[JoseMar]

LTX2: The Free Open-Source Video Model
Generating 1080p Videos With Audio

Ok HackForums people, the wait is over.🙂

As I said yesterday, I would bring you an open-source AI video generator, and here it is.

Now you have two options:
You can use it on your own PC for FREE, or you can pay to use it online. I personally preferred running it on my own PC.

To download it, I used ComfyUI. It’s very simple to install, and inside ComfyUI you just select LTX2 and let it download.

This is LTX2, an open-source video model that you can use for free. Since its release, it's been the hottest topic. And that's because LTX2 can generate high resolution videos up to 20 seconds long. It generates synchronized audio, and it does it much faster than any other video model.

What makes LTX2 so hyped right now🙂

When it comes to video dimensions, LTX2 is probably one of the few models, if not the only one, where I'm comfortable setting the resolution to 720p and also 1080p, but I preffer 720p. This works well with my PC specs, but I wasn't able to generate in 4K using my computer. And of course, if you want faster generation speed, I recommend going with 720p or even lower so you don't exhaust your PC.

You can choose from two different models. Right now I have the FP8 version. The other version is a much bigger model and will supposedly give you better quality, but I tried that and got an out of memory error every single time. So unless you're running the workflow online or have a high-end GPU with more than 48 GB of VRAM, I wouldn't bother with this.

If you're testing LTX2 for the first time, you don't need to worry so much about the settings.

The most impressive part about LTX2 is the generation speed. I generate one video in around 3 minutes. However, I noticed that the generation time fluctuates a lot. Sometimes it's slower and other times it's much faster.

LTX2 is really nice one if you use on your PC🙂

Let me explain. LTX2 is really fast. You can generate 1080p comfortably (with a good GPU). You can push longer video durations. And the fact that it can generate speech and talking head videos out of the box is impressive. And the lips movement is almost perfectly synced to the audio, still gives you cinematic video result.

LTX2 link and API price if you want to use it online🙂

• https://ltx.io/model/api/pricing
  

[Rsynced]

Thanks for the share. I am in the process of building my own AI Rig just for this purpose. Lucky to have acquired a NVIDIA GeForce RTX 3060 Dual Overclock 12GB GDDR6 PCI Express 4.0 Graphics Card

The 12 GB Cache will work great for this model!

[JoseMar]

Thanks for the share. I am in the process of building my own AI Rig just for this purpose. Lucky to have acquired a NVIDIA GeForce RTX 3060 Dual Overclock 12GB GDDR6 PCI Express 4.0 Graphics Card

The 12 GB Cache will work great for this model!

I think it will only work at 720p, but it´s OK, you will see. Let me know please 🙂

[Rsynced]

I think it will only work at 720p, but it´s OK, you will see. Let me know please 🙂

Ok, I have some other models I want o use as well.

[JoseMar]

Ok, I have some other models I want o use as well.

Try Wan 2.2 as well, I think you’ll like it even more 🙂

[TrackPro+]

Thanks for sharing. I will give this a look.

[JoseMar]

Thanks for sharing. I will give this a look.

Let me know please, also don´t forget that you need a good GPU🙂

[b001ean]

Try Wan 2.2 as well, I think you’ll like it even more 🙂

Very true, Wan2.2 is much more mature in terms of customization.