# AI · Tech watch [Canonical source]() Artificial intelligence models and applications, their capabilities and limitations. ## [Dessa learns to speak without converting audio to text]() Unlike voice systems that turn speech into text before responding, Dessa processes audio directly, end to end. Trained only on human conversation, this speech-to-speech model runs full-duplex: it can listen and talk at once, interrupt, or punctuate an exchange with laughter and backchannels. Its demos also show how pauses and intonation convey meaning. Inference is reported to run faster than real time, with local execution on a MacBook Pro M4 Max or better. This is still a research preview; the team is inviting AI safety researchers to assess the risks of a model that perceives emotion in speech. 2026-10-28 · [Markdown]() · [Original resource]() ## [Clairic: typed extraction and estimates from unstructured data]() Cloudflare introduces Clairic, an extraction model available in preview on Workers AI. From text or images, it returns typed values such as dates, numbers, and lists, and can also compute, estimate, or forecast values, with confidence scores and intervals for estimates. Its responses are well-formed, but not necessarily correct. The API can process multiple fields in parallel and return extracted text with its source offsets. Cloudflare notes limits in the current checkpoint: multi-step arithmetic, estimates without supporting evidence, and detailed forecasting call for caution. The authors recommend leaving complex calculations to code and using confidence scores and intervals to guide review. 2026-10-24 · [Markdown]() · [Original resource]() ## [Voicebox: local voice cloning and speech generation]() Voicebox is an open-source desktop voice studio for macOS, Windows, and Linux, with local inference available. It combines voice cloning, speech generation across seven TTS engines, dictation into other apps, and voice conversations with agents. 2026-10-04 · [Markdown]() · [Original resource]() ## [An open-source local suite for voice workflows]() VoiceStudio is an open-source voice suite designed for local use, covering voice cloning and design, video dubbing, dictation, transcription, and audiobook creation. The repository advertises support for 646 languages, making it relevant to teams exploring audio workflows without relying solely on a hosted service. 2026-09-13 · [Markdown]() · [Original resource]() ## [Explore the open-source HeyGem digital-avatar project]() HeyGem is an open-source repository for digital avatars and talking-video generation. It is an older fork of Duix Avatar, offering a useful trail for following the project’s development. 2026-08-20 · [Markdown]() · [Original resource]() ## [Watermarks Remover: process content you own locally]() This project offers local tools to detect and remove AI-generated watermarks from content you own. Its repository includes an app, editor integrations, and batch-processing options. 2026-08-17 · [Markdown]() · [Original resource]() ## [Duix Avatar: generate AI avatar videos offline]() Duix Avatar is an open-source toolkit for generating digital-human videos offline. Its repository includes interfaces, resources, and deployment instructions for exploring the pipeline. 2026-07-29 · [Markdown]() · [Original resource]() ## [LingBot-Map: reconstruct 3D scenes from video streams]() LingBot-Map explores streaming 3D reconstruction using a geometric-context transformer. The repository accompanies a computer-vision research project with its code and resources. 2026-04-25 · [Markdown]() · [Original resource]() ## [Explore Microsoft’s open-source VibeVoice project]() VibeVoice is Microsoft’s open-source project for voice generation. The repository provides the code and materials needed to inspect its capabilities and implementation. 2026-03-30 · [Markdown]() · [Original resource]() ## [Flora: organize AI visual creation on a canvas]() Flora brings text, image and video models into a shared creative workspace. Its canvas supports exploring alternatives and turning an idea into a repeatable workflow. 2025-07-14 · [Markdown]() · [Original resource]() ## [NDC London 2025: topics from the .NET team]() This announcement outlines Microsoft’s planned sessions at NDC London 2025. Topics included .NET Aspire, cloud-native applications and AI integration with .NET. 2025-01-19 · [Markdown]() · [Original resource]() ## [Eraser: create diagrams from text or code]() Eraser accepts descriptions, files and code to build diagrams. Inputs such as Terraform or design documents help turn architecture into visuals teams can discuss. 2025-01-06 · [Markdown]() · [Original resource]() ## [Upscayl: upscale images using AI]() Upscayl provides AI-based image upscaling and enhancement. It targets low-resolution visuals that need to be reused at larger sizes. 2024-10-10 · [Markdown]() · [Original resource]() ## [WeAreDevelopers: explore talks, workshops and engineering experience]() The World Congress site collects the program for an event aimed at developers and engineering leaders. Sessions cover software engineering and practical uses of AI. 2024-03-01 · [Markdown]() · [Original resource]() ## [DragGAN: edit images by dragging points]() DragGAN’s research repository demonstrates interactive point-based editing of generated images. It provides code, setup requirements and demos for experimenting with the technique. 2023-09-09 · [Markdown]() · [Original resource]() ## [Stack Overflow’s 2023 survey on AI adoption and trust]() The survey separates AI adoption from trust: 70% of respondents use or plan to use AI tools, while 42% trust their accuracy. It provides a dated snapshot of developer expectations. 2023-06-15 · [Markdown]() · [Original resource]()