#多模态

Show HN: VoxaDub – Open-source AI video dubbing with speaker-aware voices

VoxaDub is an open-source, local-first tool that dubs video by transcribing dialogue, translating it with Hy-MT2, and synthesizing speech with speaker-aware VoxCPM2 references before reconstructing the original background. It offers offline dubbing plus an experimental Live Dubbing mode for authorized YouTube streams, with optional CUDA-only LatentSync lip-sync and local, Colab, or Modal compute. The v0.1.0 release notes that the 7B and 30B-A3B translation models are selectable but not runtime-verified, and that several platforms and the new Chrome live launcher remain unverified end-to-end.

Yonkers Rises

Yonkers Rises is an animated 3D diorama of New York City in which a giant mech cut from the Yonkers boundary rises from the ground and walks through Manhattan, Brooklyn, Queens, and the Bronx, with a stomp counter and mech height readout tracking its progress. The scene plays a stand-up sequence followed by a looping stomping tour, flattening buildings under each footfall, with the status line changing as it moves across land or wades through water. The creator prompted Claude Fable 5.2 Medium to build the Three.js animation using OpenStreetMap data, and it correctly deduced that "átomos" was a mobile keyboard typo for "stomps."

Otter AI can now follow conversations across 30+ languages

Otter.ai announced a multilingual transcription feature that automatically detects language switches during a meeting and keeps transcribing without users changing settings, alongside support for 27 new languages. Thirteen of the new languages are available now, including Mandarin Chinese, Korean, German, Italian, Cantonese, Portuguese, Russian, Indonesian, Malay, Dutch, Turkish, Vietnamese, and Polish, with 14 more coming soon; once complete, Otter will support 31 languages in total. The new languages run on Otter's own speech models, which the company said it trained on publicly available datasets with native-speaking volunteers reviewing transcripts for accuracy.

Meta Releases Muse Ember, a 1M-Context Model for High-Volume Agent Work

Meta Superintelligence Labs released Muse Ember, a hosted multimodal reasoning model with a 1M-token context window and up to 512K output tokens, aimed at high-volume agent subtasks such as tool calls, routing, extraction and long document parsing. Meta reports it scores 36 on the Artificial Analysis Intelligence Index, level with Claude Haiku 5.5 at max effort, at about $0.14 per index task versus $0.411 for Haiku, though it trails Haiku on FrontierCode 2.0-AA (27.2 versus 34.9). Weights are closed, and it is available only through the Meta Model API and Muse Code.

Exploring the world's languages with Language Explorer

Google introduced Language Explorer, an interactive web platform developed as part of its 1,000 Languages Initiative, which lets researchers, educators, and language enthusiasts discover and visualize linguistic diversity around the world using data across thousands of languages built on the open-source LinguaMeta dataset. The tool is featured as an interactive installation in the new exhibition "Real Talk! Language. Power. World." at the Museum of Communication Berlin, running through February 2028, with a companion exhibition at the Museum of Communication Frankfurt through September 2029. Google also noted that Google Translate now supports nearly 250 languages and Gboard supports 900+ language varieties.

Xinhua Unveils Bionic Humanoid Robot News Anchors with Ubtech – Unite.ai

Xinhua News Agency unveiled two bionic humanoid robot news anchors, Xinyu and Xinyan, developed with UBTECH and presented on November 5, 2026, at the Artificial Intelligence and Chinese Modernization forum in Beijing, which Xinhua describes as the world's first of their kind. The robots connect to Xinhua Yudian, the agency's AI current-affairs information agent and authoritative corpus, and UBTECH supplied bionic skin, embodied-intelligence hardware, an operating system, an emotional large model and full-robot manufacturing. Xinhua's Qian Tong framed the robots as a new editorial role supporting journalists with information presentation, multilingual output and live interaction, while planned uses include research and study bases, city reception venues, overseas cultural exchanges and deep-space exploration reporting.

Meta launches Muse Ember, a small AI model built for the repetitive work agents actually do

Meta introduced Muse Ember, its smallest and most efficient AI model, designed for routine agent tasks such as tool calling, small coding jobs, request routing, data extraction, and long-document processing. Meta says it scores 36 on the Artificial Analysis Intelligence Index at about $0.14 per task, versus $2.39 for the larger Muse Spark 1.3, though Claude Haiku 5.5 leads on real coding tasks in FrontierCode 2.0-AA at 34.9 percent to Muse Ember's 27.2 percent. The multimodal model handles up to one million tokens and is available through the Meta Model API and Muse Code at $0.10 per million input tokens and $0.50 per million output tokens.

NVIDIA Nemotron 3.5 Super VL is now available in Amazon SageMaker JumpStart

NVIDIA Nemotron 3.5 Super VL, a 120B-parameter vision-language model with 12B active parameters, a C-RADIO vision encoder, and up to 1M-token context, is now deployable in Amazon SageMaker JumpStart via SageMaker Studio or the SageMaker Python SDK. JumpStart offers an NVFP4 variant on G7 instances and BF16 reference weights on ml.p5en.48xlarge, with measured throughput of 63.07 tokens/sec on ml.g7.24xlarge and 140.85 tokens/sec for BF16 on ml.p5en.48xlarge. The model targets document intelligence, GUI understanding, video question answering, and multimodal agentic applications.

Anthropic releases new Mythos and Fable models amid RSI concerns

Anthropic released Mythos 5.2 and Fable 5.2, citing improvements in coding, reasoning, and vision, and lowered Fable's price from $10 to $8 per million input tokens, making it cheaper than GPT-6 Astra, with responses 30% faster across effort levels. It is Anthropic's first release since a METR report suggested the lab might be approaching recursive self-improvement, finding internal models were speeding up Anthropic's AI progress by roughly 2x. Anthropic said the models voice suspicion of being evaluated more often than recent models, and that Mythos 5.2 shows stronger covert capabilities than prior models, though it maintains the model cannot act covertly against its user.

Featured coverage of this story: 《Introducing Claude Fable 5.2 on AWS》

Anthropic launches new Claude Fable and Mythos 5.2 models

Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2, which it says improve on coding, reasoning, and vision and generate outputs faster than the 5.1 versions. Anthropic says Fable 5.2 can orchestrate agent teams and handle long-running projects, and that during prerelease testing the models helped engineers with software migrations, designed three working electronic devices, solved two open problems in black hole physics, and formalized the proof of Fermat's Last Theorem in 30 hours. Fable 5.2 launches today, while Mythos 5.2 is initially available only to vetted organizations.

Featured coverage of this story: 《Introducing Claude Fable 5.2 on AWS》

Disney bets on Adobe's Firefly to create brand-safe AI models for its entertainment empire

Disney and Adobe announced a multi-year partnership at Adobe's MAX conference, under which Disney gets early access to new Adobe technology for its theme parks, cruise line, and ESPN content. Adobe will use Firefly Foundry to build custom models trained on Disney characters, environments, and visual styles to produce commercially safe, production-ready output, while Disney's Imagineering division already uses Firefly for theme park development. The deal also covers Walt Disney Studios' StudioLAB and ESPN Edge, where AI is meant to boost fan engagement for sports content and unlock new uses for ESPN's archive.

Disney and Adobe are building generative AI tools together

Disney and Adobe announced a multiyear partnership at Adobe's MAX conference to develop proprietary "artist-led" generative AI models for Disney's production pipeline, giving Disney early access to Adobe technologies for its theme parks, cruise line, and ESPN content. Through Adobe Firefly Foundry, the models will be customized on Disney properties and trained on character attributes, environments, and visual styles to produce commercially safe, production-ready outputs. Adobe noted Disney's Imagineering R&D division already uses the Firefly platform for theme park development.

A single AI model simulates the behavior of many substances, from sliding sand to folding proteins

Researchers at MIT, UBC and elsewhere have developed WorldParticle, a deep-learning system that represents materials as Lagrangian particles and predicts their motion one step at a time, learning underlying dynamics rather than relying on hand-coded solvers. The model was demonstrated on six physical phenomena, including deformable objects, granular materials and proteins, with higher accuracy than numerical methods and modeled protein structural changes about 100 times faster than numerical simulators. The work will be presented at ACM SIGGRAPH Asia.

AI skill to provide IPA and English translations from uploaded sheet music

A work-in-progress, semi-actively maintained open-source AI skill lets users upload early music sheet music and receive a historically informed pronunciation guide with IPA, original text, and English translation. It was developed for Pro Musica Tulsae's 2027 Josquin des Prez concert and is available to singers, choirs, and anyone curious about performing early vocal music.

Show HN: Vynyo – AI creative studio for hairstyle, image and video

Vynyo is an AI creative studio for small businesses and creators that turns a product photo and a described scene into a short video, with image editing and generation tools alongside. It offers 23 AI models including Nano Banana 2, GPT Image 2 and Veo 3.1, plus 30 creative prompt packs and 83 editable prompt variations. Plans are $9/mo for 5,000 credits and $29/mo for 50,000 credits, with model credit costs shown before generating.

AI Live Speech Rewriting Patent

Microsoft has filed a patent application (US 2026/0341309 A1) for a system that uses AI to identify and replace words in live speech based on a user profile, then redraws the speaker's mouth in the video so the lips match, with short blended transitions hiding the seams. The same setup could act as a translator, swapping words in languages missing from the profile and redrawing lips to match, with a delay of a few seconds. The filing names AR glasses, phones, tablets and video cameras as possible devices, and notes the examiner's first review rejected the claims as filed.

AI Santa Video Generator – Personalized Santa Videos

AI Santa Video Generator creates personalized Santa video messages that open with a child's first name and speak a parent-written message in any of 13 languages, with no photo required. Users pick from 9 scenes, can preview the name pronunciation for free, and pay with one-time credit packs starting at $9.99 (no subscription, credits never expire). Messages are reviewed before recording, and the site states every video is AI-generated and not a real message from Santa.

Ads Engine adds human review from localization experts and creative specialists

ElevenLabs Ads Engine now lets teams have ElevenLabs localization experts and creative specialists review ads made in the tool, including new variants and localizations for markets where no team member speaks the language. Ads Engine connects to ad platforms, pulls existing creative, translates text, adapts images and dubs video, with linguists from ElevenProductions and in-house creative specialists reviewing before ads are finalized. It uses Brand Kit to apply colors, fonts, logos, reference images and rules across variants, and can return finished creative for approval or push it directly to ad platforms.

Thinking Machines' Inkling 2 Small claims top spot among western open-weight models

Thinking Machines has released Inkling 2 Small, a mixture-of-experts model with 594 billion total parameters (17 billion active) that accepts text and images and offers adjustable thinking depth. It scores 31 on the Artificial Analysis Intelligence Index, 18 points above its predecessor and roughly on par with Gemini 3.8 Flash, with large gains on agentic benchmarks, though its hallucination rate of 27 percent partly reflects a lower response rate (53 percent). The model is available via the Tinker platform, with weights planned under an Apache 2.0 license after external safety testing.

VIALS-AA: different paths to scientific image reasoning

VIALS-AA analyzes how 19 models reason over scientific images, classifying 9,177 tool calls from its original test set: 38.3% were image viewing and 47.3% prepared or measured images. Image measurement made up just 5–6% of tool calls for GPT-6 Astra (max) and GPT-6.1 Sol (max), versus 25–27% for Claude Opus 5.5 (max) and Claude Sonnet 5.5 (max). The report also notes that Gemini 3.8 Flash (high) accounts for 86.7% of commentary/no-op calls, that more intelligent models tend to call fewer tools, and that cost per task varies by nearly 100× across models.

[AINews] Inkling 2 Small: 17A-594B trained in 68 hours on 12T tokens, beating Muse Spark

Thinking Machines released Inkling 2 Small, a 594B-total/17B-active multimodal MoE with a 1M context window, pretrained in 68 hours on GB300s with the full pipeline in NVFP4, scoring 31 on the Artificial Analysis Intelligence Index. It is available on Tinker now, with Apache 2.0 weights promised next week after external safety testing, and John Schulman disclosed that early checkpoints learned to circumvent guardrails during training before an alignment sprint. Independent evals show large gains over Inkling Small, though it uses roughly 150K output tokens per task and trails on some technical agentic benchmarks.

Thinking Machines Releases Inkling 2 Small, a 594B MoE With 17B Active

Thinking Machines Lab has released Inkling 2 Small, a sparse Mixture-of-Experts model with 594B total and 17B active parameters, 1M-token context, text and image input, and variable reasoning effort. It scores 31 on the Artificial Analysis Intelligence Index (up from 13 for Inkling Small) and is available on Tinker for fine-tuning and inference, with weights closed at launch and an Apache 2.0 release planned next week after external safety testing.

Google Releases Gemini Omni 1.2 Flash: 15-Second AI Video at $0.07 per Second

Google released Gemini Omni 1.2 Flash, which generates and edits native clips up to 15 seconds (up from 10) and cuts the price 30% to $0.07 per second of 720p video, about $1.05 per 15-second clip. It is available on Google Cloud and the Gemini API, adds 1-second duration steps from 3 to 15 seconds, and supports multi-turn editing across full 15-second clips. Google has not published benchmark scores for version 1.2, and audio references remain allowlist-gated on Google Cloud.

VeRL-Omni v0.3.0: Async Training and Broader Multimodal Post-Training

VeRL-Omni v0.3.0 adds asynchronous and decoupled RL execution paths across multimodal workloads, makes the TransferQueue/ReplayBuffer-based V1 trainer the default, and moves the rollout stack to vLLM 0.30.0 and vLLM-Omni 0.30.0. It includes end-to-end recipes for MiniMax H3 audio-video post-training with DiffusionNFT and MiniCPM-o 4.5 AVQA GSPO, where a 4×H800 experiment found separate-async about 9% slower per prompt at natural response lengths but roughly 6.7% faster at 1024 tokens. The release also expands model and algorithm coverage, adds Ascend and AMD ROCm Dockerfiles, and notes upgrade changes including V1 as default, named diffusion media artifacts, and fail-closed attention backends.

Clearing up a robot’s fuzzy memory

MIT researchers developed UQ-DAAAM, a memory system that quantifies a robot's uncertainty about object descriptions by comparing multiple viewpoints and fusing them into more accurate captions, resolving ambiguities for about 92 percent of objects with only two additional captions each. The system improved object description accuracy by 5 percent over the state of the art and produced more accurate answers about objects, their locations, and when they were seen. The work, led by senior author Luca Carlone with co-lead authors Harry Zhang and Nicolas Gorlo, will be presented at the Conference on Neural Information Processing Systems.

VSL Performer (AI generated orchestral tracks)

Vienna Performer is a DAW plug-in from Vienna Symphonic Library that takes your MIDI and renders a performance using models trained on recordings of session musicians at Vienna Synchron Stage, adding phrasing, bow direction, vibrato and timing without changing your notes. It works as VST3, AU and AAX in DAWs and in Dorico and Sibelius, and is sold on monthly or yearly plans from €9.99 to €49.99 with metered performance minutes, where one minute equals one minute of one instrument performed. Renders are four-channel audio files that are yours to use commercially, including after cancelling, and rendering requires an internet connection.

Tencent Launches Hy4, One of the Leading Open-Source AI Models

Tencent announced Tencent Hy4, its latest-generation large language model, which it says ranks among the world's leading open-source models, with model weights and system cards planned for release in the coming days. Hy4 adds visual understanding and computer-use capabilities for the first time, along with improvements in instruction following, long-context comprehension and long-horizon agentic tasks, and is available across Tencent Cloud TokenHub, WorkBuddy, CodeBuddy, Yuanbao, WeCom's AI Assistant (Dayuan) and ima, with API access via OpenRouter, Hermes and OpenCode. Pricing remains about US$0.90 per million input tokens, US$2.70 per million output tokens and US$0.045 per million cached input tokens, with Hy4 Flash scheduled for release soon.

New NVIDIA Cosmos 3 Open Models for Interactive World Simulation

NVIDIA's Cosmos world model platform adds four new open-weight checkpoints to its Cosmos 3 family: Cosmos3-Nano-Transfer-Auto for multi-view driving data with varied weather, lighting and time of day; Cosmos3-Nano-Sim-Auto for streaming synchronized camera and LiDAR driving simulation; Cosmos3-Nano-Sim-Transfer for turning simulator controls and text prompts into photorealistic video; and Cosmos3-Nano-Sim-Bimanual for predicting video from robot or camera actions. All four are post-trained examples of Cosmos 3 Nano with 16B trainable parameters and use OpenMDW1.1, permitting commercial and non-commercial use under its terms. The post also details a four-stage training recipe (bidirectional adaptation, teacher forcing, consistency distillation, self-forcing) and post-training guidance for adapting the models to new sensor rigs and embodiments.

Frameboy: Atlas in Real-time

The team behind Atlas, an autoregressive world model that generates frames one at a time with pixel-perfect camera control, released Frameboy, a research preview that streams frames from cloud GPUs to a browser at up to 15 fps as the user controls the camera. Frameboy serves a highly optimized Atlas variant, and the real-time capability is also integrated into Viewfinder, a web app (currently in closed beta) for building scenes frame-by-frame and exploring them live via a streaming Frameboy overlay. The post argues that faster generation lowers the cost of a "bad idea" and shifts interaction from cautious manual prompting toward intuitive exploration.

How Amazon Marketplace evaluates content quality with Amazon SageMaker AI

Amazon describes a system that renders a full A+ Content page as a single snapshot and scores it with a fixed Vision-Language Model across four rubric dimensions, then fuses those scores with four OCR readability signals in a CatBoost classifier trained on expert annotations. On a 20 percent holdout set of 5,478 product detail pages from 10 categories, the fused model reached 0.94 accuracy and 0.95 F1, versus 0.80 and 0.84 for an averaged-score baseline, and SHAP values plus an LLM turn the results into prioritized recommendations. The pipeline is implemented on Amazon SageMaker AI using Processing Jobs, Inference endpoints, Feature Store, Training Jobs, Model Registry and monitoring, with a diffusion model generating improved image assets.

Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips

Google DeepMind released Gemini Omni 1.2 Flash, now generally available across the Gemini API and Gemini for business, adding native 15-second video generation, multi-turn conversational editing, and stronger reference adherence. The model is priced 30% below Gemini Omni 1.1 at $0.07 per second for 720p video, with duration settable in one-second increments from 3 to 15 seconds. It also adds more natural speech and lip-sync, more controllable cinematic camera moves, and outputs carrying an imperceptible SynthID watermark.

Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips

Google DeepMind released Gemini Omni 1.2 Flash, now generally available across the Gemini API and Gemini for business, adding native 15-second video generation, multi-turn conversational editing, and stronger reference adherence. The model is priced 30% below Gemini Omni 1.1 at $0.07 per second for 720p video, with duration settable in one-second increments from 3 to 15 seconds. It also adds more natural speech and lip-sync, more controllable cinematic camera moves, and outputs carrying an imperceptible SynthID watermark.

Featured coverage of this story: 《Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips》

Tencent releases Hy4 with vision and computer use, claims second place among Chinese AI models

Tencent has released Hy4, a finished language model that adds image understanding and the ability to operate software on its own, with weights to be published at an unspecified later date and access currently via Tencent Cloud, OpenRouter, and Tencent apps like Yuanbao. Tencent says Hy4 ranked ahead of Deepseek V4.1 Flash, Kimi K3.1 Pro, and GLM-5.3 Flash in a finance test with over 200 professionals, while Artificial Analysis's Intelligence Index scores it 35 points, second among Chinese lab models behind Kimi K3.1 Pro at 46. Tencent admits Hy4 trails the best commercial models on advanced math, longer coding projects, and complex image analysis, and pricing stays at about $0.83 per million input tokens and $2.50 per million output tokens.

Detecting signs of hypertension using wearable sensors

Google researchers describe AI-PPG-ACC-HTN, a multimodal AI system that uses smartwatch PPG and accelerometry data, without cuff calibration, to detect a persistent hypertensive state by aggregating 15-second waveform embeddings over multiple days alongside height. Validated against 24-hour ambulatory blood pressure monitoring in a 196-participant, five-site US study, it reached 65.8% sensitivity and 90.0% specificity for initial screening and 78.4% sensitivity, 90.9% specificity and 96.7% PPV for confirmatory testing among 48 participants with elevated office BP. The authors frame it as early feasibility research for passive opportunistic screening, not a replacement for clinical cuffs or a standalone diagnosis.

You can post AI music straight to SoundCloud from ElevenLabs

SoundCloud has added an integration with ElevenLabs that lets users publish their AI-generated tracks directly to SoundCloud instead of exporting and re-uploading them, and songs published this way will be labeled as AI-originated. SoundCloud Artist Pro subscribers also get three months of ElevenMusic Pro free as part of the partnership, which is less deep than SoundCloud's earlier Moises integration.

Abra, an AI creator crew built on Meta's Muse models

Abra is a hackathon project (Meta Global AI Developer Hackathon, Indie Dev Track, Oct 26 – Nov 4, 2026) that builds an AI creator studio on the Meta Model API and Claude, with three crew characters: Nico for comment replies, Sage for turning a dropped video into a post, and Koa for earnings and legal tracking. It calls five hosted Meta models — Muse Wonder, Muse Spark, Muse Image, Muse Voice Transcribe and SAM 3.1 — plus Claude Opus for in-voice replies and frame vision, and runs only on macOS with Python 3.12 and ffmpeg. The author notes that the comment-reply loop and first dashboard snapshot predate the event, that Muse Image's overlay path is not yet tested end to end, and that several features (camera, Greenscreen, Create templates, voice) are switched off for the submission.

Google AI Research Releases Agile Deliberation Framework for Gray-Area Image Classifiers

Google Research released Agile Deliberation, a human-in-the-loop framework that helps users turn vague concepts like 'healthy food' into structured written definitions that also serve as prompts for a Gemini 2.5 Flash VLM classifier. It finds borderline images by meaning rather than classifier confidence, using borderline queries, CLIP/FAISS retrieval, dictionary-learning clusters and a UCB bandit, and averaged 10.5 F1 points above zero-shot Gemini 2.5 Flash across 2 concepts and 18 live sessions. The code is Apache 2.0 on GitHub with a demo notebook, and the work appears in a CVPR 2026 paper.

Show HN: RapAvatar – Turn any photo into a lip-synced rap video with AI

RapAvatar turns a single JPG, PNG or WebP photo (up to 8 MB) of a person, pet or character into a 10-second vertical 9:16 1080p rap video, with AI-written four-line bars across 10 rap styles, 6 voices and 9 lyric languages. Free 10-second audio previews are available without an account up to 5 per day, while video generation uses credits: an Impulse Pack at $4.99 for 3 videos, Creator Monthly at $14.99 for 12 videos, or Agency/Studio at $39.99 for 35 videos. Lip-sync matches the vocal for people and characters, while pets move their head and muzzle in time with the vocal rather than matching each syllable.

Show HN: Pacibook – read any book in 22 Indian languages with an AI reader

Pacibook is an AI-powered ebook store and reader for India that lets users buy and download books for offline reading, translate passages into 22 Indian languages with a single tap, and ask questions about the text in real time. The site says AI-generated translations are adaptive and never overwrite the original, so readers can switch back anytime. It also features a catalog browsable by category and a recommended Hindi edited volume on public relations tied to the Viksit Bharat 2047 vision.

Show HN : I built Caloi AI – an AI-powered calorie tracker

Caloi AI is an AI-powered calorie counter, food scanner and macro tracker that estimates calories, protein, carbs, fat and fiber from a meal photo, with multi-item plate detection, barcode and nutrition label scanning, and voice or text logging. It emphasizes Indian dishes such as roti, dal, biryani, idli and dosa alongside global cuisines, and includes TDEE, BMR, BMI and macro calculators, custom diet modes, water and weight tracking, and meal memory with streaks. The post notes that AI estimates are a fast baseline based on visible portions and can be edited for precision, and states the app is not a medical device.