you turn the camera ninety degrees before starting to film, you can skip ALL of the above work, and still be able to crop that into a vertical if you need to. You can move the camera, the background and everything.
Sometimes the video wasn’t planned to be used with other videos of a certain orientation so it can certainly be a handy thing. But otherwise, planning your shots if obviously the best thing to do.
Everyone talks about the “issues & problems” with the technique. 🤦🏾♂️ Damn just thank the creator and take from it what you can. I’ll say it! Thanks for spending your time creating this video and sharing your knowledge. Ignore the dweebs on here that criticize and don’t have any content of their own.
Great tip for interviews. It's wild that this is just a beta release. A lot of the results are great - some not so. Fingers crossed they'll add some video features (not requiring Premiere), at least something subtle like tree branches gently moving in the wind. That's the big tell right now. Also, I would love to see a less "professional" company (gimp?) do something like this because Adobe is crazy heavy on censorship. I tried to add legs to a character and got shut down based on the word "legs." Tried again without the keyword...and it drew legs LOL.
On the first glance you don't even notice it. The only thing that breaks the illusion is for example in the scene @7:30 the branches in the middle section are moving but the branches at the outer sections are not, if you don't look for it I'm not sure if you would notice that right away, it would be more of a unexplained feeling of something is wrong :D
No lo usaría 🙅♂ No hay vida en el fondo falso, no hay ningún movimiento que vaya acorde al video, y se nota en las sombras de los árboles el corte donde sólo tu video vertical tiene movimiento en las sombras del suelo: 8:08 Aparte para rectificar el color o agregar algún LUT ya no podrías hacer nada de eso pues son dos archivos completamente diferentes y el .PNG no soportaría una edición fluida.
I see it as a big win for creators just starting who do not have a studio as per say and only a small space to get talking head shots. At this point, even advanced creators renting studio spaces might as well just play smarter with the backgrounds for way less money overall. Not even mentioning where this might even be in the next few years ... Or imagine similar technology coming natively in Resolve for example ? This AI thing is going to get us in trouble at some point
Ok... Love this... but most importantly, I need new glasses and what you have are awesome... can I ask where they are from? Finding glasses for a bald & big headed guy like myself has been tough 🤣
The problem is, you cant walk around much x.x you are restrained to staying within that small frame you set up and then the audience will eventually noticed something is up. I also find it crazy that your doing a video that like every other content creator is also doing Dx seems like a video made because it was a easy one to produce