Luma does really good at Pixar because it was originally an image to 3d generator and a lot of its data is based on that, same with the vector graphics.
I will try my own art and be nice if I have 2 minutes of my own animation and be cool file send as slides but or the movement and reads and o I see AI and finish more extended movie
I just finished a vid. Did u try unticking the magic prompt box? I found it doesnt pan as crazy. I had a difficult shot where my soldiers were just doing goofy things. I unticked and typed slow motion, subtle camera movement and it worked ;)
Have you found any good prompts guide? im starting to get tried with my free 10 a day trial and error attempts, my game characters and their weapons screenshots keeps on transforming/morphing into different creatures/monsters.
Seems extremely promising! I'm somewhat disappointed that it struggles the most with anime-like styles, which I usually favor. Given how rapidly things progress, I'm eager to see how it develops over the next two months. Hell, probably the next couple of hours. 😅
Thanks for your research Tao, it really makes a difference to view it. I was wondering in your presentation of the various styles, are you using the enhance checked or not checked? Did you do a comparison for each style based on enhanced/Non-enhanced. I have been uploading my images and not using enhanced as I found that they follow my prompts and style of my images more closely. Like you, I notice that sometimes Luma turns an image into a cartoon which makes the video unusable.
Я понял , одним русским пришло в голову, воскрешать свою семью , смотреть и вспоминать их живыми. Остальной мир просто экспериментирует с играми и мультиками😮
You can try to reduce the camera motion by turning off the "enhance prompt" box, I also made a tutorial for how to control camera motions with different prompts here: ruclips.net/video/i4it_Nb-3PM/видео.html
4:43 - Note that the wings of the horse are actually not really truly connected to his body, they seem to appear in the background and behind the horse and not really belonging to him so that's why Luma showed them on the back of the horse and then it disappears because of that, try generating such video using a clearly visible wings that are visibly coming out of the horse's body and then it should work better I suppose.
Luma definitely needs to add more settings because I have NO clue what I'm gonna get when I hit generate. Sometimes the image doesn't move at all, sometimes it moves way too much, and sometimes the characters go completely off model, and sometimes it's good. I wouldn't be as annoyed if the generations didn't take NINE HOURS for 5 seconds. (I know the payed version isn't as slow, but I'm not paying 25 dollars a month for four second clips that I don't have any control over.
Hey, if you turn off the "enhance prompt" option on the right side underneath the prompt bar, it helps reduce the amount of motion and the result follows the prompt better.
@@taoprompts but that's just for regular prompts with no image. The image to video generations take an eternity and I have no clue what makes the generations come out good vs jankey.
For the anime ones I tested, I added the phrase "The camera does not move" and that seemed to fix it. The prompts help guide the motion you want. If you don't want much motion, just describe the scene. Otherwise you should limit the prompt to a single action, such as "kicking legs", "talking to camera", or "riding bike". Those prompts worked the best for me. "Dancing" and "jumping jacks" didn't work as well.
How to make the character blink the eyes , I tried so many times some times it does but it's all automated, secondly how to over come deformation Help plz
Very useful, thanks 👍🏽 i am yet to properly explore Luma, but i guess we dont need to mention the style of video (like pixar etc) required in the prompt ?
If you're using image-to-video (with a reference image) then it should automatically copy your image style. If you're generating videos with just the prompt then you do need to mention the style you need (like pixar)
The potential from Luma for is a lot higher than Runway imo. I think they just need to add in a few more features for it to be more consistent for us to use. But it is free to try which is always nice.
It seem that if you turn off the enhance prompt button it reduces the motion blur and stops the camera from panning out of scene, that least turning enhance prompt makes anime generations look better when you do image to video.
I think Luma is better than Pika for most situations including 3D animation. There isn't any video generator I've seen that's great with 2D animations.
@@taoprompts Ok, I'll open a new channel with just images as stories. Not sure how to animate the images I have as those are like story for people/kids sort of thing in 2D.
@@taoprompts Thank you for helping me! Now I can control the prompt so that the camera stays still when using the command. When I add an image and control the prompt to move according to the existing image, the camera still moves and does not stand still. Hope you can help me solve this problem!
I just used simple prompts for these. Like "a boy in an anime", "a woman". I made a follow up video about how to control camera motions here: ruclips.net/video/i4it_Nb-3PM/видео.html
@@taoprompts if you want to see examples on my channel, exclusively prompts except the first videos they are visible due to the quality which I find average in comparison with the 100% prompt
Luma does really good at Pixar because it was originally an image to 3d generator and a lot of its data is based on that, same with the vector graphics.
That makes sense, the pixar videos are really impressive.
I will try my own art and be nice if I have 2 minutes of my own animation and be cool file send as slides but or the movement and reads and o I see AI and finish more extended movie
They've added a video extension feature now, you can get videos up to 15 seconds long.
I hope Luma will get better at depicting anime scenes & nonhuman characters. It looks like it’s lacking in these areas.
It definitely needs some improvement for anime style scenes.
You can prompt the camera to stay still or pan slowly or zoom etc
I just finished a vid. Did u try unticking the magic prompt box? I found it doesnt pan as crazy.
I had a difficult shot where my soldiers were just doing goofy things. I unticked and typed slow motion, subtle camera movement and it worked ;)
kurt russell..nice!! subbed ya!
That's a good suggestion! It does seem to work much better.
yes, magic prompt is not recommended coz of wacko results
Damn i,wish you added your prompts input to see what you wrote
Have you found any good prompts guide? im starting to get tried with my free 10 a day trial and error attempts, my game characters and their weapons screenshots keeps on transforming/morphing into different creatures/monsters.
I will add prompts in the future 👍
Yes please @@taoprompts
input the image into chatgpt and have it describe the image and ask it to write your prompt as well
Seems extremely promising! I'm somewhat disappointed that it struggles the most with anime-like styles, which I usually favor.
Given how rapidly things progress, I'm eager to see how it develops over the next two months. Hell, probably the next couple of hours. 😅
Yeah, hopefully someone comes out with a solid anime model. A lot of us would be interested in that.
@@taopromptsgive it another few weeks. Something should preview or drop.
SORA Killer 🤣😂😅
Or 3d animator killer
It's hard to know what Sora is really like without being able to get access to it.
Thanks for your research Tao, it really makes a difference to view it. I was wondering in your presentation of the various styles, are you using the enhance checked or not checked? Did you do a comparison for each style based on enhanced/Non-enhanced. I have been uploading my images and not using enhanced as I found that they follow my prompts and style of my images more closely. Like you, I notice that sometimes Luma turns an image into a cartoon which makes the video unusable.
I have the enhanced prompt checked in this tutorial. It does seem to follow the prompts much better when I removed the enhanced prompt option.
Я понял , одним русским пришло в голову, воскрешать свою семью , смотреть и вспоминать их живыми. Остальной мир просто экспериментирует с играми и мультиками😮
Thanks for the video. I run into simelar problems with luma. And the pricing is hefty, $30 for 120 videoclips of 5 seconds is realy a lot.
I am experimenting with trying to get better camera controls, I think Luma does have a lot of potential
Glad I’ve came across your vids. I was wondering why the images would pan out randomly even though I’ve just put in the same prompt 🤣
You can try to reduce the camera motion by turning off the "enhance prompt" box, I also made a tutorial for how to control camera motions with different prompts here: ruclips.net/video/i4it_Nb-3PM/видео.html
I’m getting a lot of unnecessary movements , could u share some of your prompts structure for the image to video animations
here's a tutorial where I show some ways to manage the camera motions: ruclips.net/video/i4it_Nb-3PM/видео.html
its not working here.. its take forever to render
Their website is just overwhelmed right now, hopefully they will add more servers.
4:43 - Note that the wings of the horse are actually not really truly connected to his body, they seem to appear in the background and behind the horse and not really belonging to him so that's why Luma showed them on the back of the horse and then it disappears because of that, try generating such video using a clearly visible wings that are visibly coming out of the horse's body and then it should work better I suppose.
Thanks for the suggestion, I tried other images where the wings attachment is more clear and it does seem to work better
@@taoprompts :) Cool, glad to hear
I got a pro plan few days ago. Luma is a dud. Watermarks? No control over movement? It's so 2023
I think they released a bit too early, hopefully they listen to our feedback and add in those changes.
great demo of all the different styles, would love to see some text to vid examples, awesome vid
Thank you, I am planning to test out more prompts soon 👍
3d works perfect , flat 2d images, not at all
It does struggle with cartoony images more than photorealistic.
Hummmmm...great
can I make videos with luma 1 minute long
No, it only extends to 15 seconds
Luma definitely needs to add more settings because I have NO clue what I'm gonna get when I hit generate. Sometimes the image doesn't move at all, sometimes it moves way too much, and sometimes the characters go completely off model, and sometimes it's good. I wouldn't be as annoyed if the generations didn't take NINE HOURS for 5 seconds. (I know the payed version isn't as slow, but I'm not paying 25 dollars a month for four second clips that I don't have any control over.
Hey, if you turn off the "enhance prompt" option on the right side underneath the prompt bar, it helps reduce the amount of motion and the result follows the prompt better.
@@taoprompts but that's just for regular prompts with no image. The image to video generations take an eternity and I have no clue what makes the generations come out good vs jankey.
@@knightfallprotocol Here's a video guide with some ideas on how to control camera in image-to-video
For the anime ones I tested, I added the phrase "The camera does not move" and that seemed to fix it. The prompts help guide the motion you want. If you don't want much motion, just describe the scene. Otherwise you should limit the prompt to a single action, such as "kicking legs", "talking to camera", or "riding bike". Those prompts worked the best for me. "Dancing" and "jumping jacks" didn't work as well.
Thanks for weighing in. Simple motions do appear to work decently in the prompts.
what were your prompts for the other stuff like the western cartoon or pixar examples? Did you just write - girl stands in front of family and talks?
These are image-to-video. The prompts I used were very simple. Like "a boy in a forest" or something like that.
Agreed on all. Really need better camera control.
Camera controls are a must have for making serious videos.
the problem is you absolumtly not have any controls of what you are moving camera and subject and props....
I am making a video on how to control camera motions better
How to make the character blink the eyes , I tried so many times some times it does but it's all automated, secondly how to over come deformation
Help plz
Turn off the "enhance prompt" feature, that will help the Ai follow your prompts better
Very useful, thanks 👍🏽 i am yet to properly explore Luma, but i guess we dont need to mention the style of video (like pixar etc) required in the prompt ?
If you're using image-to-video (with a reference image) then it should automatically copy your image style. If you're generating videos with just the prompt then you do need to mention the style you need (like pixar)
It's not an art form. It's stealing bread from real artists. 👎
How yo 9:16 ratio with text to video ? Thanks
I don't think there's a way to do that right now
@@taoprompts thanks 😍
so maybe i should go ahead and stick with runway ml for another 9 months. I am sure they will greatly improve by then
The potential from Luma for is a lot higher than Runway imo. I think they just need to add in a few more features for it to be more consistent for us to use. But it is free to try which is always nice.
Trying to use it but when the queue takes hours and you're limited to so few generations it's impossible to experiment and try to get good results
Their website is overflowed right now, they'll probably add more servers as soon as they can.
It seem that if you turn off the enhance prompt button it reduces the motion blur and stops the camera from panning out of scene, that least turning enhance prompt makes anime generations look better when you do image to video.
That does seem to work!
Same issue when it comes to anime. It just pans or tilts, doesn't actually move the characters
It seems like Anime vids are one of it's weaker styles.
it's a good start. luma works much better with reference images. I hope they work hard!
Image-to-video does work pretty well. I do also plan on doing some experiments with just the prompts.
great. have you seen they also released very short prompting guide. Would like to hear more about Prompting in every video
Do you mean Luma officially released a prompting guide? Where can I find that?
So pika is better for videos? Animated movies
I think Luma is better than Pika for most situations including 3D animation. There isn't any video generator I've seen that's great with 2D animations.
@@taoprompts Ok, I'll open a new channel with just images as stories. Not sure how to animate the images I have as those are like story for people/kids sort of thing in 2D.
I like this boy.
Can you give me a prompt to control the camera? For example, you want it to stand still in one position, only the subject moves
I made a video guide on how to control the camera here: ruclips.net/video/i4it_Nb-3PM/видео.html
@@taoprompts Thank you for helping me! Now I can control the prompt so that the camera stays still when using the command. When I add an image and control the prompt to move according to the existing image, the camera still moves and does not stand still. Hope you can help me solve this problem!
It looks like Luma really needs some camera controls.
Yeah, you can control a little bit using the prompt, but controls like Runway would be better.
Great video 💡
Haha seem recognized the movie slots haa😊
Yeah, it sure did
What are the typing? Mine barely move
I just used simple prompts for these. Like "a boy in an anime", "a woman". I made a follow up video about how to control camera motions here: ruclips.net/video/i4it_Nb-3PM/видео.html
GG
definitly we need some controle - )
For sure, I think they have some updates planned soon. Maybe they will add in controls in them.
Please add prompts to these references. That would be really helpful. Thanks
I don't think he added any prompts. he generated the reference images and pasted them without text
That’s too bad
These are done with image-to-video using reference images. I didn't use any complex prompts, just simple things like "a boy" or "an animal"
I think you underestimated the 100% prompt ❤❤❤
I haven't experimented with only using the prompt by itself, there's definitely a lot of things I want to try out.
@@taoprompts if you want to see examples on my channel, exclusively prompts except the first videos they are visible due to the quality which I find average in comparison with the 100% prompt
Another wonderful tutorial, thanks 💯
Thanks Sean, Luma has some really impressive capabilities.
Great video. Thank you
Thank you, glad to know this was helpful.
Great video ma man
Appreciate it man 👍. Ai video keeps getting better.
That was awesome
Thank you 👍
Nice work
Thanks Toby! Looking forward to see your review of Luma
This is what we needed. A service that puts real pressure on OpenAI
Hopefully this pushes them to let us try Sora soon
Prompts in Luma, not really good.