This is great, really appreciate that you put the full prompts in there! I generally go for short prompts and get decent results, but I think your final one especially shows how a detailed prompt can get a more precise change in the scene.
Yeah, it definitely depends on how much of the shot needs to connect with another shot in the sequence. If it's a standalone b-roll shot, and the shots on either side are not connected, you can afford to let the AI do it's thing, but when you need continuity, the prompts just get longer and longer.
Now THIS is the video I’ve been waiting for! Thanks, Haydn. I never even thought to try the new MiniMax model on realistic film stills as they billed it for 2D animation only. And damn, don’t sleep on Vidu! Impressive results. And now Sora finally entered the arena. An amazing time for solo filmmakers! 🙏🏼
🎉the best compare vs video I’ve seen nice work. I’m on the creator group in Minimax and proud it came out on top! A lot have been saying Kling is better but for dynamic range minimax is the boss. You can dial it back a bit in you terrified prompt by adding less terrified meaning, I know minimax will at least respond to subtle to extreme in this prompt. 🎉🎉
Hey! Thanks very much! I’d love to connect with yourself and the team directly if you’re willing? I’m sure there will be prompts that I’d love to have your expert advice about in future 😃😁 you can email me directly at: haydnrushworth@me.com
They each seem to have their own strengths and weaknesses, so depending on your use, different models may be more suitable. For me, looking for authentic emotions in my characters, Minimax definitely my favourite, closely followed by VIDU.
There is a lot more control with Audio for lips movement, than words in "Text", because in audio you can control the exact moment for the lips to moves saying what ever you want over the time of the video generated.
Lip sync, yes, but just the “traditional” way, with an audio file or text to language. What I’m hoping for is to type text in the prompt and the character “mouths” those words.
@HaydnRushworth-Filmmaker you could use a text to voice tool (like eleven labs) to generate an audio file from your text and then use that as your lip sync audio file. Extra step, but might that work?
Great tests. The very last thing you say, not sure if that would work because you would have no audio. This would cause zn headache to sync up. I like the dail idea. That would great. Its a shame 11s lab dont think like this. More pause options and emotion guidance.
@HaydnRushworth-Filmmaker wow, lol. I would live to see that workflow. Every animation studio does the voice first . I think the only way it could work if the ai video company would set up a voice generator. Then you would have all the perfect pauses for precise sync. Vidu looks good as well. Might check them out. I just checked Sora, its put but not in UK ....f@@k!!!!!!
I don't know if you know, but the command structure for each program should be different. In some programs, for example, the character should be described first and then the surroundings, and in others the other way around. Therefore, testing programs with the same prompt is wrong.
You’re right, I hadn’t taken that into account. That’s a really good point. Thank you. Having said that, would a different prompt structure deliver better quality “emotion” results from characters? I’d always assumed the AI tool just didn’t know what various subtle emotions meant.
Thanks for this comparative and objective test! I test a lot of video AI, and my preference in terms of aesthetic result goes to Minimax, it's a shame to be limited to 5 seconds and not to have the tools that Kling or Runway offer
💯 absolutely, but that’s the dilemma we all face. Nine of the tools have everything we need in a single package, so we each need to prioritise the features they are most necessary for our project and make the most of them.
Great test of those four. For lip sync you need audio as text won't give you the correct intonation or spacing. It's a shame that Kling doesn't allow importing a video for lip sync as you could then create the emotions in something else and layer the lip sync on top of that in Kling.
It seems that subtlety is the biggest difference between platforms. Runway is more subtle minimax more over the top. I can more easily imagine runway fitting into an actual film but if you just want something that clearly demonstrates the emotion minimax is better.
I'm using a mix of Runway, Kling, and Minimax now and its working well. Minimax is definitely better at big emotions and big movement, but the generating speed is an issue. I can generate 10 Runway clips before one Minimax or Kling is ready.
Great upload and execution by Mr. Rushwood. However, best comparisons are candid reveal of cherry-picking attempts for all four platforms. Highly doubt the showcased emotions were all first time raw generations. Given Minimax grants uploader CPP to make unlimited generations, compared to limited credits from say Runway, Kling, or Vidu... it's more likely the relative, better generations for Hailuo Minimax would be shortlisted for final showcase, more so than the other three.
That's a fair assumption, but as it happens, I just didn't have the time to keep generating until I got a "good" one. No, it really, honestly is the case that almost all of these were first generation.
Excellent video. I did something like this only using Vidu and I was happily surprised. BTW, with Minimax you can type sentences in the prompt,but the characters only speak in Chinese. Really.
You hit the nail on the head right at the end. Until or unless we get the ability to lip synch with this stuff then it's all fantastically brilliant but also kinda useless. I know.. I know... a year ago this was all science fiction and here we are being all moaning and grouchy about it. One of my biggest peeves with Minimax is that it just can't see a mouth and keep it closed, so it's difficult to generate anything that has a face on it and not have it give the appearance of talking... but anyway... excellent comparison video, thanks so much for putting them through their paces.
Wow Great vid! I think I Will change my subscription to Minimax instead of Kling. Maybe a stupid question but Minimax is hailuoai right? Or do you have a direct link? Thanks 🤗👍
@ thanks for the reply! I tried Runway and Minimax today but wow Minimax takes ages to generate. Maybe because of my free plan? It even stopped generating images and I got a fault.
Hey man, have you been watching this video for a long time? your information is outdated yes, that's right, the voice-over function and lip sync are already available in some AI that you are dimonstrating check again
Do you mean the ability for text-to-video prompts to include dialogue words like “now what are we gonna do?” and then the character in the video mouths those words?
I’ll be honest, I don’t really keep track of the free versions any more. They all seem to have a free version which gives you the opportunity to play, but eventually, if you’re serious about creating, you’ll have to start paying for one of them in some way.
whao... What will happen to the huge budget movies industries across the globe in a few years from now if AI can create such deep emotions... uff .. just thinking about the possibilities
@@HaydnRushworth-Filmmaker Yeah, for sure... Just keep on publishing more videos like this General, I am following you like a disciplined soldier and learning from them!
Minimax is the best. But problem with it is, that all these emotions always is animated same way as it was a robotic 3d model under the choosen face skin. So imagine in one scene you make two angry people in one shot and they both will do same "angry mimics"
Yeah, I really like the massive amount of Lip Sync options, but I haven't come across a model yet that allows you to included dialogue in a prompt and have the lips move as if they're saying those words... "Let's get outta here!", or "Now what do we do?" etc.
your right, if AI doesn't understand humans, AI Understanding OFF AND ON switches, WAIT! THATS IT!, AI UNDERSTANDS MONEY! 😃👍, NO WAIT!, AI UNDERSTANDS CYCLING HISTORY SEARCH! WHAT A TIME TO BE AWAKE IN THE NIGHT.
This is great, really appreciate that you put the full prompts in there! I generally go for short prompts and get decent results, but I think your final one especially shows how a detailed prompt can get a more precise change in the scene.
Yeah, it definitely depends on how much of the shot needs to connect with another shot in the sequence. If it's a standalone b-roll shot, and the shots on either side are not connected, you can afford to let the AI do it's thing, but when you need continuity, the prompts just get longer and longer.
Mr. Haydn, you're doing a great job man... That's an impressive and very informative video!
Heyyyy!! Thanks very much 😁😁😁
Now THIS is the video I’ve been waiting for! Thanks, Haydn. I never even thought to try the new MiniMax model on realistic film stills as they billed it for 2D animation only. And damn, don’t sleep on Vidu! Impressive results. And now Sora finally entered the arena. An amazing time for solo filmmakers! 🙏🏼
Thanks very much, glad I could help. It really is an amazing time for AI Filmmakers! :-)
🎉the best compare vs video I’ve seen nice work. I’m on the creator group in Minimax and proud it came out on top! A lot have been saying Kling is better but for dynamic range minimax is the boss. You can dial it back a bit in you terrified prompt by adding less terrified meaning, I know minimax will at least respond to subtle to extreme in this prompt. 🎉🎉
Hey! Thanks very much! I’d love to connect with yourself and the team directly if you’re willing? I’m sure there will be prompts that I’d love to have your expert advice about in future 😃😁
you can email me directly at:
haydnrushworth@me.com
Do you have any idea if Minimax is planning to roll out "End Frame" capability?
@ I don’t think they have any plans for that not that I’m aware of anyway
Too Good Explanation Bro ❤🎉
Thanks so much, I really appreciate that :-)
awesome man love it!
Hey! Thanks very much indeed, I really appreciate the support :-)
kling is my favorite tool! watching your channel i thnk i can try Minimax out
They each seem to have their own strengths and weaknesses, so depending on your use, different models may be more suitable. For me, looking for authentic emotions in my characters, Minimax definitely my favourite, closely followed by VIDU.
Youre amazing. Thank you so much
Hey! Not at all, thanks for watching 😁
There is a lot more control with Audio for lips movement, than words in "Text", because in audio you can control the exact moment for the lips to moves saying what ever you want over the time of the video generated.
In fairness, I'm sure you're right, although I do love the idea of another option for mouth movement and speech.
This was a great video test. Thanks Haydn!
Doesn't Runway allow you to add lip-synced words?
Lip sync, yes, but just the “traditional” way, with an audio file or text to language. What I’m hoping for is to type text in the prompt and the character “mouths” those words.
@HaydnRushworth-Filmmaker you could use a text to voice tool (like eleven labs) to generate an audio file from your text and then use that as your lip sync audio file. Extra step, but might that work?
With the audio feature on Hailou I'm sure we will see the ability to type what we want them to say and the audio and video sync up. Can't wait.
Oooooohhh, good point. I haven't tested it yet... I'm off to check it out now :-)
Great tests. The very last thing you say, not sure if that would work because you would have no audio.
This would cause zn headache to sync up.
I like the dail idea. That would great.
Its a shame 11s lab dont think like this. More pause options and emotion guidance.
Absolutely. I guess I’m hoping that if the base video is mostly lip synced, the sync would be easier to refine with added audio.
@HaydnRushworth-Filmmaker wow, lol. I would live to see that workflow. Every animation studio does the voice first . I think the only way it could work if the ai video company would set up a voice generator. Then you would have all the perfect pauses for precise sync.
Vidu looks good as well. Might check them out. I just checked Sora, its put but not in UK ....f@@k!!!!!!
I don't know if you know, but the command structure for each program should be different. In some programs, for example, the character should be described first and then the surroundings, and in others the other way around. Therefore, testing programs with the same prompt is wrong.
You’re right, I hadn’t taken that into account. That’s a really good point. Thank you.
Having said that, would a different prompt structure deliver better quality “emotion” results from characters? I’d always assumed the AI tool just didn’t know what various subtle emotions meant.
@@HaydnRushworth-Filmmaker Watch this video. The Ultimate Cinematic Prompting Guide for Kling AI Video (And Are Pro Features worth it?)
Thanks for this comparative and objective test! I test a lot of video AI, and my preference in terms of aesthetic result goes to Minimax, it's a shame to be limited to 5 seconds and not to have the tools that Kling or Runway offer
💯 absolutely, but that’s the dilemma we all face. Nine of the tools have everything we need in a single package, so we each need to prioritise the features they are most necessary for our project and make the most of them.
That would be amazing! AI just needs more control. I 'm hoping that within 2 years this whole prompt engineering infrastructure will be a bygone era
That would be amazing!
Great test of those four. For lip sync you need audio as text won't give you the correct intonation or spacing. It's a shame that Kling doesn't allow importing a video for lip sync as you could then create the emotions in something else and layer the lip sync on top of that in Kling.
That would genuinely be really helpful!
It seems that subtlety is the biggest difference between platforms. Runway is more subtle minimax more over the top. I can more easily imagine runway fitting into an actual film but if you just want something that clearly demonstrates the emotion minimax is better.
That's actually a much better way to describe the differences... wish I'd thought of it that way :-)
I'm using a mix of Runway, Kling, and Minimax now and its working well. Minimax is definitely better at big emotions and big movement, but the generating speed is an issue. I can generate 10 Runway clips before one Minimax or Kling is ready.
But Kling has that fun lipsync which I used in my last music video :P
Nice plug for your music video 😁
@@HaydnRushworth-Filmmaker 😁hope you liked it
Thank you Sir.
Not at all… thank YOU for watching 😁😁
Great upload and execution by Mr. Rushwood. However, best comparisons are candid reveal of cherry-picking attempts for all four platforms. Highly doubt the showcased emotions were all first time raw generations. Given Minimax grants uploader CPP to make unlimited generations, compared to limited credits from say Runway, Kling, or Vidu... it's more likely the relative, better generations for Hailuo Minimax would be shortlisted for final showcase, more so than the other three.
That's a fair assumption, but as it happens, I just didn't have the time to keep generating until I got a "good" one. No, it really, honestly is the case that almost all of these were first generation.
Excellent video. I did something like this only using Vidu and I was happily surprised. BTW, with Minimax you can type sentences in the prompt,but the characters only speak in Chinese. Really.
Hahahah, that makes total sense. At least we're heading in the right direction :-)
You hit the nail on the head right at the end. Until or unless we get the ability to lip synch with this stuff then it's all fantastically brilliant but also kinda useless. I know.. I know... a year ago this was all science fiction and here we are being all moaning and grouchy about it.
One of my biggest peeves with Minimax is that it just can't see a mouth and keep it closed, so it's difficult to generate anything that has a face on it and not have it give the appearance of talking... but anyway... excellent comparison video, thanks so much for putting them through their paces.
Hey, no problem, thanks for watching. I'm glad it helped. :-)
HeyGen can provide lipsyncing. Worth checking out potentially
Brilliant, thanks for the tip-off :-)
How long does it take for MiniMax to create a video?
Not sure. I had a pro-plan access, so they were really quick (just a few minutes each), but the free plan gen times are weyyyyyy longer.
Wow Great vid! I think I Will change my subscription to Minimax instead of Kling. Maybe a stupid question but Minimax is hailuoai right? Or do you have a direct link? Thanks 🤗👍
Yeah, sorry, I really should have made that clear. Minimax is actually the parent company and Hailuo.ai is indeed the real name of the product.
@ thanks for the reply! I tried Runway and Minimax today but wow Minimax takes ages to generate. Maybe because of my free plan? It even stopped generating images and I got a fault.
Hey man, have you been watching this video for a long time? your information is outdated yes, that's right, the voice-over function and lip sync are already available in some AI that you are dimonstrating check again
Do you mean the ability for text-to-video prompts to include dialogue words like “now what are we gonna do?” and then the character in the video mouths those words?
Vidu is the best for emotional expression!
I’d definitely be happy to work with vidu’s results. Great tool. I’m sure you’ll see me gushing about it more in the future 😁
minimax>vido>kling=runway，minimax很夸张但又能抓住重点，有点像迪斯尼卡通【可能给的参照图不够写实】；vido很真实但是在一些情感表达上不如minimax准确；kling不太稳定会崩；runway从头到尾都差不多，没太大变化；
我同意，这是一个非常公平的评估。 《Runway》无疑更加微妙，但就情感真实性而言，我从未遇到过像 Minimax 这样持续令人印象深刻的结果
WORDS for tha actor to say in the prompt ...?
Now you can make your dream film
😉
💯🎉🎉🎉🎉🎉😁😁
Just remember that if you run the same prompt 10 times on the same AI, it could generate very different videos.
💯absolutely!
Which one is free ?
I’ll be honest, I don’t really keep track of the free versions any more. They all seem to have a free version which gives you the opportunity to play, but eventually, if you’re serious about creating, you’ll have to start paying for one of them in some way.
Vidu ai is also first in some emotion but you continuously giving first to minimax I don't know why ?
To be honest, it’s subjective, and I’d genuinely be really happy to use Vidu as well.
whao... What will happen to the huge budget movies industries across the globe in a few years from now if AI can create such deep emotions... uff .. just thinking about the possibilities
I can tell you with confidence that regardless of what the studios are saying publicly, they’re all working feverishly with AI behind closed doors.
@@HaydnRushworth-Filmmaker Yeah, for sure... Just keep on publishing more videos like this General, I am following you like a disciplined soldier and learning from them!
wow amazing
100% really impressive results!
Minimax is the best. But problem with it is, that all these emotions always is animated same way as it was a robotic 3d model under the choosen face skin. So imagine in one scene you make two angry people in one shot and they both will do same "angry mimics"
That sounds about right. Given that AI doesn't (and can't) understand human emotions, mimicking is the only real option it has.
Vidu is so much fun
I agree. I’ve loved using it!
Vidu or Minimax, both did well
I agree. I’d be really happy working with Vidu on a project. In fact, I’ve had terrific results already working with Vidu on one of my projects.
I prefer Vidu here. And it's a surprise.
I’ve really enjoyed using Vidu as well. Great tool.
kling has lip sync
Yeah, I really like the massive amount of Lip Sync options, but I haven't come across a model yet that allows you to included dialogue in a prompt and have the lips move as if they're saying those words... "Let's get outta here!", or "Now what do we do?" etc.
your right, if AI doesn't understand humans, AI Understanding OFF AND ON switches, WAIT! THATS IT!, AI UNDERSTANDS MONEY! 😃👍, NO WAIT!, AI UNDERSTANDS CYCLING HISTORY SEARCH! WHAT A TIME TO BE AWAKE IN THE NIGHT.
Hahahaha, you're absolutely right... time for bed :-)
Can we get a new comparison that includes Sora, Runway and Minimax? Please try to fix the teeth problems, otherwise the footage is unusable.
Absolutely, good call. I'll add teeth problems to my AI tool wishlist :-)