VeeshanVault

I wanted Spang to see this so added his name to the title.

I made a quick example of WAN2.2's ability to use first/last frame to continue a video seamlessly between still images.

Discussed in detail here:

viewtopic.php?t=26681&start=100
I took these three images of Jenny Ortega since she's popular with season 2 of Wednesday out.

I left a super simple prompt : "A woman model posing" for all steps.
Click above image to see the animation

I had to convert to gif and optimize the shit out of it to get it to post on this forum so the quality isn't there.

The key thing to look at is how WAN2.2 figures out the movements between images. Look at her hair for example in first and second image. See how WAN2.2 moves her hair with the hand so it matches? In fact, WAN2.2 had to find a reason for her hair to fall in her face so the AI had her touch her hair. Same with 2nd to third image where she moves the hair out off her face. That's not prompted. I just said "A woman model posing". All the smooth zooming, etc is WAN2.2 AI doing its thing.

This is a simple 10 second example but that's the power of WAN2.2. You can take any photos that are semi related and make a video out of them.

Even with the goofy arm/hand positions of those images, WAN2.2 can figure out the proper motions to get to the next image pose.

I've made multiple 3-5 minute videos using various subjects. It really does bring still images to life in a very realistic way

click to see animation

To expand on what it can do. You can use a single image and take it any direction you want'

A man kisses a woman. The woman kisses a man. tongue kissing. The man holds her shoulder. A second man approaching from the right reaches out and touches the woman's hair with his hand.

Again, ridiculously optimized gif to get it to post here so quality is way better than this.

I used this image because she is wearing a goofy outfit. It's to demonstrate that WAN2.2 can handle odd looking things and still put it in motion.

In this case it followed instructions pretty well. They kissed. He did touch her shoulder (kinda, it's upper arm). The other dude did reach out and touch her hair.

I know I know. SHOCKED! APPALLED! The lesson is don't trust any images or videos ever again. You can start off with any image and have that person doing unspeakable things with anyone/situation etc. And it only takes a couple minutes do to it.

The news/gossip sites already sucked. Now you can be 100% confident not to trust anything instead of wondering about it.

Thankfully this WAN2.2 wasn't around during that crazy Aunt Jemima and Spang fling awhile back!

Some more examples.

The other amazing AI image model recently released is Flux Kontext. with that you can change whatever you want with an image. background, clothes, etc. zoom out, etc.
With this one, changed background to hotel room and zoomed out to show mini skirt:

The woman partially lifts up her shirt and wags her finger at the viewer with a seductive smile

This one was a weak effort but I guess shows some water physics. The bikini needed a regen.

The woman opens her mouth as she is sprayed with water from a hose in her face and mouth. The water splashes off her body and mouth

Anyway. There's a lot you can do working off a single image.

The real power is if you wanted or felt like it. You could take that image and have her do a modeling shoot. go to her hotel room, change into her bikini, then go out to the beach and make it all one long video sequence.

There is a learning curve and it can be steep. Once you know what you're doing, the actual images and videos and editing are quick and easy. And of course you need a 24GB VRAM card because otherwise it's not as fun due to all the waiting for things to generate.

I’ve seen more of your AI slop. Now what?

Spang wrote: ↑August 13, 2025, 9:40 am I’ve seen more of your AI slop. Now what?

Simply observe how it improves over time and maybe give any thespian friends you have a dollar because they'll be out of work sooner rather than later.

On an image posting site a few years back there were tons of negative comments any time AI images were posted. Today it's rare to see a negative comment because the AI images are better than what the artists can make and faster and the smart "artists" use AI as much as they can.

The people with no ability to project the rapid improvement of AI understandably whined about 6 fingered hands etc. instead of marveling that it could actually create an image at all. The bright people were noticing the improvements. Each rapidly developed improved AI models released. Now anomalies are rare. Hands are great even in AI videos with crazy movement/positioning.

You are like a Senior in high school hazing a freshmen when you hate on AI. The only reason you are is because you are a couple years older and it's tradition to do stupid things. That freshman will eventually be as good or better than you. Enjoy your senior moment!

AI haters are like religious freaks from the past. Every time science proved their religious beliefs wrong, they made up some excuse to integrate the scientific discoveries into their religion. Right now you are at the AI equivalent of believing the Sun revolves around the Earth.

AI will be smarter than humans in every way and so much faster. That especially includes "creativity".

Winnow wrote:give any thespian friends you have a dollar because they'll be out of work sooner rather than later.

I don’t know if you’re familiar with long-form improv, but it involves a group of actors who, after being given a simple prompt from the audience, create a scene from scratch. So that a scene doesn’t become stale, someone off stage will run across signaling that the scene has ended and to begin a new scene. This will go on for 10-15 minutes or so and usually be very funny, at least at the improv theater near me. AI won’t be able to do that. AI won’t be able to replace plays, drag shows and other forms of live entertainment.

I imagine at some point a movie or tv show will be released that is completely AI, and it will likely flop like one of those Christian movies starring Kirk Cameron. I don’t believe there has ever been a good Christian film. AI flicks will suffer the same fate.

Spang wrote: ↑August 15, 2025, 11:22 am

Winnow wrote:give any thespian friends you have a dollar because they'll be out of work sooner rather than later.

I don’t know if you’re familiar with long-form improv, but it involves a group of actors who, after being given a simple prompt from the audience, create a scene from scratch. So that a scene doesn’t become stale, someone off stage will run across signaling that the scene has ended and to begin a new scene. This will go on for 10-15 minutes or so and usually be very funny, at least at the improv theater near me. AI won’t be able to do that. AI won’t be able to replace plays, drag shows and other forms of live entertainment.

I imagine at some point a movie or tv show will be released that is completely AI, and it will likely flop like one of those Christian movies starring Kirk Cameron. I don’t believe there has ever been a good Christian film. AI flicks will suffer the same fate.

I disagree with the first point. AI can do improv. You even mention the actors being given a single "prompt" which is exactly what AI takes as input. Every "prompt" you give AI, it instantly changes to that prompt and creates whatever it is...story, image, video, etc. AI is literally designed to take prompts and immediately act on them.

Your second point, much like the original AI art 2 years ago and older AI videos, perhaps the first all AI video will flop and you will rejoice but that is exactly like the original AI images that everyone hated, now you can't tell the difference if you randomly see a recent one that isn't on an AI thread.



Since I suck at giving my own examples. Look what this guy was able to do with WAN2.2:

It's an hour+ long because he goes into how he did it but watch the first 3 minutes to see what any nerd can do already in a weekend locally on your own PC. This guy didn't even have a fancy 24GB VRAM card. He had a "crappy" MAC. He was able to insert himself in as a kid and even as a female.

Go ahead and nitpick it, but that's really freaking good especially if you look at the time progression/improvement of AI video over just one year.

winnow wrote:I disagree with the first point. AI can do improv. You even mention the actors being given a single "prompt" which is exactly what AI takes as input. Every "prompt" you give AI, it instantly changes to that prompt and creates whatever it is...story, image, video, etc. AI is literally designed to take prompts and immediately act on them.

There is no way AI can walk onto a stage in a theatre in front of actual people and perform improv. That would require additional technology that, like AI, no one is asking for.