Jack Dorsey’s Block made an AI model that talks to you like an awkward new friend
Block announced Dessa, a speech-to-speech AI model trained from scratch on human conversation, showcased through chats that can sound natural but occasionally awkward.
#multimodal
41 articles tagged with “multimodal”.
Beyond tree cover: mapping the world’s forest types
Google DeepMind describes a global map that classifies each 10-meter patch of land on Earth into distinct forest types, going beyond basic tree-cover data.
Google’s more personalized Gemini AI is heading to your car
Google is extending Gemini's Personal Intelligence to cars, letting the assistant draw on Gmail, Calendar, Maps, Photos, YouTube, and Search to answer context-aware questions while driving.
Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and …
NVIDIA released Nemotron 3.5 Super VL, an open 120-billion-parameter model, alongside two new Nemotron 3.5 Bolt edge models aimed at local agentic workloads.
ChatGPT’s ‘Intelligent UI’ update fills its responses with pictures, charts, and but…
OpenAI is adding an 'Intelligent UI' feature to ChatGPT that pairs text replies with interactive visuals like charts, forms, and buttons, rolling out to all users alongside GPT-6.
Google’s new Guided Vision feature can help you read the fine print
Google is launching Guided Vision in Gemini Live on compatible Android devices, using AI to give real-time audio descriptions of what the phone camera points at.
Introducing Quine: An AI research system designed for the complexity of biology
Microsoft Research unveiled Quine, an early-stage effort to build a multimodal world model of biology that connects insights across biological scales to help scientists prioritize hypotheses before lab testing.
I created an interactive digital avatar of myself — and you can talk to it
A TechCrunch AI writer created an interactive digital avatar of themselves and trained it to discuss venture fraud, reflecting on the trade-offs of AI clones.
Accelerate multimodal RL training with SkyRL on Amazon SageMaker HyperPod
AWS details how to use SkyRL on SageMaker HyperPod to post-train a Qwen3-VL-8B vision-language model with GRPO, covering setup, job execution, and adapter hosting.
Gemini 3.8 Live with Live Avatar gives Google’s AI a face
Google's Gemini 3.8 Live update introduces Live Avatar, an animated persona that lip-syncs and shows facial expressions in real time, initially limited to Gemini Enterprise customers.
Introducing Kimi K3 on Amazon Bedrock
Moonshot AI's Kimi K3 is now available on Amazon Bedrock as an open-weight option for coding and knowledge work, featuring native vision, a 1M-token context window, and prompt caching.
Snap is launching a new Specs AI tool, and it’s coming to iOS and Mac
Snap is introducing Specs Intelligence, an AI assistant that connects digital accounts to help with tasks and travel, launching alongside its Specs AR glasses and available on iOS and Mac.
Fashion app Daydream uses Apple Intelligence to help you shop the outfits in your camera roll
Daydream's fashion app now uses Apple Intelligence in iOS 27 to convert saved outfit photos into shoppable results and let users search products through Siri.
Recreating a 70-year love story frame by frame
Google DeepMind and filmmakers used AI to recreate an unrecorded 70-year love story in the short film "Love, Rendered."
Amazon Prime Video’s new AI tech matches lips to dubbed audio
Prime Video is rolling out AI technology that matches actors' mouth movements to human-dubbed audio, debuting on the German series Maxton Hall with plans to expand.
ChatGPT Sketch turns your bad drawings into detailed AI images
OpenAI announced ChatGPT Images 2.5, including a new Sketch feature that converts user doodles drawn inside ChatGPT into AI-generated images based on text instructions.
Deploy a multimodal WhatsApp ordering assistant with Amazon Bedrock AgentCore
AWS explains how to build a WhatsApp ordering assistant on Amazon Bedrock AgentCore with Amazon Nova 2, accepting text, voice notes, and real-time calls on one number with shared memory across channels.
NeoMME: an efficient Multimodal-native and Multilingual Encoder
Hugging Face has unveiled NeoMME, a new encoder designed for efficient multimodal and multilingual processing.
Modernizing and scaling support operations with generative AI on AWS
AWS describes building a generative AI support operations platform that converts training videos into structured SOPs, applies RAG to guide ticket resolution, and predicts SLA risk to prioritize work.
Google’s AI note-taking app now allows you to interact with books
Google's Gemini Notebook can now import purchased Google Play Books via a new Expert Intelligence feature, letting users query titles and generate plans, infographics, and AI podcasts from their contents.
Google’s new AI transcription edits out your ‘ums’ and ‘ahs’
Google's new Gemini 3.5 Transcribe model adds transcription that detects specialized jargon, supports more than 85 languages, and edits out filler words like 'ums' and 'ahs'.
Implement vector-prompt document classification using Amazon Bedrock
AWS outlines a multi-agent document classification approach on Amazon Bedrock that combines text analysis with Claude Haiku 4.5 and visual similarity search using Titan Multimodal Embeddings to classify insurance documents.
Twitch streamers can now opt out from training Amazon’s AI
Twitch has added an option letting users exclude their content from future training of Amazon's generative AI models, while other AI-supported features continue to work.
Google’s Gemini app surges to 1 billion users
Google's Gemini app has reached 1 billion users, with the company reporting 63% of users engage via voice and more than 150 million images generated daily.
Google DeepMind’s new AI model can control a robot’s entire body
Google DeepMind says Gemini Robotics 2 can control a humanoid robot's entire body, expanding beyond the previous model's upper-body focus to full-body motion.
Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot …
Google DeepMind introduced Gemini Robotics ER 2, a model aimed at helping robots reason, collaborate, and solve real-world tasks through improved video understanding and tool orchestration.
Claude’s voice mode is now available for Opus and Sonnet
Anthropic is bringing Claude's voice mode to its Opus and Sonnet models, moving beyond the faster Haiku model, and extending voice into apps like Gmail, Slack, and Canva.
Here’s what Samsung’s smart glasses actually look like
Samsung has shown its upcoming AI smart glasses in person, unveiling two new designs and initial specs including 9-hour battery life, with a fall launch planned.
Meta made its own AI detection system. It should have just used Google’s
Meta unveiled Content Seal, an invisible watermarking tool that flags images from its AI model, but the tool is seen as less accessible and reliable than existing options like SynthID and C2PA.
Adobe’s ‘natural look’ camera app embraces generative AI
Adobe is adding a generative AI suite called AI Playground to its experimental Project Indigo camera app, offered as an opt-out experiment to a small percentage of users.
At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI
NVIDIA used SIGGRAPH to showcase graphics and simulation advances driven by agentic and physical AI, spanning media, content creation and robotics.
Introducing Grok on Amazon Bedrock
Amazon Bedrock now offers access to Grok 4.3, positioned for agentic and enterprise workloads with features such as configurable reasoning effort, tool calling, and multi-turn conversations.
Google Vids now lets you star in your own AI videos
Google Vids is introducing personalized AI avatars that let users appear in their own AI videos, along with Gemini Omni-powered generation and editing from prompts and reference images.
Google is renaming NotebookLM to Gemini Notebook
Google is rebranding NotebookLM as Gemini Notebook. The app remains standalone but will integrate more closely with Gemini and Google Search.
Agentic vision: Building visual intelligence with Amazon Bedrock and MCP servers
AWS describes a Computer Vision MCP Server built with Amazon Bedrock, offering a standardized interface for AI systems to process visual information and make decisions.
How Speech-to-Text Works: From Sound Waves to Transcripts
Speech recognition went from a brittle research problem to a solved-ish commodity in about three years. Here is the pipeline that made it work — and where it still breaks.
Waze is getting a bunch of new AI-powered features
Waze is gaining four new updates, two of which use Google's Gemini assistant, including conversational voice reporting and a new Destination Search feature.
How CLIP and Multimodal Search Work: One Space for Text and Images
Searching photos by typing a description, or matching an image to a caption, relies on one clever idea: putting text and images into the same mathematical space. Here is how CLIP made that work.
Multimodal AI Goes Mainstream: One Model, Many Senses
The leading AI models no longer just read and write text — they see images and hear audio too. Here is what 'multimodal' means and why it has become the default expectation.
Stable Diffusion: The Open Model That Democratized AI Image Generation
Stable Diffusion made high-quality text-to-image generation open and runnable on consumer GPUs, sparking a vast ecosystem of tools and fine-tunes. Here is what it is and why it mattered.
CLIP: Teaching AI to Connect Images and Words
CLIP learned to link pictures and language by studying hundreds of millions of image–caption pairs from the web. It quietly became the foundation for image search and generation.