Up next in 10
Create stunning videos and images with AI - https://lanhive.com
Show More Show Less View Video Transcript
0:00
We have a new AI model from LTX. It can
0:03
generate up to 4K videos. It's amazing.
0:07
Try it online.
0:08
Light tricks introduces LTX2, a video
0:11
demonation model with amazing features.
0:14
It comes with synchronized audio, up to
0:16
4K video resolution, up to 50 frames per
0:20
second, up to 10 seconds video duration,
0:22
and it's open source. I tried it and I
0:25
was amazed with the results. Here are
0:27
some of the examples I tried. By the
0:29
way, you can use LTX2 and many other
0:31
video generation models on Lanhive.
0:34
Let's look at the examples. So on LHive,
0:36
you can just go to create video and here
0:38
on the models, you can just click the
0:40
drop-down, scroll down, and choose
0:43
either LTX2 Pro or LTX2 fast. So for the
0:46
first one, I started with text to video.
0:48
And here is the prompt that I used. I'll
0:50
leave all the prompts that I've used in
0:52
the description below. This was the
0:54
result.
0:58
Hi, I see you here every morning. You
1:00
stay around here?
1:02
No, just love it here.
1:05
This is very good. It followed the
1:07
prompt well. The only thing I noticed
1:09
that this bench was much on the way
1:11
rather than on the site here, but
1:13
everything is perfect. For the second
1:15
one, I used image to video. Image to
1:17
video gives you more control than text
1:19
to video. This is the image that I used.
1:22
And this was the prompt. And this was
1:23
the result. They told me that Earth is
1:26
primitive, that I would not be detected.
1:28
Here I am locked in. But I like it here.
1:32
Okay, I like this one. Especially the
1:34
last part where he says but then takes a
1:37
pause and finish the sentence. It's
1:39
really good. According to the prompt, he
1:41
was supposed to be moving around this
1:42
cell. But despite that, I love how this
1:45
came out. For the next one, I tried this
1:47
one, the image of the giant there and
1:49
the meta human. Then this is the prompt
1:51
that I had. Then I got this result. You
1:53
are ready. Your time has come. Go and
1:57
conquer.
2:01
Okay, this one started so well. I love
2:03
it. The lifttoff was good, but the
2:05
ending not as very good. Also, if you
2:08
noticed, the words that the giant was
2:10
saying were the same words that the meta
2:12
human said as well. The metauman was
2:14
supposed to be quiet, but I would not
2:16
complain about that. This is good. Using
2:18
the same image, I tweaked the prompt a
2:20
little bit so that the meta human can
2:22
list his trident and someone lightening
2:25
like that. Let's see how this came out.
2:27
You are ready. Your time has come. Go
2:30
and conquer.
2:35
Okay, this was beautiful. I like the
2:37
lightning. I like the dust on the
2:39
ground. But again, the meta human was
2:40
saying the same words that the giant was
2:43
saying. And you saw as the creep ended
2:45
the giant draw a little bit backward
2:47
just to give space to that lightening
2:49
and the dust which is very very good.
2:52
And so far this may be one of my
2:53
favorite. And staying on the topic of
2:56
giants I wanted to see how the model
2:58
would handle fighting. This didn't go as
3:00
I expected. And here is the result.
3:11
So you can see it's not very realistic.
3:13
It took a lot of time just to start the
3:15
action. The fall from the giant and the
3:17
physics of it is all good, but the
3:19
entire scene is not what I expected.
3:21
Then I tried this prompt. I've used it
3:23
in other models like V3 and it was good.
3:26
I wanted to see how this one would do
3:28
here. The results were good enough.
3:31
I hope this temple was not built by
3:33
aliens.
3:38
Okay, not bad. But according to the
3:39
prompt, you are supposed to be
3:40
whispering and show fear. But I can't
3:43
see fear in the way he's moving and in
3:45
the way he's speaking. And also, he's
3:47
not whispering. Despite that, the
3:49
generation is good. Everything is good.
3:51
For this one, I wanted to try a small
3:53
dialogue to see how the model would
3:55
handle it. So, we had these two ladies.
3:57
The first lady speaks, then the last one
3:59
speaks later on. Let's see what we get.
4:02
This was my best semester ever. I loved
4:04
every moment of it. I am so happy for
4:06
you.
4:09
Except for the awkward silence at the
4:11
end, the rest was okay. But you could
4:14
see that as the first lead was speaking,
4:16
part of the last part, she said it
4:18
together with this one. So, not bad, but
4:20
it could have been better. So, for this
4:22
one, I wanted to try a time lapse. The
4:24
result was not bad, but it's better if
4:26
you're using first frame and last frame.
4:28
In this case, we're only using the first
4:30
image. The model doesn't know how this
4:32
house will be at the end. It's relying
4:34
on the prompt. And you can see here the
4:36
prompt that I've used does not have a
4:38
lot of details. So let's see the result
4:40
we got.
4:49
Okay, as you can see, this is not the
4:50
modern house that I expected, but in all
4:53
it's not bad given that we didn't
4:54
provide so much details in the prompt.
4:57
with a little bit of extra details would
4:59
definitely get good results. Next, I
5:01
tried this shoot announcement from this
5:03
guy and I would say I loved how the
5:05
result came out. Here's what we get.
5:07
We have a new AI model from LTX. It can
5:11
generate up to 4K videos. It's amazing.
5:14
Try it on Lana.
5:15
I love how the initial zoom in happened
5:18
and after that there was subtle zooming
5:20
in as the character was speaking. I
5:22
loved this result. Although at the end
5:24
he didn't finish the web plant hive but
5:26
ultimately this is a very good result.
5:29
For the last one I tried this. This man
5:31
drinks from this can then it says some
5:33
words and continues drinking. I love
5:35
this one. It was good. I think you would
5:37
have many application from this one.
5:40
H Fruto mix fruit is so amazing and so
5:43
healthy. You should try it.
5:47
The words on the can are not readable,
5:49
but it's okay because probably this was
5:51
far away and the model cannot see what
5:53
is written there. But all in all, the
5:55
character did what we requested, said
5:58
everything that we requested and this is
6:00
very good. I also love the subtle
6:02
details in the background like how these
6:04
guys are moving the cars in the far
6:06
background. It's really good. Overall, I
6:09
think OTX2 is one of the best models out
6:12
there for video generation. I think I'll
6:14
be using this a lot. It's a cheap
6:16
alternative to V3.1 and Solar 2 and many
6:19
models out there. And the fact that it
6:21
comes with up to 4K resolution makes
6:23
this a great choice, especially for
6:25
storytelling, film making, or anything
6:27
like that. If you want to try this, I'll
6:29
leave the link in the description below.
6:31
And if you want to see how you can use
6:33
AI for motion graphics, that video is up
6:35
here. or see over there.