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.

Gemma 4 12B in action: Object detection, function calling, voice command, segmentation, language switch, translation - all of this and much more without vision/audio encoders! (Inputs and outputs are real, but FC2 data shown as code, and generation speedified)

Andreas Steiner

1,308 subscribers

10,448 views • 1 month ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

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

72,238 views • 4 months ago

New Google Gemma 4 12B claims near-26B performance - we tested both! We ran both models locally on one RTX 4090 and gave each the same task: write a self-contained HTML5 canvas animation with real physics in one file without libraries. Three scenes - a Galton board, two blocks colliding off a wall, and a chaotic triple pendulum Outputs: Gemma 4 26B-A4B: 15 GB VRAM usage, 6.9k tokens, 138 tok/s Gemma 4 12B: 9 GB VRAM usage, 8.9k tokens, 80 tok/s Same Gemma 4 family, but the 26B-A4B won every scene and ran ~1.7x faster - on just 4B active params. The 12B stayed very close though, on almost half the VRAM - which makes it the ideal model for a 16 GB laptop

New Google Gemma 4 12B claims near-26B performance - we tested both! We ran both models locally on one RTX 4090 and gave each the same task: write a self-contained HTML5 canvas animation with real physics in one file without libraries. Three scenes - a Galton board, two blocks colliding off a wall, and a chaotic triple pendulum Outputs: Gemma 4 26B-A4B: 15 GB VRAM usage, 6.9k tokens, 138 tok/s Gemma 4 12B: 9 GB VRAM usage, 8.9k tokens, 80 tok/s Same Gemma 4 family, but the 26B-A4B won every scene and ran ~1.7x faster - on just 4B active params. The 12B stayed very close though, on almost half the VRAM - which makes it the ideal model for a 16 GB laptop

atomic.chat

151,491 views • 1 month ago

Ever wondered what robots 🤖 could achieve if they could not just see – but also feel and hear? Introducing FuSe: a recipe for finetuning large vision-language-action (VLA) models with heterogeneous sensory data, such as vision, touch, sound, and more. Details in the thread 👇

Ever wondered what robots 🤖 could achieve if they could not just see – but also feel and hear? Introducing FuSe: a recipe for finetuning large vision-language-action (VLA) models with heterogeneous sensory data, such as vision, touch, sound, and more. Details in the thread 👇

Carlo Sferrazza

46,111 views • 1 year ago

🗓️ Release Week Recap Big week. mlx-audio and mlx-vlm are now among some of the fastest-growing OSS projects. Here’s what we shipped last week. Gemma 4 on Apple Silicon Two awesome releases by our partner Google DeepMind & Google Gemma : > Gemma 4 12B — their new dense, unified multimodal model. It uses an encoder free audio path and simplified vision encoder. > Gemma 4 QAT — quantization-aware training checkpoints, optimized to run locally on consumer GPUs and edge devices, compressing the model while preserving the quality you expect from Gemma 4. On the audio 🎧 side we added support for 15+ new TTS, ASR & VAD models, faster long-form transcription, and an expanded OpenAI-compatible audio server. All local on Apple Silicon. Huge thanks to every contributor and my co-maintainer Lucas Newman. 🙏🏽

🗓️ Release Week Recap Big week. mlx-audio and mlx-vlm are now among some of the fastest-growing OSS projects. Here’s what we shipped last week. Gemma 4 on Apple Silicon Two awesome releases by our partner Google DeepMind & Google Gemma : > Gemma 4 12B — their new dense, unified multimodal model. It uses an encoder free audio path and simplified vision encoder. > Gemma 4 QAT — quantization-aware training checkpoints, optimized to run locally on consumer GPUs and edge devices, compressing the model while preserving the quality you expect from Gemma 4. On the audio 🎧 side we added support for 15+ new TTS, ASR & VAD models, faster long-form transcription, and an expanded OpenAI-compatible audio server. All local on Apple Silicon. Huge thanks to every contributor and my co-maintainer Lucas Newman. 🙏🏽

Prince Canuma

11,731 views • 1 month ago

The new episode of #TheAgentFactory is live! We’re diving into Gemma 4, the powerhouse open model family for single GPUs. See it in action: ✅ Apache 2.0 license ✅ Native function calling ✅ 100% offline agents And, watch it build in real time →

The new episode of #TheAgentFactory is live! We’re diving into Gemma 4, the powerhouse open model family for single GPUs. See it in action: ✅ Apache 2.0 license ✅ Native function calling ✅ 100% offline agents And, watch it build in real time →

Google Cloud Tech

15,420 views • 3 months ago

Tracking Anything with Decoupled Video Segmentation paper page: Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To 'track anything' without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation.

Tracking Anything with Decoupled Video Segmentation paper page: Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To 'track anything' without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation.

AK

305,633 views • 2 years ago

Microsoft has launched a powerful new data analysis tool! Introducing Data Formulator, a 100% open-source LLM-powered, no-code tool that transforms data in a snap and creates stunning visualizations. Key features include: 🤖 AI-powered data transformation 🖱️ Interactive drag-and-drop UI for visualizations 💬 Seamless blend of UI & natural language inputs But that’s not all: You can even create charts beyond your initial dataset. Data Formulator automatically identifies extra computation needs, generates fields for you, and outputs the final visualization. Find the GitHub repo in the next tweet! _____ Find me → Akshay 🚀 ✔️ For more insights and tutorials on AI and Machine Learning.

Microsoft has launched a powerful new data analysis tool! Introducing Data Formulator, a 100% open-source LLM-powered, no-code tool that transforms data in a snap and creates stunning visualizations. Key features include: 🤖 AI-powered data transformation 🖱️ Interactive drag-and-drop UI for visualizations 💬 Seamless blend of UI & natural language inputs But that’s not all: You can even create charts beyond your initial dataset. Data Formulator automatically identifies extra computation needs, generates fields for you, and outputs the final visualization. Find the GitHub repo in the next tweet! _____ Find me → Akshay 🚀 ✔️ For more insights and tutorials on AI and Machine Learning.

Akshay 🚀

280,449 views • 1 year ago

🔴 The Pain: Running local MLX models is incredibly fast and private. But let's be real - testing tool calling via terminal is clunky, and there's zero good UI for it. 🟢 The Fix: rapid-mlx share Just ONE command gives you a polished web chat + seamless tool calling (works beautifully with gemma-4-12b-qat). We are proud to be the ONLY MLX inference engine in the community shipping this. ⚡️ 👇 Try it now: brew install raullenchai/rapid-mlx/rapid-mlx

🔴 The Pain: Running local MLX models is incredibly fast and private. But let's be real - testing tool calling via terminal is clunky, and there's zero good UI for it. 🟢 The Fix: rapid-mlx share Just ONE command gives you a polished web chat + seamless tool calling (works beautifully with gemma-4-12b-qat). We are proud to be the ONLY MLX inference engine in the community shipping this. ⚡️ 👇 Try it now: brew install raullenchai/rapid-mlx/rapid-mlx

raullen

22,680 views • 1 month ago

Infinite Photorealistic Worlds using Procedural Generation paper page: introduce Infinigen, a procedural generator of photorealistic 3D scenes of the natural world. Infinigen is entirely procedural: every asset, from shape to texture, is generated from scratch via randomized mathematical rules, using no external source and allowing infinite variation and composition. Infinigen offers broad coverage of objects and scenes in the natural world including plants, animals, terrains, and natural phenomena such as fire, cloud, rain, and snow. Infinigen can be used to generate unlimited, diverse training data for a wide range of computer vision tasks including object detection, semantic segmentation, optical flow, and 3D reconstruction. We expect Infinigen to be a useful resource for computer vision research and beyond.

Infinite Photorealistic Worlds using Procedural Generation paper page: introduce Infinigen, a procedural generator of photorealistic 3D scenes of the natural world. Infinigen is entirely procedural: every asset, from shape to texture, is generated from scratch via randomized mathematical rules, using no external source and allowing infinite variation and composition. Infinigen offers broad coverage of objects and scenes in the natural world including plants, animals, terrains, and natural phenomena such as fire, cloud, rain, and snow. Infinigen can be used to generate unlimited, diverse training data for a wide range of computer vision tasks including object detection, semantic segmentation, optical flow, and 3D reconstruction. We expect Infinigen to be a useful resource for computer vision research and beyond.

AK

275,335 views • 3 years ago

Smart Turn v2: open source, native audio turn detection in 14 languages. New checkpoint of the open source, open data, open training code, semantic VAD model on Hugging Face, fal, and Pipecat AI. - 3x faster inference (12ms on an L40) - 14 languages (13 more than v1, which was english-only) - New synthetic data set `chirp_3_all` with ~163k audio samples - 99% accuracy on held out `human_5_all` test data Good turn detection is critical for voice agents. This model "understands" both semantic and audio patterns, and mitigates the voice AI trade-off between unwanted turn latency vs the agent interrupting people before they are finished speaking. Training scripts for both Modal and local training are in the repo. We want to make it as easy as possible to contribute to or customize this model! Here's a demo running the smart-turn model with default settings, aimed at generally hitting 400ms total turn detection time. You can tune things to be faster, too. You can help by contributing data, doing architecture expermints, or cleaning open source data! Keep reading ...

Smart Turn v2: open source, native audio turn detection in 14 languages. New checkpoint of the open source, open data, open training code, semantic VAD model on Hugging Face, fal, and Pipecat AI. - 3x faster inference (12ms on an L40) - 14 languages (13 more than v1, which was english-only) - New synthetic data set `chirp_3_all` with ~163k audio samples - 99% accuracy on held out `human_5_all` test data Good turn detection is critical for voice agents. This model "understands" both semantic and audio patterns, and mitigates the voice AI trade-off between unwanted turn latency vs the agent interrupting people before they are finished speaking. Training scripts for both Modal and local training are in the repo. We want to make it as easy as possible to contribute to or customize this model! Here's a demo running the smart-turn model with default settings, aimed at generally hitting 400ms total turn detection time. You can tune things to be faster, too. You can help by contributing data, doing architecture expermints, or cleaning open source data! Keep reading ...

kwindla

42,219 views • 1 year ago

OpenAI just dropped a completely new kind of model gpt-realtime-translate takes in speech audio from any language and outputs speech in your target language LLMs are great, but you need specialized models for specialized use cases We're running this on our smart glasses

OpenAI just dropped a completely new kind of model gpt-realtime-translate takes in speech audio from any language and outputs speech in your target language LLMs are great, but you need specialized models for specialized use cases We're running this on our smart glasses

cayden 凯登

100,395 views • 1 month ago

Excited to introduce GPT-4o. Language, vision, and sound -- all together and all in real time. This thing has been so much fun to work on. It's been even more fun to play with -- with moments of magic where things feel totally fluid and I forget I'm video chatting with an AI.

Excited to introduce GPT-4o. Language, vision, and sound -- all together and all in real time. This thing has been so much fun to work on. It's been even more fun to play with -- with moments of magic where things feel totally fluid and I forget I'm video chatting with an AI.

Rowan Zellers

128,683 views • 2 years ago

Real-time transcription just got a significant upgrade. Universal-3-Pro is now available for streaming — bringing AssemblyAI's most accurate speech model to live audio for the first time. Developers building voice agents, live captioning tools, and real-time analytics pipelines now get three things they've been asking for: 🔹 Best-in-class word error and entity detection across streaming ASR benchmarks 🔹 Real-time speaker labels — know who said what, as it happens 🔹 Superior entity detection for names, places, orgs, and specialized terminology in real-time 🔹 Code-switching and global language coverage built-in

Real-time transcription just got a significant upgrade. Universal-3-Pro is now available for streaming — bringing AssemblyAI's most accurate speech model to live audio for the first time. Developers building voice agents, live captioning tools, and real-time analytics pipelines now get three things they've been asking for: 🔹 Best-in-class word error and entity detection across streaming ASR benchmarks 🔹 Real-time speaker labels — know who said what, as it happens 🔹 Superior entity detection for names, places, orgs, and specialized terminology in real-time 🔹 Code-switching and global language coverage built-in

AssemblyAI

15,016 views • 4 months ago

i spent the afternoon experimenting with Gemma 4's vision capabilities made an app that uses roboflow RF-DETR for a first pass of object detections and Gemma to summarize the scene in one sentence for fun i asked Gemma to "describe what you see as if you were a medieval bard" all made with free local AI models (running via webgpu in the browser thanks to transformers js) lots more possibilities to explore with this

i spent the afternoon experimenting with Gemma 4's vision capabilities made an app that uses roboflow RF-DETR for a first pass of object detections and Gemma to summarize the scene in one sentence for fun i asked Gemma to "describe what you see as if you were a medieval bard" all made with free local AI models (running via webgpu in the browser thanks to transformers js) lots more possibilities to explore with this

AA

700,459 views • 3 months ago

Voice‑driven, real‑time arbitrary action generation😁 Using external voice commands, G1 is directly controlled to generate a wide range of actions in real time. This video was recorded in a single take, with on‑site audio recording. Because the actions are autonomously generated by AI in real time, there may be slight latency, and the smoothness of the movements may be somewhat reduced.

Voice‑driven, real‑time arbitrary action generation😁 Using external voice commands, G1 is directly controlled to generate a wide range of actions in real time. This video was recorded in a single take, with on‑site audio recording. Because the actions are autonomously generated by AI in real time, there may be slight latency, and the smoothness of the movements may be somewhat reduced.

Unitree

29,017,924 views • 1 month ago

🧾 [Initiation File #ΞL-010] Project: ETERNA – DAO Initialization Status: LIVE Protocol: DAO launch sequence: active Objective: 101 SOL resource intake to seed the network. Primary function: - Coordinate experimental trials - Fund research and compound testing - Log and govern all outputs on-chain Data stream: online Capital flow: active Access point: Please note: community inputs are still under review and new subjects are being added to the whitelist as the protocol evolves. Subjects: you are now part of the experiment.

🧾 [Initiation File #ΞL-010] Project: ETERNA – DAO Initialization Status: LIVE Protocol: DAO launch sequence: active Objective: 101 SOL resource intake to seed the network. Primary function: - Coordinate experimental trials - Fund research and compound testing - Log and govern all outputs on-chain Data stream: online Capital flow: active Access point: Please note: community inputs are still under review and new subjects are being added to the whitelist as the protocol evolves. Subjects: you are now part of the experiment.

Eterna Labs

43,725 views • 11 months ago

🎵 ✨The world’s most flexible sound machine? With text and audio inputs, this new #generativeAI model, named Fugatto, can create any combination of music, voices, and sounds.🎹 Read more in our blog by richardkerris ➡️ #NVIDIAResearch Note: Some music, sounds, and the voice of NVIDIA CEO Jensen Huang used in this video are AI generated.

🎵 ✨The world’s most flexible sound machine? With text and audio inputs, this new #generativeAI model, named Fugatto, can create any combination of music, voices, and sounds.🎹 Read more in our blog by richardkerris ➡️ #NVIDIAResearch Note: Some music, sounds, and the voice of NVIDIA CEO Jensen Huang used in this video are AI generated.

NVIDIA AI Developer

90,328 views • 1 year ago

Google presents AudioPaLM: A Large Language Model That Can Speak and Listen paper page: introduce AudioPaLM, a large language model for speech understanding and generation. AudioPaLM fuses text-based and speech-based language models, PaLM-2 [Anil et al., 2023] and AudioLM [Borsos et al., 2022], into a unified multimodal architecture that can process and generate text and speech with applications including speech recognition and speech-to-speech translation. AudioPaLM inherits the capability to preserve paralinguistic information such as speaker identity and intonation from AudioLM and the linguistic knowledge present only in text large language models such as PaLM-2. We demonstrate that initializing AudioPaLM with the weights of a text-only large language model improves speech processing, successfully leveraging the larger quantity of text training data used in pretraining to assist with the speech tasks. The resulting model significantly outperforms existing systems for speech translation tasks and has the ability to perform zero-shot speech-to-text translation for many languages for which input/target language combinations were not seen in training. AudioPaLM also demonstrates features of audio language models, such as transferring a voice across languages based on a short spoken prompt.

Google presents AudioPaLM: A Large Language Model That Can Speak and Listen paper page: introduce AudioPaLM, a large language model for speech understanding and generation. AudioPaLM fuses text-based and speech-based language models, PaLM-2 [Anil et al., 2023] and AudioLM [Borsos et al., 2022], into a unified multimodal architecture that can process and generate text and speech with applications including speech recognition and speech-to-speech translation. AudioPaLM inherits the capability to preserve paralinguistic information such as speaker identity and intonation from AudioLM and the linguistic knowledge present only in text large language models such as PaLM-2. We demonstrate that initializing AudioPaLM with the weights of a text-only large language model improves speech processing, successfully leveraging the larger quantity of text training data used in pretraining to assist with the speech tasks. The resulting model significantly outperforms existing systems for speech translation tasks and has the ability to perform zero-shot speech-to-text translation for many languages for which input/target language combinations were not seen in training. AudioPaLM also demonstrates features of audio language models, such as transferring a voice across languages based on a short spoken prompt.

AK

290,517 views • 3 years ago

Today we’re launching our first homegrown AI model: an open source turn detection model for building voice agents. Instead of relying solely on voice activity detection (VAD), which only considers when a user is speaking, our model also considers what has and is being said in the context of a conversation and predicts when a user is finished expressing their thoughts before the agent responds. Conversations with AI voice agents using this new model flow much more naturally without constant interruptions from the AI— check it out (more videos, details, and code in the thread):

Today we’re launching our first homegrown AI model: an open source turn detection model for building voice agents. Instead of relying solely on voice activity detection (VAD), which only considers when a user is speaking, our model also considers what has and is being said in the context of a conversation and predicts when a user is finished expressing their thoughts before the agent responds. Conversations with AI voice agents using this new model flow much more naturally without constant interruptions from the AI— check it out (more videos, details, and code in the thread):

LiveKit

126,860 views • 1 year ago