Natural Language Processing news and analysis on Web Pulse: 3979 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Gigatoken-rb, a Ruby port of the gigatoken BPE tokenizer, achieves 12 GB/s (2.78 billion tokens per second) on a Mac Studio M4 Max, running 1.6x faster than the upstream Python package on the same Rust engine. The gem is…
A developer built iTube, a pipeline that turns entire novels into narrated, lip-synced motion videos on a single 16GB GPU using FLUX, Wan2.2, PuLID, MuseTalk, and ComfyUI. The key insight is that audio length dictates sc…
A developer built Type Mac, a native macOS app that turns AI prompts into reusable keyboard shortcuts. The app removes the copy-paste cycle by letting users assign a shortcut to a prompt, such as fixing grammar or summar…
A new approach to reducing hallucinations in retrieval-augmented generation (RAG) pipelines involves using typed generation contracts that force large language models to adhere to strict schemas, such as Pydantic or JSON…
Knowledge-and-Memory-Management v0.0.2 has been released, stripping all hardcoded personal paths and replacing them with the portable $AGENT_HOME environment variable. The tool provides three collection pipelines for web…
A developer recounts spending weeks cycling through BM25, hybrid search, cross-encoders, and multiple embedding models for RAG performance, only to find that answer quality remained flat because the retrieval stage faile…
An AI notepad combines real-time transcription with human-guided notes to produce structured, searchable meeting documentation, unlike basic transcribers that dump raw text. The tool lets users write what matters while A…
Granola, an AI notepad for meetings, enhances user-written rough notes with specific quotes, figures, and context from the full transcript, aiming to close the gap between what is said and what ends up in the record. The…
The Poro 2 Long family of models, developed by LumiOpen in collaboration with TurkuNLP and the OpenEuroLLM project, extends the context window of the Poro 2 base model to 128k tokens while maintaining short-context perfo…
SSML (Speech Synthesis Markup Language) provides fine-grained control over AI speech output, including pronunciation, pacing, volume, pitch, and pauses, with support across major TTS engines like Google Cloud TTS, Azure …
A 2026 guide from this+that identifies 12 AI agents for real estate lead follow-up, citing that 80% of sales require five or more follow-ups yet 44% of agents give up after one contact, and reporting that brokerages usin…
Anthropic has expanded Claude's voice mode to support its Opus and Sonnet models, ending a period in which voice conversations were limited to the lighter Haiku model, unlocking longer reasoning and tool-heavy tasks via …
Willow, a voice dictation startup, announced a new platform on July 23rd, 2026, that measures success by a "zero edit rate" — the proportion of transcriptions that require no user edits before sending. The company argues…
HubSpot launched Agent Hub and Agent Builder in public beta on July 23 for Professional and Enterprise customers, providing a central console to build, monitor, and coordinate AI agents across marketing, sales, and servi…
Narya.ai launched Mumble Dictation, a local-first dictation app for Mac that keeps audio on-device and learns users' vocabulary, priced at $50 one-time. The app runs its full pipeline locally, including ASR and transcrip…
DKSplit v1.0.2, a fast character-level word segmentation library for Python, achieves 86.5% strict exact match accuracy on a 1,000-sample benchmark of domain prefixes, outperforming WordSegment (65.2%) and WordNinja (51.…
Amazon is launching an update to Alexa Plus that enables the assistant to connect with smart home devices from Bosch, Delta, Ecovacs, iRobot, Yale Home, Whirlpool, Tapo, Eufy, and others, automatically routing requests t…
Google's latest Google Home update gives Gemini for Home a 15-minute conversational memory, allowing users to issue follow-up commands without repeating context. The update also brings automation warnings in the Google H…
Researchers built RAGAL, a fully local retrieval-augmented assistant for the technical-support team of AFIR, the Romanian Agency for Financing Rural Investments, using only an 8 GB consumer laptop and zero data egress. O…
OpenAI announced on July 23, 2026, that ChatGPT Voice is now available in its desktop app, enabling users to control their computer and direct multiple agents running in ChatGPT Work or Codex using voice commands. The fe…