Thanks for posting - starting to seriously get my head around AI tools now, I notice Pika has 'PICK YOUR PAID PLAN' now to get rid of watermarks etc; has that just been introduced as the free plan comes with 300 credits then you would have to pick a plan to continue...
@@AIAnimationStudio Nope. Kaiber absolutely sucks. No amount of prompt get anywhere close to what I requested. Why loop stupid animations over and over in every single clip. It's probably my worst purchase of all of 2023 I really wished I never bought. Utterly useless. Tech support is awful.
Very cool and informative Videos. Do you have any tips for creating Image to Video (lipsync) longer than 1minute videos, without using expensive options like d-id or else?
Thanks. Check out synclabs.so/ which is using Wav2Lip (They have a google Collab for this too). I've only briefly played with it, but might help you a lot.
@@AIAnimationStudio So far it seems pretty cool. You have to give a Video as Input (no bigger than 250 MB, 30fps, and 1080p), as well as providing the audio file you want to use. So in combination with Pika and 11Labs, I figured that works pretty well and gives you cool results. You can get up to 5 minutes per month on the Free Plan, which seems pretty generous, I might say. After that, with the Subscription plans, it's basically 1 Dollar per minute. Some things to consider are, that it can only do one person per clip, so if you have a slightly longer video with scene changes, it struggles a lot. So you're better off doing videos scene by scene and with no more than 1 Person in the scene. But if you keep that in mind it gives an overall good Quality.
It depends on what you want to create. For absolute photorealistic content none of these can do it good enough.. you have to have a midjourney created image or use AE or additional software to „repair“ the messed up motion I hope when MJ introduced its own text to video we get better movement as MJ does not rely on Stable Diffusion in its creation.😇
Yeah 100% agree. Using MJ images in After Effects, potentially with some depth displacement for a little movement in the scene is great. As for actual scene/character animation they all have a way to go for flawless output. 👍
That guy walking with the green screen looks awesome. So does it transfer an image onto the footage. Thos could be very clever for generating better human motion shots.
Yeah that seems like a really powerful approach for the future. It would be great if we could combine the video source with an image reference to have a little more control over the output. It seems to just reference the layout/motion in it's own generations but without taking much (if any) of the image style which is impressive.
Right now for ease. domo ai does a great job. I don't think they offer much or any free usage anymore but the results are good. The alternative would be a workflow using ComfyUI.
Yeah, I run up against that a lot with various tools. Midjourney would'nt make a 'globulous' monster. Try a few times or try rewording. Definitely a reason to champion open-source tools like Stable Diffusion (which typically have less over zealous safety filtering in place).
I’ve been experimenting with pixverse and getting pretty good to mixed results. Some of the girls look absolutely fantastic and soon look terrible either eyes out of focus or weird teeth, or weird where the eyes look etc. bad with hands, movement, etc. but it’s free and unlimited and does a great job sometimes. Another weird thing is that they are very strict on their moderation rules, like can’t say cleavage, or lingerie etc, can’t say naked or nude obviously, but often they generate naked or nude models themselves. Sometimes innocent enough like cute beautiful woman in a hot tub and suddenly she’s topless. Which is fun I guess but unusable for work haha. Make sure to add some details about a bikini or else be warned may get some nsfw videos.
I have a real problem with paying real money to generate trash that you can't use or isn't what you asked for. I will NEVER EVERY pay any of those websites to generate trash. I will struggle with local generations. I get no watermark and I don't have to worry about wasting money. The only struggle I have is shitty resolution and I have to work out the best settings myself.
🙂fair. Both have their strengths and I guess you could say Runway had the top spot until Pika came out on Discord for free and the quality improved pretty quickly. I'm a big fan of the new Pika website and will be happy to pay a subscription to remove the watermark when we can. 👍. Long way to go for flawless output but it's edging closer all the time.
For me personally, one of the most ANNOYING things all of these sites/companies/platforms do is they make you pay for that demo video. I'm sorry but that is just bad business. I should be able to AT LEAST do one animation video for free. IDC if it's even like 240 res, I want to see the capabilities.
Yeah, a lot of those that have been around for a year have moved to paid teirs only. But some newer things still have short trials. I think some of it is due to cloud computing companies (AWS, Azure, Google) offering LOTS of free credits as incentives for companies to build on their platforms... but the free credits dry up and the companies have to charge and not al have the backing to fund trials. It is a shame these AI processes still cost so much to run.. hopefully it'll drop dramatically over time but I reckon we'll constantly expect more and more from them.
Thanks for posting - starting to seriously get my head around AI tools now, I notice Pika has 'PICK YOUR PAID PLAN' now to get rid of watermarks etc; has that just been introduced as the free plan comes with 300 credits then you would have to pick a plan to continue...
Yeah that came in over the last two days. So no more unlimited free generations.
Great overview and analyse of the different AI's. 👍
You got a new suscriber from Argentina. Great video!
Missing Kaiber, which imo is the best current option for AI video
That's one I really need to revisit. Not checked out Kaiber for 6+ months.
@@AIAnimationStudio Nope. Kaiber absolutely sucks. No amount of prompt get anywhere close to what I requested. Why loop stupid animations over and over in every single clip. It's probably my worst purchase of all of 2023 I really wished I never bought. Utterly useless. Tech support is awful.
Very cool and informative Videos. Do you have any tips for creating Image to Video (lipsync) longer than 1minute videos, without using expensive options like d-id or else?
Thanks. Check out synclabs.so/ which is using Wav2Lip (They have a google Collab for this too). I've only briefly played with it, but might help you a lot.
Thanks alot, i am going to check that out and report back :)@@AIAnimationStudio
@@mitchibob . Cool let me know. It's been on my list for a future video for a while but wanted to tie it in with some other processes too.
@@AIAnimationStudio So far it seems pretty cool. You have to give a Video as Input (no bigger than 250 MB, 30fps, and 1080p), as well as providing the audio file you want to use. So in combination with Pika and 11Labs, I figured that works pretty well and gives you cool results.
You can get up to 5 minutes per month on the Free Plan, which seems pretty generous, I might say. After that, with the Subscription plans, it's basically 1 Dollar per minute.
Some things to consider are, that it can only do one person per clip, so if you have a slightly longer video with scene changes, it struggles a lot. So you're better off doing videos scene by scene and with no more than 1 Person in the scene. But if you keep that in mind it gives an overall good Quality.
It depends on what you want to create.
For absolute photorealistic content none of these can do it good enough.. you have to have a midjourney created image or use AE or additional software to „repair“ the messed up motion
I hope when MJ introduced its own text to video we get better movement as MJ does not rely on Stable Diffusion in its creation.😇
Yeah 100% agree. Using MJ images in After Effects, potentially with some depth displacement for a little movement in the scene is great. As for actual scene/character animation they all have a way to go for flawless output. 👍
Very good tip. Any more tools?
Thank you for the detailed explanation and excellent comparisons! Loved it
Loyal to Leo but might try Pika to compare. Your videos are most helpful❤
amazing tutorial bro thanks a lot
That guy walking with the green screen looks awesome.
So does it transfer an image onto the footage. Thos could be very clever for generating better human motion shots.
Yeah that seems like a really powerful approach for the future. It would be great if we could combine the video source with an image reference to have a little more control over the output. It seems to just reference the layout/motion in it's own generations but without taking much (if any) of the image style which is impressive.
You’re the best, ty!!
i am really impressive by pika - I discover this tool. ! Thanks
What would you say is the best program to create video to video (3D cartoon)?
Right now for ease. domo ai does a great job. I don't think they offer much or any free usage anymore but the results are good. The alternative would be a workflow using ComfyUI.
Thanks for this Video, am excited to see how much AI Video Generation have advanced so much in just a short 6 months!
Great stuff, thank you.
Great video
Subscribed
Cheers.... have some cool new content in the works.
Very nice. Thank you
Love ✊🏿
Thank you!
WOW!!
Recreating iconic scenes from YJHD with a GenAI twist! ✨ #yjhd #ai #genai #shorts
ruclips.net/user/shorts1RbvONx_h14?feature=share
I wanted Runway Gen-2 to create a red paint splash video but it got flagged by content policy!? !?
wtf?!
Yeah, I run up against that a lot with various tools. Midjourney would'nt make a 'globulous' monster. Try a few times or try rewording. Definitely a reason to champion open-source tools like Stable Diffusion (which typically have less over zealous safety filtering in place).
I’ve been experimenting with pixverse and getting pretty good to mixed results. Some of the girls look absolutely fantastic and soon look terrible either eyes out of focus or weird teeth, or weird where the eyes look etc. bad with hands, movement, etc. but it’s free and unlimited and does a great job sometimes.
Another weird thing is that they are very strict on their moderation rules, like can’t say cleavage, or lingerie etc, can’t say naked or nude obviously, but often they generate naked or nude models themselves. Sometimes innocent enough like cute beautiful woman in a hot tub and suddenly she’s topless. Which is fun I guess but unusable for work haha. Make sure to add some details about a bikini or else be warned may get some nsfw videos.
❤
good
nice
Title should actually be "Which Ai video generation platform sucks the least?" If that's the case then it would be Leonardo.
😅 .... fair. Either way, I hope you got something a little bit useful out of the video.
I have a real problem with paying real money to generate trash that you can't use or isn't what you asked for. I will NEVER EVERY pay any of those websites to generate trash. I will struggle with local generations. I get no watermark and I don't have to worry about wasting money. The only struggle I have is shitty resolution and I have to work out the best settings myself.
pika labs are the best
Are you deadpool ???
Runway never had the top spot. Pika is and was always better.
🙂fair. Both have their strengths and I guess you could say Runway had the top spot until Pika came out on Discord for free and the quality improved pretty quickly. I'm a big fan of the new Pika website and will be happy to pay a subscription to remove the watermark when we can. 👍. Long way to go for flawless output but it's edging closer all the time.
@@AIAnimationStudio I agree. Thank you for your content👍👍
For me personally, one of the most ANNOYING things all of these sites/companies/platforms do is they make you pay for that demo video. I'm sorry but that is just bad business. I should be able to AT LEAST do one animation video for free. IDC if it's even like 240 res, I want to see the capabilities.
Yeah, a lot of those that have been around for a year have moved to paid teirs only. But some newer things still have short trials.
I think some of it is due to cloud computing companies (AWS, Azure, Google) offering LOTS of free credits as incentives for companies to build on their platforms... but the free credits dry up and the companies have to charge and not al have the backing to fund trials.
It is a shame these AI processes still cost so much to run.. hopefully it'll drop dramatically over time but I reckon we'll constantly expect more and more from them.
THE MOVEMENT IS NOT PERFECT YET
Not yet, its' come a long way in a year but still a long way to go to fix finer details.