21 results for “michabbb”
🎭 Pro Voice reproduces a speaker's character, emotional range and speaking style across languages from 30 minutes of audio, aimed at brand voices and professional use cases.
🤖 The voice agent SDK is built on #LiveKit. Stages stream in parallel, so each turn completes in under one second, and the agent stops talking when interrupted. It supports function calling, web search, retrieval, SIP telephony, WebRTC and existing IVR/contact-center software. LLM, voice and language are swappable.
🇪🇺 Voxtral 3 runs via Mistral's EU, US and global API endpoints or self-deployed on-premises, with control over audio retention. Voice agent API: $0.04/min with Mistral Small 4, $0.08/min with a larger model. #DataSovereignty
👂 Voxtral Small Transcribe 3 covers 50+ languages, including all 24 official EU languages, with diarization (who said what) and context biasing for domain vocabulary without fine-tuning. On FLEURS it reaches 4.1% WER, ahead of Gemini 3.5 Transcribe, Scribe v2, GPT Transcribe and Whisper large-v3. The 4B Mini variant costs $0.0035/min.
⚡ Voxtral Realtime Transcribe 3 is built for agents: sub-500ms latency in 35+ languages, native end-of-turn detection from acoustic and semantic signals, so the agent answers fast without cutting users off. Price: $0.008/min.
🗣️ Voxtral Mini TTS 3 adapts emotion, pace and emphasis to the conversation across 35+ languages and 50+ dialects, with explicit style controls, phoneme hints for unusual names and better handling of numbers, dates and email addresses. Price: $13 per 1M characters.
🎙️ #MistralAI released #Voxtral 3, a family of speech-to-text and text-to-speech models plus a voice agent SDK that runs listening, reasoning, tool calling and speaking as one real-time streaming pipeline. Target use cases: contact centers, service desks, virtual assistants and hands-free copilots.
🖨️ Light and dark themes, plus printable A4 PDF sheets (poster, technical drawing, explainer).
🔍 Self-verification built in: a geometry audit proves no part passes through another and every overlap is drawn in correct depth order. Headless Chrome scripts capture close-ups of every joint, and a line check fails any line that ends in mid-air.
🔧 Install via npx skills add wheresryan22/anatomy -g, then prompt: /anatomy explain how a token bucket rate limiter works. Needs #NodeJS 20.10+ with zero npm dependencies, plus Chrome for the checks.
⚠️ Limits: not a chart library, no people or exact maps, and builds are deliberately slow. 3D mode is beta. MIT licensed, 17.1k stars.
#DevTools #DataViz #JavaScript
https://github.com/wheresryan22/anatomy
💡 The core trick: it invents a physical mechanism whose behavior IS the concept. A cache key that skips unchanged work becomes a pin-tumbler lock that only turns when every pin matches. One GPU context serving many canvases becomes a gantry plotter whose single print head visits 16 wells.
🧱 Each figure is assembled from dozens of small shaded solid parts, framed in a card with title, hint and live readout, and moves with eased motion on pointer and keyboard. Every number shown comes from the same small model that drives the parts.
🎨 Three output kinds: plain SVG (standalone HTML, a single .svg or #React / #NextJS components), SVG + WebGL for fire, water, steam, caustics and glow, and a 3D beta mode where moving parts are re-projected and re-shaded every frame.
🔩 #Anatomy is an #opensource #AgentSkill that explains a technical idea by building it as a real, working object: interactive isometric illustrations of machines and systems in #SVG, with optional #WebGL shaders and 3D effects. Made for #ClaudeCode, also runs in #Antigravity, #OpenCode, #Codex and MiniMax Code.
🧵👇
🤯 Surprises: Voyage's code-specialized models score poorly here (they target retrieval for coding agents), while the general-purpose voyage-4-large rivals Jina. Gemini's semantic-similarity task variant sets a new bottom in the ranking
📏 256-dimension vectors show no quality loss compared to full size, often slightly better, and cut CPU and RAM costs for all-pairs comparison on large codebases
🛠️ Recommendation: jina-code-embeddings-0.5b with task code2code.query via API, voyage-4-large as alternative. For self-hosting: jina-embeddings-v2-base-code (Apache-2.0, runs on CPU via #Ollama). Script and test data are on GitHub
🔍 Scored as near-identical: reformatted code, ~15% additions, reordered statements, simple refactors. Not detected as duplicates: copies with renamed variables, and different algorithms solving the same problem. Identifiers weigh heavily: OpenAI's text-embedding-3 models rate unrelated code with shared names higher than identical code with renamed identifiers
🧪 Which #embedding models can actually detect duplicated code? A focused benchmark tested 12 models from #JinaAI, #VoyageAI, #OpenAI, #Cohere and #GoogleGemini on 19 cases × 10 code pairs, giving 190 data points per model. It is supplementary research for the #opensource duplication detector Slopo #AI #codequality
🧵👇
📐 The score is the gap between average cosine similarity on 13 high-similarity cases and 6 low-similarity cases. The wider the gap, the more precise a duplicate threshold can be. Adversarial cases (unrelated code with shared identifiers, unrelated code with the same framework boilerplate) penalize models that only compare surface text
🎙️ Portfolio: Mercury Voice minimizes time to first answer token for voice agents. Mercury Decide is a decision model for high-volume classification
🔌 Available on Inception API Platform, #OpenRouter, #Baseten and #Vercel. OpenAI API compatible, no rewrites needed. Enterprise options include dedicated capacity, autoscaling and compliance controls
https://www.inceptionlabs.ai/blog/introducing-mercury-3
💰 Pricing: $0.20 per 1M input and $0.70 per 1M output tokens. 50% launch discount ($0.10 / $0.35), cached tokens 90% off the input price, and 100M free API tokens to start
🤖 In production: browser agents completed 50% of 50 live Mind2Web tasks, ahead of GPT-6.1-Sol. Tutoring apps return hints in about a second, search agents hit 0.9s median generation time, plus voice and chat support
🌊 #InceptionLabs releases #Mercury3, the most capable diffusion-based #LLM so far. Benchmarks put it on par with or ahead of GPT-6 Luna and Claude Haiku 5.5 while running 4–8x faster #AI
🧵👇