Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Check out this amazing combo using Gemma4 + Falcon Perception for video tracking! 1️⃣Give Gemma 4 video frames 2️⃣It describes what it sees 3️⃣Falcon Perception takes those descriptions, segments the objects, and tracks them across the video! The best part? All running locally!

Google Gemma

89,259 subscribers

83,744 views • 3 months ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Next mlx-vlm release will ship with continuous batching support on the server 🚀 What's coming: → Continuous batching — new requests join the active batch immediately, no waiting. Mixed image + text batches supported → OpenAI-compatible API — field-for-field match with mlx-lm, reasoning/content split for thinking models, tag-aware streaming → Multi-turn tool calling — full tool use support across streaming and non-streaming, works with Gemma4 and other templates → Vision feature caching — cache image embeddings across turns. Gemma4: 228x speedup, Qwen3.5: 23x on cache hit All running locally on Apple Silicon. Check our this demo running 4 concurrent requests (mixed image + text) to gemma-4-26B-A4B-IT by Google Gemma in bf16 using Pi + MLX-VLM server on my M3 Ultra. One of the requests ingests a 8K resolution image!

Next mlx-vlm release will ship with continuous batching support on the server 🚀 What's coming: → Continuous batching — new requests join the active batch immediately, no waiting. Mixed image + text batches supported → OpenAI-compatible API — field-for-field match with mlx-lm, reasoning/content split for thinking models, tag-aware streaming → Multi-turn tool calling — full tool use support across streaming and non-streaming, works with Gemma4 and other templates → Vision feature caching — cache image embeddings across turns. Gemma4: 228x speedup, Qwen3.5: 23x on cache hit All running locally on Apple Silicon. Check our this demo running 4 concurrent requests (mixed image + text) to gemma-4-26B-A4B-IT by Google Gemma in bf16 using Pi + MLX-VLM server on my M3 Ultra. One of the requests ingests a 8K resolution image!

Prince Canuma

82,169 views • 3 months ago

Check out this Circulus Pion robot at Computex 2026. It’s running on a Panther Lake processor, using P-cores for real-time control, E-cores to collect sensor data, the GPU for vision-language-action models, and the NPU for continuous perception.

Check out this Circulus Pion robot at Computex 2026. It’s running on a Panther Lake processor, using P-cores for real-time control, E-cores to collect sensor data, the GPU for vision-language-action models, and the NPU for continuous perception.

Intel Technology

10,752 views • 1 month ago

Best UFO ever recorded?, someone on Reddit posted this video, and it's amazing. Check out the 59th second.

Best UFO ever recorded?, someone on Reddit posted this video, and it's amazing. Check out the 59th second.

Neil Goodman

102,777 views • 18 days ago

Running Gemma 4 12B locally at 28T/s. Facing issues with vision in LM Studio, but text performance is quite solid. Asked it to create a landing page for a tech event. Video is fast forwarded for the demo.

Running Gemma 4 12B locally at 28T/s. Facing issues with vision in LM Studio, but text performance is quite solid. Asked it to create a landing page for a tech event. Video is fast forwarded for the demo.

AshutoshShrivastava

15,686 views • 1 month ago

All 4 tracks have performance video!!!! 🐶: First of all, I thought about how to repay the fans for their support, and it felt like it's 'stage/perform" , so we filmed 4 performance videos this time. 🐶: So, we filmed performance videos for all the tracks, including choreography. 🐶: And by doing so, we thought that giving the fans more to see would be the best way to repay them.

All 4 tracks have performance video!!!! 🐶: First of all, I thought about how to repay the fans for their support, and it felt like it's 'stage/perform" , so we filmed 4 performance videos this time. 🐶: So, we filmed performance videos for all the tracks, including choreography. 🐶: And by doing so, we thought that giving the fans more to see would be the best way to repay them.

Moori ʚ🐥ɞ

14,987 views • 1 year ago

Introducing DeepSeek R1:1.5b Running Locally in Cursor! In less than 4 minutes, I set up the DeepSeek R1:1.5b model, download it, and run it locally to seamlessly work with code in Cursor The video is shown in real-time, with the model impressively running on an Intel Core i5 😱 using Ollama and the CodeGPT extension 👇

Introducing DeepSeek R1:1.5b Running Locally in Cursor! In less than 4 minutes, I set up the DeepSeek R1:1.5b model, download it, and run it locally to seamlessly work with code in Cursor The video is shown in real-time, with the model impressively running on an Intel Core i5 😱 using Ollama and the CodeGPT extension 👇

Daniel San

584,081 views • 1 year ago

XRP And Wal-Mart! What - Check This Out! Thanks X Finance Bull...for the inspiration for this video!

XRP And Wal-Mart! What - Check This Out! Thanks X Finance Bull...for the inspiration for this video!

Austin Hilton

12,076 views • 1 year ago

We're tracking active weather this week. Check out what to expect across the state in the video explainer below ⬇️⬇️⬇️ - High Surf 🌊 - Strong winds (some wind gusts up to 40 MPH) 🌬️

We're tracking active weather this week. Check out what to expect across the state in the video explainer below ⬇️⬇️⬇️ - High Surf 🌊 - Strong winds (some wind gusts up to 40 MPH) 🌬️

Drew Davis🤙

13,924 views • 1 year ago

If you were curious what the Crimson Wolf weapon skin looks like on every Wuyang skin in the game, you can check out this video I made that showcases them. Joker is absolutely the best skin combo, OWCS Away is my 2nd favorite with Cozy being the 3rd! Full Video Below 👇👇👇

If you were curious what the Crimson Wolf weapon skin looks like on every Wuyang skin in the game, you can check out this video I made that showcases them. Joker is absolutely the best skin combo, OWCS Away is my 2nd favorite with Cozy being the 3rd! Full Video Below 👇👇👇

ExO Chantal | ObsSojourn

18,577 views • 4 months ago

Your phone can build voxel world now! 👾 This is an open-source tool that converts phone video into 3D point clouds. You can run Depth Anything 3 locally. Basically just local GPU inference. The interesting part is how it handles messy input. Video windows overlap, scale gets converted to real meters, drift gets corrected after inference. The system stitches everything together even when footage is noisy. Objects far away might fail strict confidence filters but still show up as recognizable buildings. That's huge for small robots because perception becomes cheap and stays local. Obviously it wouldn't be a phone for robots, but it shows the potential of edge/local perception processing. A robot can understand its environment without offloading to the cloud. Real-world video has artifacts and occlusions. The system doesn't need perfect input to reconstruct scene structure. It works with what's there. For robotics, this means small platforms can build 3D maps autonomously using hardware they already have. Here's the repo: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Your phone can build voxel world now! 👾 This is an open-source tool that converts phone video into 3D point clouds. You can run Depth Anything 3 locally. Basically just local GPU inference. The interesting part is how it handles messy input. Video windows overlap, scale gets converted to real meters, drift gets corrected after inference. The system stitches everything together even when footage is noisy. Objects far away might fail strict confidence filters but still show up as recognizable buildings. That's huge for small robots because perception becomes cheap and stays local. Obviously it wouldn't be a phone for robots, but it shows the potential of edge/local perception processing. A robot can understand its environment without offloading to the cloud. Real-world video has artifacts and occlusions. The system doesn't need perfect input to reconstruct scene structure. It works with what's there. For robotics, this means small platforms can build 3D maps autonomously using hardware they already have. Here's the repo: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

50,190 views • 15 days ago

Google just launched Universal Commerce Protocol. I created this video using Google NotebookLM to explain what UCP is, how it works, I like the video, it came out really well and explains everything clearly. Hope you all enjoy it. Note: I slightly edited the video for my YouTube channel.

Google just launched Universal Commerce Protocol. I created this video using Google NotebookLM to explain what UCP is, how it works, I like the video, it came out really well and explains everything clearly. Hope you all enjoy it. Note: I slightly edited the video for my YouTube channel.

AshutoshShrivastava

36,315 views • 6 months ago

Check out this video for a quick walkthrough on how to dive in and get started in the new game mode "Descent". Do you think you've got what it takes to defeat the Nemesis? #TheDivision2

Check out this video for a quick walkthrough on how to dive in and get started in the new game mode "Descent". Do you think you've got what it takes to defeat the Nemesis? #TheDivision2

Tom Clancy's The Division

132,836 views • 3 years ago

I'm obsessed with pushing local small models to their limits. Qwen3.5:0.8b doing real-time video captioning on a Mac Studio M2 Ultra, streaming descriptions as the video plays. Under 1s per frame — 269 frames captured & described from a 3m49s video. Pause anywhere and read the captions, it describes every frame surprisingly well. This model is barely 1GB. Local AI is moving absurdly fast.

I'm obsessed with pushing local small models to their limits. Qwen3.5:0.8b doing real-time video captioning on a Mac Studio M2 Ultra, streaming descriptions as the video plays. Under 1s per frame — 269 frames captured & described from a 3m49s video. Pause anywhere and read the captions, it describes every frame surprisingly well. This model is barely 1GB. Local AI is moving absurdly fast.

stevibe

106,441 views • 4 months ago

Today we’re introducing Gen-4 Turbo. The fastest way to generate with our most powerful video model yet. With Gen-4 Turbo it now takes just 30 seconds to generate a 10 second video, making it ideal for rapid iteration and creative exploration. Now rolling out across all plans.

Today we’re introducing Gen-4 Turbo. The fastest way to generate with our most powerful video model yet. With Gen-4 Turbo it now takes just 30 seconds to generate a 10 second video, making it ideal for rapid iteration and creative exploration. Now rolling out across all plans.

Runway

93,662 views • 1 year ago

🚨 The Gridlock Practice Weekend begins TOMORROW! For all the details, and to know what you can expect going into this weekend, check out the video below ⬇️

🚨 The Gridlock Practice Weekend begins TOMORROW! For all the details, and to know what you can expect going into this weekend, check out the video below ⬇️

Gridlock

26,805 views • 1 month ago

How has the Singapore F1 track layout changed for this year, and what does it look like from the cockpit? Check out this video from The Athletic and EA SPORTS F1.

How has the Singapore F1 track layout changed for this year, and what does it look like from the cockpit? Check out this video from The Athletic and EA SPORTS F1.

Madeline Coleman

27,366 views • 2 years ago

My camera was shooting still images at 30 frames per second, so I put them all together into a video of the snowy owl's flight. I used Final Cut Pro to slow the video to 25% using machine learning analysis to smooth the movement. It actually came out pretty well.

My camera was shooting still images at 30 frames per second, so I put them all together into a video of the snowy owl's flight. I used Final Cut Pro to slow the video to 25% using machine learning analysis to smooth the movement. It actually came out pretty well.

Rob Wright Images

20,058 views • 4 months ago

i made a computer vision system to monitor the situation - live webcam feed and roboflow RF-DETR for object detection - inputs are fed into SmolVLM which describes what it sees in one sentence - dataviz with vanilla js all running locally in real-time on my macbook air m2

i made a computer vision system to monitor the situation - live webcam feed and roboflow RF-DETR for object detection - inputs are fed into SmolVLM which describes what it sees in one sentence - dataviz with vanilla js all running locally in real-time on my macbook air m2

AA

72,465 views • 5 months ago

When the NotebookLM team was building video overviews, they wanted to combine the best of Gemini's multimodality into one feature. The AI host "sees" your sources, processes the information, and then is able to discuss what is truly unique about them. For example, we uploaded one of Google DeepMind's blog posts and created a video overview to help distill the complex scientific information into a visually engaging summary, check it out!

When the NotebookLM team was building video overviews, they wanted to combine the best of Gemini's multimodality into one feature. The AI host "sees" your sources, processes the information, and then is able to discuss what is truly unique about them. For example, we uploaded one of Google DeepMind's blog posts and created a video overview to help distill the complex scientific information into a visually engaging summary, check it out!

Google AI

142,258 views • 11 months ago

🌀 Tuesday video discussion on Tropical Storm #Erin, tracking westward across the eastern Atlantic. 🌀 Erin's forecast track takes it north of but fairly close to the Leeward Islands of the northeastern Caribbean. This region should monitor the forecast and be prepared for potential impacts.

🌀 Tuesday video discussion on Tropical Storm #Erin, tracking westward across the eastern Atlantic. 🌀 Erin's forecast track takes it north of but fairly close to the Leeward Islands of the northeastern Caribbean. This region should monitor the forecast and be prepared for potential impacts.

Dr. Levi Cowan

41,987 views • 11 months ago