Googles New Text To Video BEATS EVERYTHING (LUMIERE)

Welcome to our channel where we bring you the latest breakthroughs in AI. From deep learning to robotics, we cover it all. Our videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on our latest videos.

Was there anything we missed?

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[zuzelstein]

I'm sick of them showing charts and releasing papers where they are claiming to be the best without public demo.

[tukanhamen]

Google loves their papers and videos. Ill be impressed if they release it and its as good as they say.

[DanyAI]

To be honest, I really hope that other companies beat this thing and actually release it to the public, because the only thing that I don't like about this is that it was done by Google, who are known for their _"Show but no Release"_ slogan.

[BruceWayne]

Next week we will find out that they handmade the video from pictures. I’ll believe it when I see it released. It looks amazing, as long as there is no man behind the curtain trying to pull the wool over our eyes again.

[geirwiwansivertsen]

Probably a good presentation, but I cut it off because of the terrible intro “music.» Will look for another presentation about the same topic.

[Lugmillord]

Two years ago we didn't even have Dall-E 2. LET THAT SINK IN. AI progress is so insanely fast.

[splunge]

There could be a thousand reasons why Google hasn't chosen to release this. It might take too much compute, it might not be stable, it might have IP issues, they might be cherry picking the good output, it might be built on deprecated libraries so they can't maintain it, the senior architect might not still be at the company... etc., etc..

[notthebestpic]

🎯 Key Takeaways for quick navigation:

00:00 🎥 *Introduction to Google's text-to-video generator*
- Google Research presents a state-of-the-art text-to-video generator.
- Introduction to the impressive video demo showcasing the capabilities.
02:02 📊 *Lum's Performance in User Studies and Benchmarks*
- Lum outperforms other models in user preference for both text-to-video and image-to-video generation.
- Benchmark results indicate Lum's superiority over competitors like imin, PE collabs, zeroscope, and Gen 2.
- Highlight of Lum's consistency in video quality across different models.
04:15 🧠 *Lum's Unique SpaceTime Unit Architecture*
- Lum utilizes the SpaceTime unit architecture, generating the entire temporal duration of a video in one go.
- Integration of temporal downsampling and upsampling for more effective full-frame rate video generation.
- Leveraging pre-trained texture image diffusion models for enhanced generative capabilities.
05:38 🌐 *Lum's GitHub Page and Advanced Video Examples*
- Overview of Lum's GitHub page and access to impressive video examples.
- Examination of specific examples, such as a rotating Lamborghini and pouring beer, showcasing the model's advanced capabilities.
- Discussion on the significance of rotations and realism achieved in the generated content.
09:19 🎨 *Lum's Stylized Generation and Style Drop Reference*
- Lum's effective stylized video generation demonstrated through various examples.
- Reference to Google's previous work, "Style Drop, " indicating the incorporation of style-based generation.
- Exploration of Lum's potential in creating diverse video styles and its impact on the overall effectiveness.
13:16 🖌️ *Lum's Cinemagraphs, Video Inpainting, and Image-to-Video*
- Discussion on Lum's ability to animate specific regions within an image (cinemagraphs).
- Comparison with Runway's similar feature release.
- Exploration of Lum's advanced video inpainting and image-to-video capabilities.
15:38 🌊 *Lum's Performance in Animating Specific Scenarios*
- Examination of Lum's effectiveness in animating specific scenarios, including waves, giraffes, and happy elephants.
- Analysis of the model's capability to generate realistic movements in different contexts.
- Contemplation on the potential release and application of Lum in real-world scenarios.
17:58 🚀 *Speculations on Google's Future Plans*
- Speculation on Google's potential plans for the release of Lum.
- Discussion on the difference between AI research and practical product development.
- Reflection on the exciting advancements in the AI space and anticipation for future developments.

Made with HARPA AI

[labmaier]

Google is going crazy right now! I hope they accelerate their current workflows.

[psylocyn]

The “dude, trust me” method of product launch

[EC]

WHOA ! omg this is what I was imagining text to AI to become in the future...as coherent and crazy as this..and now this turns into reality ! wow its next gen next level AI animations, super impressed ! cant wait for it to get released, I might even be willing to pay for a service of theirs in case it wont be free but only if it would be really cheap (say not more than 5$ per month or so) the future is here and its only getting crazier and more amazing by the day :) thank god I'm living to witness this AI craziness and this whole AI revolution.

[bkzzzzz]

This is paradigm shift. This the moment like ChatGPT for Videos. what will be the future of Hollywood now? what will the Entertainment and video production companies do? this good and bad at the same time.

[Josytt]

Google haven’t released anything, they’ve been too focused on ad revenue over the years

[Charvak-Atheist]

Why don't they release it for public use ?
They have not even released Gemini yet.

[angel_cheon-sa]

I believe the reason they are not releasing it yet, is because they need to build censors around it because they do not want the publicity of people doing dumb things with it. (nudity, gore, misinformation, etc)

[JasonRule-]

Pronunciation would be something like:
"Loo-mee-yair." In French the letter "i" is pronounced with a hard "e" sound. I added the "y" to my pronunciation to help the syllables to flow together. To say it quickly in French It might be more like "Loom-yair" with the second syllable dissipated in the process. If you're really interested check out the pronunciation in this video, online:

Of course it's possible that Google may choose its own pronunciation for the word.

[tariq]

Google's AI capabilities are considered state-of-the-art; however, the advertising department appears to lag behind, lacking the ability to discern deepfake advertisements even after reporting. This raises doubts about the genuineness of Google's AI advancements, as it seems that these may be more of a presentation gimmick than practical advancements in addressing current challenges.

[logancade]

Guys, this means all of everyone's physical photos can be brought to life. *Hype.*

[DarolTuttle]

Alas, another "Google AI Rocks but is not available yet" release. Meanwhile, Microsoft keeps chugging along. Mr. Google! Better late than never doesn't apply anymore

[oooooooo]

your rate of output is incredible! thanks for all the videos!