in this video we're going to take a look at the best AI video generators of 2025 I'll take a look at each tool talk about the pros and cons and rank them based on my super scientific french fry scale with curly fries being at the top and floppy crinkle cut fries being at the bottom let's get started the first tool that we're going to take a look at is Sora Sora is an incredible tool developed by open AI when it was announced in 2024 our Jaws dropped it looked incredible but now we have access to
the tool and the question is does it stand up well let's find out so during this testing process I'm going to basically lay out a pretend scene so that we can have some variety to see the Cinematic potential of each AI video generator so we have this establishing shot here it's basically this kind of like abandoned factory setting we have this shot of this man looking over a computer he's just kind of worried right like something's going on and then we have the shot here of basically kind of a SWAT team maybe they're about to
uh get to that researcher and attack him or something and so we're just building tension from just a few shots here and basically I'm going to use each one of those images to test the Cinematic potential of each of these generators so let's get to it to create a video and Sora all you have to do is go to their website click on the plus icon at the bottom and you can upload an image or a video which is pretty cool I'll use this video of a SWAT guy walking towards the door and we'll go
ahead and click open once the image is loaded up now you can type in your prompt for our prompt I'll say a soldier walks down a hallway the camera tracks with the soldier as they approach a door handheld camera so the cool thing about using Sora is you have the ability to adjust quite a few different settings that aren't available other tools so for example you can select different aspect ratios we'll say 16 by9 of course uh you can also change the resolution so you can go all the way to 1080p which is great because
most of the AI video generators can only do 720p you can also change the duration with some of the resolution settings you can go all the way up to 20 seconds it's pretty cool and then finally you can change the number of variations now with the 1080p version you can only create one variant but with 720 you can do two and with 480 you can do four which is really powerful so when you're ready to render go ahead and click create video I went ahead and rendered out the other two clips as well and combined
them together into a scene with some sound effects let's go ahead and take a look copy that okay so as you can see it does look fairly cinematic the establishing shot looks pretty good I like the the dynamic movement I like that the clouds in the background are billowing and it really does feel like an establishing shot we also have the shot here of the soldier walking to the door which is pretty good I should note that we had to edit around a spot where the soldier in the background actually just kind of disappeared so
that's just kind of part of working with Sora it sometimes takes your image that you uploaded and changes it a bit whenever it starts to render but you know looks pretty good the alarm you know kind of Clicks in a little late there but it's not bad and then finally we have the shot of the guy looking up and yeah there's like anxiety on his face looks like he's kind of in a hurry the camera Shake looks very very natural so I think it did a pretty good job now of course the problem with working
with Sora is because it tends to be one of the more creative tools it can also get the strangest whenever you get into your generation so for example here's one clip that we had and you can see that the background character like disappears behind some random door and then that door closes and then he is approaching this door and same thing over here he just kind of covers up the guy and then a huge door appears which is uh just kind of weird so a lot of times when you're working inside of Sora that type
of stuff can happen and it can really eat into your generation credits the strangeness that you get from Sora can be extra annoying sometimes because the tools also pretty expensive the most expensive version is $200 a month which is definitely cost prohibitive for a lot of Independent Artists on the positive side Sora makes it super easy to upload images and videos and you do have the ability to key frame your videos using a storyboard feature which is very helpful when Sora hits it hits incredibly well and the video quality can be incredible so it's definitely
an amazing tool for getting really good results and the fact that you can render 1080p natively is really nice of course you probably want to upscale that using an outside tool but the fact that you can render the core video in 1080 is really awesome now the cont to using Sora is exactly like I said it's really expensive which is cost prohibitive for a lot of folks and then the fact that it can tend to cut to random shots and really just waste a lot of time that can be a challenge you also don't have
the ability to upload photos of people if you aren't on the most expensive plan so again that makes it really challenging if you're an independent artist using the tool so for all of these AI video generators I'm going to rate them based on three categories the first is realism the second is camera and character physics and the third is workflow and ease of use so for realism inside of Sora I would say eight out of 10 it does a really good job it's not perfect and even the best generations of Sora if you really begin
to examine them there can be some strangeness that you begin to uncover but it does a really good job now for character and Camera physics I will say that because Sora can get really strange really quickly and it tends to want to change your scene again and again I have to give it just a 4 out of 10 and for workflow and ease of use I'm going to give it a 5 out of 10 because while the platform is incredibly easy to use and intuitive the fact is the content moderation filters are too extreme with
Sora and it makes it really challenging when you're working on a film project so that gives Sora a total score of 17 out of 30 putting them into the sweet potato fry level next up we have Google vo so VO2 was announced recently and the quality of the results that we've seen from the tool are incredible let me show you how to use it so this is what VO2 looks like I'm in Early Access so this isn't publicly available so I'll move quickly here weird thing about using vo is the fact that you don't have
the ability to upload your own images at this time you have to create the image on their platform and then you create the video so that can be really tricky especially if you're looking for character consistency lighting consistency color all of those things so you should definitely know about that limitation and it is going to make it incredibly hard to create films with the tool in its current capacity but that being said let me show you the workflow here so I have basically a prompt where I'm going to create an image of a soldier walking
down the abandoned hallway let's go ahead and create that image okay we have an image here if you wanted to you could start over I'm just going to go ahead and move forward we'll go to motion here and I'm going to say a soldier walks down a hallway the camera tracks as the soldier approaches the door and we'll go ahead and create that video I did the same thing for the other three shots let's take a look at our result copy that it's pretty wild I mean the establishing shot looks great that shot with the
SWAT guy looks amazing and then this is probably one of the more expressive AI videos that I've seen up to this point but the problem that I'm sure you're noticing here is the three shots feel completely different they feel like they're from three completely different films and really no matter how much prompting you do it's really tricky to get the consistency without more advanced tools for doing style consistency location characters all of those things so that's a huge limitation to using vo however the physics engine that's powering vo is probably the best out of any
AI video generator so for overall realism I would say with vo 10 out of 10 it is the best AI video tool for creating realistic videos for character and Camera physics again I would say 10 out of 10 I think it does an amazing job not only with camera Shake tracking shots all of those things but also with the actual expression of the characters directing the movement all of those things come through when you're working with vo and it's incredible all of that being said for workflow and ease of use people really don't have access
to vo right now so that's not very helpful and then it's also very tricky to work on a scaled up project inside of VO especially because you're going to want consistency you're going to want that film to feel like it lives in these same world and universe and that's just really tricky so that gives vo a score of 23 out of 30 and they are now in the waffle fry level all right good job moving on let's take a look at Runway Runway is an incredible company that really supports the AI film making community and
they have a ton of tools outside of just AI generative video alone you can use runway for lip syncing in painting there's a ton of tools that are really interesting to create a video inside a Runway all you have to do is go to generative session and you can upload your image or video and of course we can type in a prompt here I'm just going to say camera tracking shot Soldier walks down the hallway and you can of course change the model that you're using gen 3 Alpha is their best model at the time
and you can change it from 5 Seconds to 10 seconds we can just do 5 seconds for now and go ahead and click generate so let's watch your scene copy that okay so you can see there are things to like and also things not to like with this overall scene here uh so if we kind of go to the beginning here the first thing you're probably going to notice is a severe loss in resolution now it's really interesting because a lot of the AI video generators produce results that are around 720p but for some reason
the Fidelity of the generations from Runway really don't look great A lot of times uh for example there's a lot of really fuzzy areas you see here with this establishing shot that if an audience was watching this they would really uh not be very forgiving uh for those uh changes in resolution there you also have some really weird physics so you see this character here he's kind of gliding in a weird way this guy in the background he looks pretty good although he's not in much of a hurry but that's okay uh but this guy
he kind of seems like he's gliding and a bit like connected to the camera as opposed to his like natural footsteps so that's a little weird and then the final shot here of the guy looking up I I like that Dynamic camera movement obviously there's not a lot of expression on his face it is kind of flat so you'd probably need to use Runway act one to like get a little more expression there uh but it it did an okay job the thing to like about Runway is the fact that it's really quick you can
generate like 20 video clips in the time it would take you to generate one inside of cing or minia Max and you also have the ability to use all of those features that I talked about that go Way Beyond just the video generation alone so I really recommend using Runway if you are a filmmaker but let's get to the score for overall realism with Runway I would say that it's kind of just a five out of 10 I was really thinking that the results would be better but it seems like the algorithm has really stagnated
recently hope hopefully there will be new announcements soon but at this point it's really beginning to fall behind some of the competitors for character in camera physics again I have to give it a five out of 10 the characters inside of our scene some of them looked good others the physics just didn't look right and the overall camera movement just didn't really make a lot of sense especially with that second shot and for workflow and ease of use I'll say eight out of 10 giving Runway a total of 18 out of 30 putting them strongly
inside of that sweet potato category next up we have Luma Luma dream machine released in summer 2024 and it really was one of the first AI video generators that existed in this kind of gen 3 era the results that we got from Luma far exceeded the other results that we were getting from other AI video generators but the question is how does the Luma hold up today Luma is incredibly easy to use all you have to do is click the plus icon on their website go to the image and upload load your image you also
have the ability to set an inframe so you can interpolate between the first frame and the last frame and it'll basically animate between them which is pretty cool all we have to do is type in our prompt and go ahead and click render I put all the clips from Luma together let's go ahead and take a look copy that okay so as you can see it's not super real istic an audience would not fully buy the scenes that we're getting from Luma I do think the establishing shot looks pretty good but that second shot with
the soldiers here I mean it just looks like this guy in the background is being the creepiest little creepy SWAT guy ever and this guy while the initial turn looks good it just it it's coming across a slow motion which a lot of the AI video generators will give you slow motion footage and that's not really helpful if you're working on a project that's supposed to have realistic Dynamic physics and then the final shot here of the guy his eyes looks weird it looks like he's like a weird skeleton guy uh I think was it
X-Men where the guy like blinked in two different ways it kind of looks like he has like a weird blink going on right now so it really just did not do an amazing job so while Luma is fast and easy to use and their API is also incredibly quick the fact is that realism is really bad you tend to get a lot of slow motion video clips from Luma and the physics just aren't really that realistic so for overall realism I'm going to give Luma a two out of 10 for character in camera physics again
a two out of 10 and then for workflow and ease of use I'm going to say five out of 10 the actual interface is incredibly easy to use but when you actually get into the resolution that you're getting from Luma the overall Fidelity the quality of the generations and then also just the feasibility of the tools that are there the first frame and last frame feature is great in theory in actuality it doesn't actually work out so good most of the time so I'm actually only going to give that a five out of 10 so
that means that the total score for Luma is 9 out of 10 putting it in the steak fry category although some french fries with some A1 that would probably be pretty solid next up is minia Max also known as hilu it is a chinese-based video generator that popped up in 2024 and the results are pretty impressive let me show you how to use it to use Minimax all you have to do is go to the hilu AI video website and go ahead and click quick create from here you can drag in an image and we'll
paste in our prompt and the cool thing is you do have the ability to generate up to three video clips at a time which is pretty cool and go ahead and click generate so let's watch back our results from Minimax copy [Music] that so as you can see there's some things to like and some things not to like about this the first is the overall movement of the camera especially on the establishing shot it's not super Dynamic and the things that should be moving for example the clouds in the background there's really just not a
lot of movement it it's okay but it's not the best you also have the second shot here and let's just note this guy who is just being a like a sneak King over here he's just like sneaking around creeping in the weirdest possible way and I should also note that we prompted for the soldier to basically be walking towards the door and he really just kind of like checks his gun and like looks around there's really just not a lot of movement there so not the best result from that shot and then finally we have
this shot here and you can see that the man he's kind of talking here I don't really know what he's saying we didn't prompt in for talking but for some reason Minimax really likes making everyone talk so again minia Max is incredibly easy to use and the physics are decent especially if you have shots that aren't getting enough movement from some of the other video platforms Minx tends to give you the most overall movement and so it can really just add life to your animation just think about it as H giving your subject some coffee
that's kind of what you're doing whenever you use minia max now the con to using Minx is the fact that it the overall realism is just not amazing compared to the AI video generators that are now available it used to be amazing just a few months ago it really was one of the best tools but as it stands right now it's not giving us the best results and the overall resolution is also a bit lower from miniac so you're going to have to upres the result and then also probably do a little bit of sharpening
because it does tend to be a little soft so for overall realism from minia Max I'm going to say seven out of 10 for character in camera physics we're going to do 6 out of 10 and for workflow and ease of use it's incredibly easy to use we're going to say8 out of 10 which gives Mini Max a score of 21 out of 30 and they are officially a sweet potato next up we have cing like minia Max clling is a Chinese video model but they have many more features that are interesting you have the
ability to train models on someone's face so that you can have them in the video again and again there's a lot more control over the videos that you generate and there's just more settings that can adjust including negative prompting to get exactly what you're looking for to create a video inside of cing all you have to do is go to the AI videos button and you can drag and drop your image and I'm using the latest version of cing that's 1.6 again you type in a prompt you can also change the duration the number of
generations and then also the creativity strength which is pretty cool and then they do have a spot down here to type in negative prompts so things like Blurry Distortion weird all of those words that you typically would want to include in a negative prompt you can paste down here and when you're ready of course you can click generate so let's take a look at the results here copy [Music] that okay I got to say I'm really impressed from the result from cling here I think that it really embodies a lot of the Cinematic Aesthetics and
elements that I would want to see from a video generator for example if we go back to the beginning here you can see that the smoke is billowing from this Factory and we have that nice movement of the overall scene which is cool we also have this second clip here it's pretty interesting because you have the soldier that holds up the gun and then it rack focuses on this guy in the background and his movement if you compare that to some of the other tools the guy in the background's movement is very realistic it really
looks like there's a lot of weight behind it steps obviously we'd probably want to prompt in a little more urgency from the way he's approaching that door but it is what we prompted and the actual added camera movement looks good and I should also note that on the edges here you're seeing some chromatic aberration it's taking on Lind Dynamics like that's really uh a very cinematic and uh a much more detailed result that we're getting from cing versus other tools like Luma you couldn't even begin to think about chromatic aberration like that because the resolution
is already so soft that that's that's a something to focus on in the future whenever their model gets as good as clean and then we have the shot here of the guy anxiously looking at his computer and I should note that when we prompted in for you know someone anxiously at their computer the system actually understood what anxiety is when it's captured on camera there's usually a pushing in feels like the world is closing in around them and uh really the overall uh cinematic result helps to reinforce the story that we're trying to tell so
it's kind of like having a bit of an AI Director of Photography helping you out during the production process here so for overall realism I would say eight out of 10 for cing it does a really good job most of the time character in character physics again I would say eight out of 10 and then for workflow and ease of use again eight out of 10 so eights across the board that gives cing a total of 24 out of 30 putting it in the waffle fry category and then finally we we have hyper 2.5 hyper
is a super fast and easy tool to use but can it stand up to the other tools let's find out so I'm on Hyper's website and all you have to do is go to image to video and let's go ahead and upload our image so our image is uploaded and then we have our prompt here so you do have the ability to change the resolution if you want and also the duration up to 6 seconds long and go ahead and click create let's take a look at our final result [Music] here copy [Music] that okay
so as you can see the result from hyper is not amazing the first shot is just incredibly static there's just no movement whatsoever the second shot I don't know what's happening this guy's not even moving and this guy the the weight behind the guy's steps is just strange he's like floating and then it just like randomly Zooms in for some reason at the end so no idea what's going on there and then finally we have the shot of the guy kind of rising up and I suppose it's okay the only problem is when he blinks
it is like it's like too fast and too robotic it's like it didn't blink for enough frame so it looks like he's I don't know some sort of like robot character that's been programmed to Blink as opposed to a real uh video clip here so I just don't think it did an amazing job so while I definitely think hyper is a very fast and easy tool to use the truth is it's very unrealistic and the physics are honestly pretty bad at this point so for overall realism I'm going to give hyper a one out of
10 for character and Camera physics I'll give it a two out of 10 I think that last shot was okay but not the best and for workflow and ease of use I'm going to say two out of 10 mainly because the overall Fidelity and resolution of those final videos that you get they're not amazing and there's not a ton of advanced features on their platform so that gives hyper a total of five out of 30 putting it unfortunately in our floppy crinkle cut category if you want to learn how to get maximum cinematic results from
prompting to video generation to even compositing upscaling and distribution be sure to check out our AI film making course over on Curious Refuge we would love to have you inside of the program and of course if you want to get the latest AI tutorials and news directly to your email inbox you can subscribe for free by going over to curious Refuge just click that blue button in the top right corner and you can fill out your email information and then you'll get our Weekly Newsletter which has the latest news you need to know about and
then finally of course you can subscribe here on YouTube to get the latest tutorials and AI news directly here on the platform before we go I just want to say that the thoughts and prayers of everyone at the Curious Refuge Community goes out to people affected by the fires in Los Angeles I know we have a ton of members of the Curious Refuge community in La so we hope you are safe please reach out to us on Discord to let us know that you are doing well this community is incredibly loving and supportive and when
things are safe we would love to get together again and see all of you thank you so much for watching this video we will see you in the next one
