New Promising AI Video Generator! VEO 2 Alternative?

Hey everyone, welcome back to the channel! In today's video, I dive into LumaLab's new AI video generation model, Ray 2. There's a lot of hype around this model, and I put it to the test to see if it can serve as a viable substitute for Google's VEO 2, which isn't publicly accessible yet. We explore Ray 2's claims of realistic visuals and strong understanding of text instructions, supported by various example prompts like gorillas surfing and cinematic movie footage from Egypt's heyday. We also discuss the model's pros and cons, including its strong cinematic capabilities and issues with generation speed.

▼ Link(s) From Today’s Video:

-------------------------------------------------

▼ Extra Links of Interest:

Let's work together!

Thanks for watching Matt Video Productions! I make all sorts of videos here on Youtube! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!

All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.

00:00 Introduction and Welcome
00:12 Introducing LumaLab's Ray 2
01:28 Ray 2's Capabilities and Examples
08:35 Hands-On with Ray 2
17:36 Community Feedback and Issues
23:05 Final Thoughts and Conclusion

[TheinterfaceTvSeries]

Actually I’ve found the most difficult things for AI to generate are simple things like an insert shot of a foot pressing a gas pedal, someone opening a door and walking into a room, getting in or out of a car. Things that I call the connective tissue of visual storytelling. Generative video is great at creating remarkable visuals but not so good at simple nuanced visuals.

[optimalopus]

I've been following your channel for a while. It's amazing to see how all of these tools are evolving.

[badmovi]

I tried ray 2 with text to video (because image to video is unavailable) and the speed to render was slow and the render was absolutely bad. Stick to kling 1.6

[angryox]

I’m just ready for local run open source ai video.

[Davidturner-qy]

why can't prompt to 3d models solve the problem of consistent characters in prompt to video ?

[chilloutloops]

That fencing example video could literally be mistaken for the real thing. I didn't even know AI video models could do anything with 'realistic' combat yet, at least I haven't had much luck. So that blew me away. And fencing foils are kinda wobbly, so that might have been intentional and not an AI artifact of the blade physics.

[JonnyCrackers]

I tried to use it today but the two prompts I submitted are still generating 8 hours later.

[pdjay]

모두들 발전하네요. veo2를 아직 못써본 저로서는 아직까지는 Kling이 My One Top :)

[Jackofficial]

Mat video generator is but what do you think about Ai song dubbing

[AvoidTheseMemes]

I would need unlimited access to Luma and Sora. I already have unlimited full early access to VEO 2 which absolutely rocks!

[MabelYolanda-ci]

I just saw a new YouTuber using your footage and even copying word by word some of your videos, replaced with an AI voice. I tried to add the link here, but YouTube doesn’t let me, so be on the look. Perhaps it’s time to start using watermarks or something.

[GabrieloTomorrow]

If there’s something that I learned from Sora, is that a healthy dog is much better than a dying lion.
And that applies to Veo2…

[MojaveHigh]

45 of your viewers have actually tried Ray 2, Veo 2 AND Vidu 2? Seriously?

[RMCanimationOFFICAL]

Where’s the intro, did you forget the intro?!?

[PewnaOsoba-cx]

Why these fonts on these modern sites are so big? I can't read nothing from this.

[YourBrandDead]

I bet half the hair commenters are bald 🤯😱

[TheGalacticIndian]

First things first: what happened to your HAIR??... just kidding, Matt😉

[Zerobytexai]

Matt, you're a cool guy, but you over exaggerate everything. It's kind of annoying and it's not realistic.

No this is not even close to Veo 2. Please stop blowing smoke up everyone's butt all the time and exaggerating and fabricating everything. It's so obvious that it's cringy at times.

You've been doing this a lot in your videos, and it makes your channel more amateur.

[jonnyjonbo]

What are these video apps where you can make it look like someone is kissing a random person ? you've got to have seen the videos😅

[christiandarkin]

you can't judge a model you can't test. In production, text to video is meaningless.

image to video and responsiveness to prompt are what really matters - even if the output often fails, you can do multiple generations and pick the best. But if it doesn't follow the input image or fails to listen to the prompt, then it can't be used.