Luma Dream Machine

[Mr. X]

https://lumalabs.ai/dream-machine
Put in a pic and a prompt describing what happens and it AI animates the pic. Its pretty picky on prompts.
https://storage.cdn-luma.com/lit_lite_i ... d2b216.mp4

[Mr. X]

https://storage.cdn-luma.com/lit_lite_i ... 5657fc.mp4
She became Miss Piggy but still kind of cool

[wunderladyyy]

https://storage.cdn-luma.com/lit_lite_i ... aaf34f.mp4

The weird transition freaked me out
But I guess I expected it given the picture I used.

[Mr. X]

Better than my results. You can tell it to not move camera etc.

[wunderladyyy]

Better than my results. You can tell it to not move camera etc.

Thank for the awesome link and prompt tip

[jlocke]

This looks really cool. I've had the most success so far with keeping things simple:
https://storage.cdn-luma.com/lit_lite_i ... 743df8.mp4

[jlocke]

Thanks again for sharing this site. I'm having a blast. Quick heads up, you only get 30 total creations per month. I did 10 in 1 day because I thought they reset every day. I've had the most success keeping the camera fixed. "No pan, no zoom, camera no movement." Having the model move too much has been disastrous. Slow motion running on a beach turned into legs whirl around in a freakish tornado.

If anyone has some success with the prompts, please post. I'd like to keep refining my approach. Also if there are any sites like this that animate still images, I would love to try it out.

AI Superheroine Comes To Life.gif (18.26 MiB) Viewed 11518 times

[Mr. X]

That looks fantastic! Its really bad at obtuse angles but good at straight on shots. My guess is it needs to figure out were bones go so it can rig the figure.

BTW I posted this thread in the wrong forum. It should be in lounge. Maybe a mod can move it.

[phopho]

https://storage.cdn-luma.com/lit_lite_i ... 6939ba.mp4

I can't really get much movement, I got some thickness on the heroine though! How do you export as a gif?

[jlocke]

https://storage.cdn-luma.com/lit_lite_i ... 6939ba.mp4

I can't really get much movement, I got some thickness on the heroine though! How do you export as a gif?

I downloaded the mp4 and then converted with ezgif.com
They are also great for converting webm to jpg

[jlocke]

This site might be worth checking out. It desperately wants to morph my image to something else. But if you use the command "Do not change model" you can get some interesting results:

I upscaled an ai generated image and used face swap ai to get this image:

Jasmine-SH-02-j.jpg (213.14 KiB) Viewed 11435 times

And Haiper.ai was able to produce this:

Victory-01.gif (18.46 MiB) Viewed 11435 times

I might have specify, belly button does not change. Eyes do not become demonic.

[superheroinelinks]

[superheroinelinks]

[superheroinelinks]

[superheroinelinks]

[superheroinelinks]

[superheroinelinks]

[Mr. X]

I think I recognize this from the boys with Stormfront and Homelander in an alley killing a mugger.

[superheroinelinks]

[superheroinelinks]

[superheroinelinks]

You have to hear this!

[superheroinelinks]

[BlackOut]

Just a little something to add to the collection here.

[DonShip]

Just a little something to add to the collection here.

Very sexy

[superheroinelinks]

Dead link? Nothing opens?

[BlackOut]

Further experiments occasionally yield pleasant results.

[jlocke]

Further experiments occasionally yield pleasant results.

Well done! Love it! Would you mind sharing what prompt you used?

[BlackOut]

Further experiments occasionally yield pleasant results.

Well done! Love it! Would you mind sharing what prompt you used?

Camera remains stationary. Gal Gadot is Wonder Woman. Wonder Woman is captured and bound by rope. Wonder Woman's arms are tied behind her back. Wonder Woman is on her knees. Muddy. Photorealism. Worried. Prisoner. Squirming.

I did a little research and according to one source AI responds better to positive prompts than negative prompts. So for example I've been using Camera remains stationary instead of no camera movement.

[sneakly]

Further experiments occasionally yield pleasant results.

That does look pretty good. Nothing weird going on with her features, she does look real. This is going to get better.

[jlocke]

Camera remains stationary. Gal Gadot is Wonder Woman. Wonder Woman is captured and bound by rope. Wonder Woman's arms are tied behind her back. Wonder Woman is on her knees. Muddy. Photorealism. Worried. Prisoner. Squirming.

I did a little research and according to one source AI responds better to positive prompts than negative prompts. So for example I've been using Camera remains stationary instead of no camera movement.

Fantastic! Thanks for sharing, I'll give it a try with my favorite Wonder Woman.

[BlackOut]

Minus the wonkiness of the hands at times, I kind of found this to be rather nice, especially the motion captured. Plus, ya know, Lynda Carter.

[superheroinelinks]

[jlocke]

Minus the wonkiness of the hands at times, I kind of found this to be rather nice, especially the motion captured. Plus, ya know, Lynda Carter.

Love it! I think the problem with AI hands are the number of bones and joints. They don't have a great solution at the moment, but I think they will only improve. I think it's amazing that we can create new videos/photos of actresses that are retired or at least no longer look like they used to. And we don't even need an expensive, experimental Time Machine that might destroy our entire reality.

[superheroinelinks]

[BlackOut]

Terrifyingly intriguing to think where this kind of technology might be in a year or twos time - as this little homage to a favorite cosplayer/actress can attest to.

[superheroinelinks]

[superheroinelinks]

[BlackOut]

This has become addictive

[BlackOut]

Very.

[BlackOut]

Addictive.

[jlocke]

Very.

Love this one! What prompt did you use?

[BlackOut]

Static camera shot of a dreamy and subdued Wonder Woman played by Lynda Carter. Wonder Woman is being hypnotized by a mysterious older man barely seen in the foreground. Translucent golden swirls of enchanting magical color swirl towards the center of her head. Wonder Woman has a vacant look of surrender as she is mesmerized. The center of her tiara pulses red. Photorealism style.

To be honest I don't know really how much influence the prompts have over animations that are rendered from images. Personally I think the camera lingo is fairly reliable. Also I did one of the purchase plans so my typical method is to just try again, and again, and again, and... until I get a result that I'm happy with or just abandon the effort.

Camera control also seems to be more accurate when unclicking the "enhance prompt" option - yet sometimes you get far more animated results when clicking it.

Mostly it seems to be random luck sprinkled with a little bit of tried and true discoveries made along the way. Also the image itself has quite a lot to do with the final results. Regardless I am pretty optimistic that the technology is going to only improve over the course of time.

And eventually take over the world.

[superheroinelinks]

Minus the wonkiness of the hands at times, I kind of found this to be rather nice, especially the motion captured. Plus, ya know, Lynda Carter.

What app or website did you use to do this? LOVE IT!
https://lumalabs.ai/dream-machine
yes or no?
c.

[BlackOut]

Minus the wonkiness of the hands at times, I kind of found this to be rather nice, especially the motion captured. Plus, ya know, Lynda Carter.

What app or website did you use to do this? LOVE IT!
https://lumalabs.ai/dream-machine
yes or no?
c.

I've been using Luma exclusively, although I wouldn't mind looking into other options. This is kind of my first foray into the whole AI scene - but it's certainly got me hooked.

Speaking of which this turned out well.

[superheroinelinks]

Can you make these still images of Batgirl come to life? Or teach me how to do it?

If you teach a man to fish......

I found this one website is good but only does faces

[BlackOut]

Luma has an option to use a starting image and an end image that you provide for reference. So for example I took two of the pictures you provided and Luma rendered this.

[jlocke]

Luma has an option to use a starting image and an end image that you provide for reference. So for example I took two of the pictures you provided and Luma rendered this.

Fuck yeah! That is awesome!

I'm not seeing that option, could you help a brother out?

[BlackOut]

I kind of discovered it one day just by accident. I'm not sure if it's a free option? But after you add your first image you should see a second window pop up to add an end point.

[superheroinelinks]

BlackOut this is AMAZING! I Really would love to be able to do this too. If you can help us, so I don't have to keep asking you how to do it.

[jlocke]

BlackOut this is AMAZING! I Really would love to be able to do this too. If you can help us, so I don't have to keep asking you how to do it.

I think I figured it out. Go to Luma Dream:
https://lumalabs.ai/dream-machine/creations

Click the image icon to the left of the "Type some text or add a picture" Add your first image and then it auto adds an option to Add Last Frame. I'm trying one with Lynda Carter from her season 2 Wonder Woman promos.

They did not have this option last time I played with Luma. This is really cool. I can't believe how well that Yvonne Craig video turned out. I think this has A LOT of potential!