AI Video Generators Ranked from Worst to Best (2025)
By Youri van Hofwegen
Summary
Topics Covered
- Open Art Unifies Fragmented Testing
- Sora 2 Sets Realism Benchmark
- Newer Versions Crush Older Siblings
- Audio Transforms AI Video Immersion
- Avoid These Three Total Flops
Full Transcript
Lately, it feels like every single day there's a new AI video generator being released. There are literally dozens of
released. There are literally dozens of them online, and it can be impossible to pick out the best one, or even one that's actually worth your money. So,
after getting disappointed by yet another AI video generator, I decided to test them all because I wanted to know which of these tools actually deliver and which ones should never be used. In
this video, I'm testing 16 of the most popular AI video generators side by side, using the exact same prompt, and ranking them from worst to best. The
tool I'm using to rank all these AI video generators is called Open Art. And
the reason I'm using it is because it lets me access all of them in one simple workflow. I don't have to go out there
workflow. I don't have to go out there and buy a bunch of separate subscriptions just to test each model, which honestly makes this whole process way easier. So, with that being said,
way easier. So, with that being said, we're going to start off by testing the Cling family, which has become one of the most popular and consistent AI video generator lineups out there. The newest
version, Cling 2.5, is known for its movie-like movement, high realism, and surprisingly affordable price. I'm
already in the prompt field for Cling 2.5. And the prompt I'm using for all of
2.5. And the prompt I'm using for all of them goes like this. A young marine officer stands on the deck of a wooden sailing ship under bright midday sun overlooking a calm turquoise sea. White
sails billow gently in the warm breeze.
Seagulls circle overhead, light reflecting off the water. camera slowly
tracks around him from a low angle to a midshot. Warm golden sunlight, crisp
midshot. Warm golden sunlight, crisp photographic realism, cinematic wide shot. I decided to go with this prompt
shot. I decided to go with this prompt because it basically includes all the main things you'd want to get from your video generator. Realism, nice textures,
video generator. Realism, nice textures, and just an overall pleasant outcome.
So, I'm just going to go with 10 seconds for this prompt right here. Then, I'll
click create, and let's take a look. And
this came out looking stunning. The
realism here is spot-on. Our character
looks super realistic and the seagulls look great, too. When you add in the fact that it's one of the cheapest options out there, I've got to rate Cling 2.5 as S tier. Next, we have Cling
2.1, which was the flagship model before Cling 2.5 came out. First, I'm going to switch to 2.1, which is really easy to do in Open Art. Then, I'm going to just click create, and let's take a look.
Even though you can spot a few small differences like slightly weaker realism and a bit more color saturation, it still does a really good job overall because it costs nearly four times as
much as Cling 2.5. I'm placing this one in the B tier. And finally, we're testing the oldest one in the lineup, Cling 1.6. This model's been around for
Cling 1.6. This model's been around for a while, and you can tell from the interface. It's even got some odd
interface. It's even got some odd toggles like creativity level. I'll bump
that up and click create. Let's see what it can do. And as you can already see, the results just don't hold up. I
honestly can't even tell what's happening in this generation. There are
glimpses of what I asked for, but overall it's just not as good as the other ones. So, this is one of those
other ones. So, this is one of those models I would definitely skip and I'm putting this one in F tier. For the next tool, I'm going to be testing out Cance.
Seance is less known than Clling, but still a good AI video generator that was built for creating multi-shot videos.
And while it used to be pretty competitive, it's fallen off a bit over time. But, I still wanted to see how it
time. But, I still wanted to see how it holds up today. I'm going with the pro version here, maxing out all the settings. Clicking create and let's see
settings. Clicking create and let's see what we get. Seedance actually manages to surprise here because the result looks really good. The only problem I have though is the seagulls flying in the background. It doesn't look as
the background. It doesn't look as realistic as it did with Cling. Aside
from that, everything looks really good.
The textures, the lighting, the color grading, and it rendered really fast, too. So, for that reason, I'm placing
too. So, for that reason, I'm placing Cadence in A tier. Now, we're really bringing out the big guns because the next one we're testing is Sora 2 by OpenAI. Sora 2 is one of the newest and
OpenAI. Sora 2 is one of the newest and most advanced models out right now.
Known for its incredible realism, smooth motion, and even integrated audio and dialogue support, it's a high-end model that costs quite a bit per generation, but the results usually speak for themselves. I'm running this at its
themselves. I'm running this at its highest settings to really see what it can do.
[Music] and it looks stunning. The built-in
sound adds a ton of immersion, especially with how the surroundings move. And the way our marine looks
move. And the way our marine looks around is really well done. It really
stands out above everything else we've tested so far. So, I'm putting Sora 2 in S tier. Now, let's move on to the Juan
S tier. Now, let's move on to the Juan family of models. Juan is a Chinese-made model known for focusing heavily on realism and fluid motion. And there are three versions currently available: Juan
2.1, Wan 2.2, and Wan 2.5. Each version
is meant to be better than the one before it. So, I'm curious to see how
before it. So, I'm curious to see how much it's actually improved. I started
off with Juan 2.1, which is one of the older versions. The generation came out
older versions. The generation came out surprisingly good. The realism is there,
surprisingly good. The realism is there, the motion feels natural, and the lighting stays consistent across the water and deck. It's definitely not the most advanced in terms of detail, but for being an older version, it holds up
better than expected. I'd actually place one 2.1 in Btier because it gave me a very solid result. Next, I tested one 2.2, 2, which was supposed to be the upgrade, but for some reason, the output
just didn't look right. The motion feels a bit laggy, and the realism is worse than the previous version. It actually
feels like a step back. Because of how inconsistent and laggy the result looks, this one, unfortunately, has to go into F tier. Finally, I tested one 2.5, which
F tier. Finally, I tested one 2.5, which is the newest model in the lineup.
[Music] This one immediately looks far better.
The colors are cleaner, the motion is smoother, and it even has integrated audio built in, which really adds a lot of immersion and realism to the whole scene. The sound of the water and the
scene. The sound of the water and the wind in the background gives it an entirely different level of depth. And
just to give you a perspective on this, as you can now see on the screen, it easily beats all the other one models.
I'm putting one 2.5 in and a tier mostly because of the added audio and the solid overall quality of the generation. Next,
we're going to take a look at the Google VO lineup. These models are known for
VO lineup. These models are known for their cinematic tone, strong motion control, and consistently realistic lighting. We'll start with one of the
lighting. We'll start with one of the older versions, V2, and then move on to V3 and the latest release VO3.1 to see how much has changed. Starting off with V2, the result actually looks good,
especially considering how old this model is. The movement's actually really
model is. The movement's actually really solid, and it already looks better than a few of the weaker ones we tried, like one 2.2. But the textures on everything
one 2.2. But the textures on everything make it look a bit too AI, which takes away from the realism. It's a decent result overall, but compared to what's available now, it's just not on the same
level. I'm putting V2 in Ctier. Moving
level. I'm putting V2 in Ctier. Moving
on to V3, I kept all the settings at the highest possible level and ran the same prompt.
[Music] The generation came out really well. The
video quality is sharp, the motion is smooth, and the lighting feels cinematic. It's a clear improvement from
cinematic. It's a clear improvement from the older version. Everything looks
great, the motion, the effects, the colors, but we've still got one more to try out. Finally, I ran a generation for
try out. Finally, I ran a generation for VO 3.1, which is Google's newest model.
This one really stands out. The sound
design in 3.1 sells the immersion a lot better than 3.0. The audio of the sea and the seagulls sounds much more natural, and the added environmental detail makes a huge difference. The
realism, motion, and lighting are all a noticeable step up. So, for that reason, I'm placing VO 3.1 in S tier and V3 in A tier. Now, let's test out all three of
tier. Now, let's test out all three of the Hyo models. We've got Minax Hyo 2, the brand new Miniax Hyo 2.3, and Hyo Standard. They're all made for realistic
Standard. They're all made for realistic motion and smooth textures, and all of them are in Open Art. So, starting off with Minax Hilo. I set it to 6 seconds and 1080p. This one's not as pricey as
and 1080p. This one's not as pricey as some of the others we've tested, which already gives it a bit of an advantage.
I scheduled Hyo Standard, too, just to compare them side by side. The Hyo
Standard model definitely came out looking the worst. The textures feel flat and the lighting doesn't really hold up compared to any of the newer models we've seen. Because of that, I'm immediately placing Hyo Standard in F
tier. On the other hand, Minimax Hyo 2
tier. On the other hand, Minimax Hyo 2 looks much better. It feels realistic with solid texture detail on the character and the water, but there are still a few things that don't look right. The sky in the background looks
right. The sky in the background looks off, and the motion in the water doesn't quite match the rest of the scene. So,
while it's definitely usable, it's not quite at the top level. I'm putting
Minamax Hyo in Btier. Now, Hyo 2.3 is one of the newest generators we're testing today. In the upgrade, they made
testing today. In the upgrade, they made major improvements in movement, realism, and physics. So, this will be an
and physics. So, this will be an exciting one to test. I'm pasting in the same prompt, setting it to 6 seconds, turning the resolution up to 1080p, and clicking create. And this one really
clicking create. And this one really stands out. It looks the best out of all
stands out. It looks the best out of all the other Halo models. It's not perfect, and it's still a step below the top generators like Sora 2 or V3.1. But for
what it costs and how smooth it looks, I'm ranking Hyo 2.3 in a tier. Next, I
tested out Pixverse 5. This one's a standalone model that isn't part of a bigger group of models, but it's been getting a lot of attention for its balance between price and quality. I set
the generation to 5 seconds and 1080p, then clicked create. This one really surprised me with how good it looks.
It's not a generator I use often, but I've got to say I was really impressed by how realistic everything turned out.
The movement is smooth, the lighting feels natural, and the overall textures and color tones look great. But since it is just not as good as the S tier models, I'm placing Pixver 5 in a tier.
All right, now let's test out the final two contenders. Video Q1 and Honeyan.
two contenders. Video Q1 and Honeyan.
Video Q1 is actually a model aimed at very quick and stylized outputs, meaning you get pretty good results really fast.
Let's take a look at it first. There
aren't a lot of settings you can play around with here. So, I'm just going to click generate and let's see what we get. After generating for only about 30
get. After generating for only about 30 seconds, this is the result we get back.
And it's actually pretty surprising, especially considering it almost didn't spend any time generating. It's not
perfect. There's not a lot happening in the video, and it looks more like a static image with a bit of movement added, but all the textures look good, and the person in the video itself looks surprisingly solid. So, overall, this is
surprisingly solid. So, overall, this is a really impressive result. And for that reason, I'm placing VidQ1 in Btier. Now,
Huan, this is a hybrid model that supports both real and virtual styles as well as animated ones. So, it'll be pretty interesting to see how it handles our prompt. I'm running everything with
our prompt. I'm running everything with the same settings and clicking create.
Before we look at the result, I just want to mention that this model was one of the slowest ones by far. Cling 2.5
and V3.1 finished about 2 minutes faster than this one. So, that's definitely something it's going to lose points for.
But with that being said, let's take a look at the actual video. Honestly, it
didn't load great, especially with how long it took. The result overall looks decent, though. In terms of quality
decent, though. In terms of quality alone, I'd probably place it in Btier.
But just because of how long it took to generate, I'm going to have to push it down to Ctier. Still, since the quality was okay, I don't want to drop it all the way down to F tier. And now, let's look at the rankings. The best three
overall are Sora 2, Cling 2.5, and Google Vo 3.1. Easily the top for realism and quality. Some models did come pretty close like Pixverse, Hyo 2.3, and V3. They all performed really
well, but the ones that struggled the most were Cling 1.6, Wan 2.2, and Hyo Standard. Those are the ones I'd really
Standard. Those are the ones I'd really recommend avoiding altogether. All the
models that landed somewhere in the middle are honestly up to you. If you
think some of these looked better than others, go ahead and test them yourself.
But the best part is with Open Art, you can test all these models in one place.
No need to buy 10 different subscriptions. It's super convenient and
subscriptions. It's super convenient and you can easily switch between each generator, compare results, and find the one that fits your style best. If you
want to test all the models on your own, just sign up for Open Art using the link down in the description. Thank you for watching, and I'll see you guys in the next one.
Loading video analysis...