Kling AI 1.6 is HERE: BEST AI Video Generator of 2024? (Prompt Battle)

in this video we will take a detailed look at the recently released clink AI version 1. 6 we will compare the new model update with previous version 1. 5 in the prompt battle we will test clink 1.

6 against Sora Runway Genry Minimax and who you on and as we approach the end of this year I will share my personal ranking of significant video models from 2024 let's get started firstly if you want to access new clink version you can click on AI videos and select version 1. 6 from the model selection drop down on top this new model promises greatly improved prompt adherence for more consistent and dynamic results in this first example I'm focusing on prompt understanding and I'm testing version 1. 5 and version 1.

6 to understand the Improvement or if there's any difference just as a reminder I'm using the official prompting structure recommended by the clink prompting guide and the structure they also use in their examples as well also another rule that I'm not rrolling these prompts I'm bringing the results from first ever try so I wrote woman is in the Metro while camera is pulling back for the first example version 1. 6 did a good job it understood the camera movement in the first track whereas version 1. 5 struggled a little bit it didn't really brought me the camera pullback the camera motion was hit or miss with version 1.

5 most of the time it was a Miss but I'm positively impressed that in the first try 1 Point six managed to bring me the camera motion the next example is about a ballerina dancing in an old castle here I wanted to observe how good the new model is in comparison to the previous one when it comes to motion and action so again I'm using same prompt for both models and you will realize the ballerina in the version 1. 6 is much more Dynamic and active the motion looks smoother and natural and 1. 5 output in comparison looks really static again these are first try results and I'm already positively impressed with version 1.

6 in this next example we are testing prompt understanding again and I'm trying to understand the coherence and how closely and consistently camera can track the subject here you will observe a positive Improvement again with version 1. 6 where we have pretty coherent result where this robotic firefighter is carrying the dock on the other hand 1. 5 cut little bit weird and they kind of enter to the burning house and dissol there I can say there is a clear imp improvement with the new model here and in this last example again I'm focusing on action and motion while prompting for clink we always start with describing the scene first this is recommended by their guidelines and in the second part of our prompt we are describing either the event happening in this scene or the camera motion or cinematic motion that we want to see in the shot so in this example you will realize that both models struggled little bit with the final swing of the katana I would argue that version 1.

5 struggle much worse whereas you can see version 1. 6 did okayish job until the end and in the end it struggled very severely but just trying this four quick prompts you will realize that there's a significant Improvement for 1. 6 so kudos for clink team I think they did a good job here now let's jump into prompt battle and we will start with text to video in text video I'm comparing five models clink 1.

6 Sora Runway Gentry minia Max and who new one the first prompt is famous stylish woman walks down a Tokyo Street you can see the full prompt on the screen I'm running same prompt for all these models I will start by the worst one from my perspective Sora looks the worst on the runway output I see little bit of coherence issues particularly on right hand carrying the back minimix output looks surprisingly colorful and saturated in this particular output that diminishes the realism for me a little bit but but if we focus on the Walking the motion and the fluidity of action it looks really good I think two of the best jobs here done by clink 1. 6 and huan especially cink's world looks really alive even in this short clip you can see people are walking around unfortunately I had a shot from back but if we solely focus on walking it looks really natural and correct I think coherence is really high andan result it's just looks so realistic almost like recorded by a real professional camera it's not as alive as the clink output but realism looks really whole new level so I'm really happy with the huan result and clink result I would argue that for this first test it's a tie between clink and huan next prompt is a difficult one that I wanted to a little bit challenge the models we have an otus Siege of Istanbul it's an historical prompt where we are looking at the many things a lot of stuff is happening in this prompt it's a long one we are looking at photo realism prompt adherence World understanding all together in this particular example I was impressed with the Sora output that it looks cinematic and realistic so thumbs up for Sora for this video and of course the closeups were not part of the deal I didn't mention any closeup in my prompt so if you look holistically the output didn't deliver what we asked for but there are bits and pieces and small frames that looks really impressive the runway output looks the worst from my perspective this just looks like a modern day stumble and it has nothing to do with what I asked for very poor result the Hunan output managed to render the realism and of course you have walls and um you have Soldiers the flags and Genies are not represented here so World understanding wise what we asked for isn't here but it's a realistic looking video at Le I can give you that here I would argue that Minimax output and clink output come closest to what we ask in our prompt it's a historical scene you have soldiers in front of the walls at least they have the traditional red outfit of gen series and minimex output again looks really a little bit cartoonish it's not as realistic as huan output but at least there is some kind of world understanding looking at huan output with weird physics and just the soldiers and horses are moving abruptly right or left I would argue that clink and Minimax are winners of this challenge yet it's difficult to say one model out performed the other one dramatically next prompt is hand shuffling cards touring test for AI video models simply all models are struggled with this prompt it's a very good prompt to test these models I haven't seen any AI video model who passes this test so easily except Google V2 the clink output looks really strange the anatomy of hands looks good but the cards look strange they look more like a brochure or I don't know they they don't really look like the playing cards Sora output is just pure chaos the runway output morphs and kind of like changes shape it's just weird punan output is ultra weird as well even though it's not perfect I think the Minimax output comes closest to we could say acceptable quality here the next challenge is my and yours favorite yes it's a relable prompt with a massive meme making potential you can generate so many memes for AI video models using this challenging prompt so we have Vol bable match between two professional teams it's not Beach oall this time and struggle is real let's start with clink output here it managed to put some kind of net at least between players there are two balls and one of them morphs and it has like strange physics it's just not working the Sora output I think there is not much to add here yeah again the time moves weird multiple balls you have actually one net disappears another net comes into the scene Runway Gentry completely ignored volleyball part or let's say merged basketball with volleyball field together and created a new sports Minimax output has at least some kind of world understanding this player is unfortunately morphing so heavily so it's really making this shot unusable you can see it's really like morphing like crazy I think for this challenge huan comes closest to what we say is acceptable of course you have some issues here again kind of like a strange frame jump and the jerseys doesn't make sense right and you have a mix of teams in the pitch but I would argue huan comes closest to an acceptable result here in this challenge I aimed for something cinematic a chain of actions where the subject follows a sequence step by step so we have a detective walks into frame shows his batch and then camera zooms to the batch for this particular prompt huan and Runway gen 3 struggled quite a lot I'm not Ultra happy with the results came from these models so I would probably put them aside for this particular prom I think clink 1. 6 did a good job except the badge looks a little bit weird and not really like a detective badge as we know it I like so much that clink 1.

6 managed to follow prompt step by step the whole sequence walking into frame showing the batch is here the S output looks Ultra cinematic as well you can see that the person walks into frame shows the batch and we have a zoom to the batch and minim maxer took an interesting approach here and got creative and I think did also a fantastic job for this particular prompt I think clink and Sora both did a great job then clink 1. 6 is little ahead in terms of prompt following but Sora output definitely didn't disappoint here but if we say let's pick the best prompt following output I would say definitely clink and with this we are finishing the text to video comparison and now we will jump to image to video by the way if this video is helpful for you so far please give a thumbs up and subscribe to the channel for image to video again we are taking the first output without rerun and for comparison now I'm dropping H Newan because H one doesn't have image to video mode yet for the first challenge I have a train speeding up I'm trying to understand how well model is able to understand what's in the image and create emotion out of it I think all models did a good job here except Sora among all these models sora's output looks by far the worse I like the most personally clink output here because this is exactly what I imagined the runway and Minimax output is also good except you have a bit of a morphing for Minx output there and generally historically when it comes to action and motion way is just creates extremely slow video footage it's just moving so slow so it doesn't really give you that strong the natural feeling of motion that I could only take care from clink here it just feels like Sora or I'm guessing this model is actually sort turbo it's not the full release this just this really feels like a inferior model in comparison to other state-of-the-art models especially in comparison to clink 1. 6 in this next example we have a couple of different sub objects we have two dogs and one woman walking and I would love to run this on Sora as well but unfortunately being able to use images with people in it is behind the pay wall so therefore I couldn't run this with Sora I will start with the worst one again the Minimax here so action the walking motion it's very natural and it looks great but we are seeing strong coherence problems with the face so face looks really blurry and we are losing the coherence here when it comes to best output for me it's clink 1.

6 without adopt the runway output is great as well but it's not as good as clink output and again you have this slow motion thing it just doesn't flow you know and in Runway output for the dog on the right side you will see an additional leg appears so winner here is clink in next video I'm starting with this image and I would like to test the camera orbiting around my subject so again I started with describing My Scene little bit we have a woman watching the city from a high angle while camera orbits around her I think all models here delivered a great result a usable result I personally found the clink output the best because I also like the wind in her hair but looking at Runway output and Minimax output I think they all did a great job here visually I like the clink output the most but I will accept all three of them as winners here this video is actually funny and I will be honest with you I tried two times with two different prompts here actually I just wanted to have a simple cinematic Focus change from man's face to the gun or vice versa but I struggled with that with all models I couldn't have that nice cool cinematic Focus change so then instead of that I decided to test simply camera pans right and I hope that while camera is panning the focus offer changes strangely Minimax gave me the shot the rifle fires I didn't ask for for it Runway Gentry Alpha gave me a really nice smooth camera pan but unfortunately it added an additional Barrel a wooden sniper to the end of the rifle of my character so that's really nonusable here again I would argue that here clink 1. 6 did the best job even though it's not the best camera pan I will still accept it relatively looking at the other results however the Minimax output the rifle is firing and after that you have a nice Focus change focusing on character's face it's very usable it's very cinematic shot I think it would be fair to say that this challenge is a tie between clink and Minimax so in next video we have an cinematic action scene and here Minimax again started with firing rifle that's something we didn't ask for in our prompt and yes we mentioned a shootout but we didn't specify that the rifle needs to fire here Gentry Alpha output looks weird the legs the movement the motion looks really off from my perspective so for my own AI film I wouldn't use this output again here clink 1. 6 gave the most cinematic best shot as a clear winner here in comparison to other models in this next challenge we have an event happening in the scene we are starting our scene with an ancient city and then we have a flat hitting the city unfortunately Runway just didn't get the flood and couldn't reflect what we asked for and people's movement looks natural but it's not what we asked for here we got a fantastic result from Minimax as I mentioned few times in my videos Minimax exess with this kind of things if you have an event happening in your scene when you start with image to video Minimax generally always delivers a good result before clink was struggling with this but now I see Clink started to catch up as well so even though it's not a massive flood rendered here still managed to give us what we asked for I'm really in between Minimax and clink output here but I personally feel Minimax output did a better job here so therefore I will choose Minimax over clink in this next video we have a viking and we would like to understand in which model camera follows the subject best I think Minimax did a great job here only in the first frame you have a bit of a Time freezing problem and it almost feels like director said action and then actors started to walk so that's funny and this happens with minimix sometimes the initial shot kind of freezes for like 1 seconds and then it starts again I like here that the runway output looks really cinematic like the the water physics look fantastic but again it's not what we wanted and the clink output here looks really good but the camera doesn't follow the subject it stays static so it completely ignored what we asked for so for this challenge clear winner for me is minia Max probably you need to cut the first second of the video though in this next challenge I didn't write any prompt I I just decided to give the initial image leave the prompt empty and see what happens and which model does the best job without prompting anything again starting with Runway I don't know why it's always slow motion in this video I'm definitely Laing little bit of dynamism I think here minimex output would be great but this weird red UFO appears in the end that's kind of strange and that kills the shop for me I'm afraid to keep repeating myself but clink did the best job here with dynamism people walking rendering the smoke we have characters walking in different speeds we have UFOs slowly hovering in the scene it looks so alive and dynamic and I would say clink is the clear winner here in the next challenge I'm testing the Human Action motion anatomy and physics we have a simple prompt she gets on the horse I think it's a simple yet effective way to test models uh minimex output is not usable because a second horse appears from nowhere Runway output is cute there's a nice cinematic Dolly out and being fair it managed to keep the coherence of the scene and character but it's not what we wanted here again clink like a soldier really loyal to the prompt woman gets on the horse a physics and kind of like how she's climbing is little bit weird and little bit strange but clink delivered what we ask for with relatively okayish looking physics next challenge is similar again we would like to see that our monk gets on the dragon or since there are no wings here we could say kind of like a big lizard it's again a sequence of chain of actions so we want first to see him walking towards the white dragon or white lizard and then gets on it the Minimax output is a little bit strange where he kind of flots Runway output again looks very cinematic the lizard doesn't move at all so it looks very static you know it's a very nice coherent over theh shoulder shot but again it's not what we asked for and clink I'm kind of tired of repeating this but did the best chop here our character walks towards the lizard physics looks okay it's not the best but at least it climbs over in the next challenge again we are testing the physics understanding of the models so we have a woman gets in the car starting with Minimax output how the doors of the car opens in the same time and she walks to the front of the car it's just strange Runway output is also Ultra strange where she looks like she's trying to run away from the car where car is driving over her this is the clink output first try again she opens the door and in the middle of the shot video stops because 10 seconds limit so in comparison to the other two it looks much better I think still it was going to struggle a little bit with the end of the action that I'm not super sure she would be able to really fully get into the car but I will accept it as a winner next video I wanted to test motion and action again and physics a minimix generally ex us with this kind of videos of course if you ignore that the coherence change of the face probably this is something they need to fix because model does an okayish job generally with physics and motion but the coherence of face changes way too much so they probably need to improve that Ramy output started well but then character changes she becomes smaller bigger and then a sunglass appears in between so we completely lose coherence of the character here unfortunately it's an unusable shot here again clink did the best job here like waves and the characters motion with the surfboard the coherence everything looks good it would be of course nice to see what comes out from that wave if model would still keep the coherence until the end that would be a big surprise for me but again clink did the best job here I don't know I kind of started to feel like I'm just repeating myself here in the next challenge we would like to understand if models understand the hand gestures and I just want to try something simple like hand heart gesture and most of the time AI video models are good with it so Minimax nailed it a perfect output here exactly what we asked for Runway I don't know a completely unnecessary dolly in here that it's not something we asked for completely ignore my prom and clink made a heart here so I'm like where is the rest of that so Minimax is the clear winner here for this challenge in this next challenge I'm starting with this closeup portrait and I want drone camera to move up here Minimax actually understood what I asked for but then coherence completely failed my character started to fly Genry Alpha switched to low angle it's not what we asked for unfortunately all three models failed for this prompt in the next chenge I have an event happening in the scene I'm starting with the simple my character is walking in the jungle and I have a tiger jumping into the scene so Minimax tiger looks really strange this whole shop looks weird and gentry Alpha gave me a nice tracking shot completely ignored tiger you know what you don't need tiger here's a nice cinematic tracking shot for you and clink gave me the Tiger in the end in the very end well we could say we could accept that clink delivered what I asked for even though maybe it's not the best most usable shot here you will realize there's also a bit of a struggle with the scale here like woman gets smaller and smaller almost like she shrinks in the shot and tiger is just way too huge the next challenge is about understanding how models reflect the human emotion and render I think both Minx and cling did a fantastic job though clink output looks much more dramatic where Minx output is little a bit more subtle and yeah as you realize Runway just couldn't render it so yes she looks sad but she's not crying I'm between cling and Minimax for this particular challenge but to be fair with you I think clink won this one again as you understand we have a great model with clink 1.

6 it's definitely a massive Improvement I want to test this to Wi as well and I know the upcoming comments that yes vo is not publicly available I agree uh so we have clink already right it's publicly released we can use it today so it's already amazing so just for fun we can test it against Google's wi2 so I want to start with famous tomato cutting scene this is the clink output the beginning of the shot is unfortunately unusable and yes VR shot looks really like NextGen here and if you cut maybe the beginning of cling shot it's usable but still it's not as good as VI unfortunately in this next prompt we have a cat Roars while looking at its reflection in the mirror instead sees itself as a lion roaring I think this is a really nice challenging prompt because you need to render cat lion the whole reflection Roar many things happening and this is a good prompt for testing AI video models I think clink here got a really difficult time it kind of morphed cat into a lion and made like a very big cat whereas same prompt in vo rendered extremely accurately I would say Google's VO2 output is shockingly good for this difficult prompt and lastly giraffe gymnastics another fantastic prompt I mean imagining giraff gymnastics is even difficult for me as a human yet for AI video models and as you can see both models struggled here it's a difficult prompt and I like it so much I think both models honestly struggle with this so it's good to see you know that Google VO2 has a limit as well but if you look at like coherence and generally photo realism video still VO2 I would say is superior and many people in the comments ask for my personal choice of AI video models so I hope this section will be a great Consolidated reply for everyone my personal ranking of video models as now looks like this so if you are watching this video I don't know in March 2025 things can be quite different because this space is moving so fast but this is a snapshot from December 2024 of course obviously based on what we saw so far the Google V2 looks like the best model but it's not available yet for public only selected people can try it and therefore I grade it out I can happily change it once I have access to it but for now I can't use it therefore even though it looks super promising and it looks NextGen I can't Clearly say that this is the winner for 2024 I would like to give overall best available model award to clink 1. 6 it overall offers the best quality and you also saw in the compar comparisons with other models it has also decent lip syncing it is going to satisfy you overall with its capabilities I think it will offer you best value for your cash and since I repeated so much I feel urged to say that this is not a paid partnership or anything I'm just trying to put like an objective test to help my community to make best decision with their cash I don't have any organic connection with clink team so to the third place I put both Minimax and Runway Genry I placed them together because it's really up to your needs meaning Runway is optimized for image to video so it really lacks in the text to video so as a Creator if you are using more text to video I think you shouldn't go to Runway the cool thing about Runway prompting is you don't need to describe the image in the image to video mode you just need to explain the cinematography which makes the runway for me one of the best cinematic AI video tool when it comes to controlling the Cinematic camera but you want to control the subject control the action control the events happening in the scene then I would recommend you to go into the direction of Minimax it's it's a fantastic tool when it comes to controlling the subjects and action if you want to control the Cinematic camera go to the runway and Runway also offers a very decent lip syncing tool I made videos about act one already the problem with minimex is face coherence problem right and it's there for a while and they didn't solve it yet meaning that face changes so much especially if it is far from the camera if they fix their face coherence problem then they will be a serious competitor for clink if you say look I want to control both cinematic camera and I want to control my subjects in the scene and action and events well unfortunately we don't have a tool for this yet every tool has his own advantages and disadvantages maybe clink 1. 6 and V2 comes closest to control everything type of AI video tools a few words about speed because this is something like people really care today as December 2024 Minimax for me is the fastest tool right now clink 1.

6 Generations at the moment take around 10 minutes for 5 seconds videos and 14 minutes for 10 seconds videos when 1. 5 released first time it was exactly same problem it was really slow and then by time they made it much faster so if you go back to my clink 1. 5 video you will see in the comments many people were complaining about speed problem and later this complaints diminished I think it's going to be same with 1.