与「tronscription」相关的搜索结果

🚨50,00 * 10 $TRXI giveaway🚨 - Follow @trximarkets ✅ - Like and Retweet this Tweet ✅ - Tag 6 friends with hashtag #trxi#, #tronscription#, @trximarkets - Drop TRON Address. ✅ Winner announce in 24 Hour 🔥🔥🔥

⚪️ New App: Slate — a voice journal built on one rule: nothing leaves your phone. 100% Apple, top to bottom. Transcription by SpeechAnalyzer on device. Reflection by the 3 billion parameter Apple Intelligence model on the Neural Engine. Storage in SwiftData. Swift 6, SwiftUI, Liquid Glass. Zero third-party code in the binary. There's no account. No analytics. No tracking. No third party SDKs. No network calls. The app is private because there is nothing to collect. Turn on airplane mode and everything still works. There's no server. You speak your mind and slate writes it down. It notices patterns and reads your week back to you in a summary every Sunday. The local model is an observer. It never replies or advises or comforts. When it has nothing real to say, it says nothing.

Clone any voice. Keep it local. Voice Clone Lab turns 5–15 minutes of audio into a voice model on your own GPU, cleanup, transcription, fine-tuning, and generation included. Open source: github/tetsuo-ai/voice_clone_lab

Grok Build just received another major update, introducing powerful conversation navigation, enhanced enterprise controls, smarter terminal workflows, and extensive reliability improvements Release Notes: v0.2.102 — 2026-07-16 Breaking Changes: • --minimal and --fullscreen flags now apply only to the current session. Features: • New /jump slash command lets you quickly jump to any previous turn in the conversation. • New /timeline sidebar shows a clickable tick rail for fast navigation between conversation turns. • grok login now requests Grok Projects scopes so workspace listing works after consent. • Permission mode can now be set fleet-wide via remote config when no local setting exists. • Edit tool output has a setting to show a compact one-line summary instead of always-expanded diffs. • Tab completion in !bash mode now works like a normal terminal (prefix fill, dropdown, directory drill-down). • Enterprise deployments can now disable voice dictation via requirements.toml so /voice and Ctrl+Space are hidden for everyone. • User prompts now appear bold only in --minimal mode; fullscreen keeps normal weight. • grok plugin install now accepts a marketplace's registered name as a qualifier. • Consecutive edits to the same file now collapse into a single scrollback row when collapsed edit blocks are enabled. • Local sessions now inherit your shell environment variables and keep the current directory across commands. Bug Fixes: • Login and re-login no longer stack multiple device-code polls or leave stale flows running. • Background task tools now render with correct icons and titles instead of the generic MCP wrench. • Task tool now correctly validates and displays allowed model slugs for subagents. • Rewind now correctly handles bash transcripts, permission follow-ups, and sessions that mix old and new prompt markers. • Re-login during a session now immediately uses the new token instead of requiring a new session. • Terminal commands using globs now behave the same on zsh as on bash and no longer fail with shell errors. • Installer no longer replaces stowed shell configuration symlinks with plain files on upgrade. • Voice transcription now works with enterprise API bases and API-key authentication. • Fixed crashes on some network-mounted home directories by using a safer SQLite journal mode. • Home and End keys now move to the ends of the current wrapped line in the prompt. • Arrow keys and Esc now work correctly inside viewers opened from the dashboard. • Warns at startup when user and project sandbox profiles define the same name differently. • Billing upgrade links now show the full URL in the transcript (and copy it) when a browser cannot be opened. • Fixed Ctrl+Y yank no longer working after sending a prompt. • No longer shows permission prompts seconds after a turn was cancelled with Esc or Ctrl+C. • Page Up and Page Down now move the highlighted entry to the top or bottom of the visible scrollback area. • Conflicting project and user sandbox profiles now show a clear warning on the welcome screen. • OAuth login URLs no longer contain duplicate referrer parameters. • File links in official VS Code Remote-SSH terminals now use VS Code's native path handling. • Minimal mode now shows the folder-trust prompt after sign-in when required. • Skills whose names collide with built-in slash commands are now reachable via qualified names. • Fixed background task tracking when using grok -p --no-wait-for-background so tasks are properly reaped on exit. • Rate limit errors (429) now show specific server messages (capacity, team limits, free-usage) instead of generic upgrade prompts, with correct copy based on auth type. • /copy slash command is now available in minimal mode. Performance: • Improved recap and compaction behavior.

QVAC SDK 0.15.0 is live. This release adds multiple prompts batching, brings a native AMD GPU backend to the stack, moves more vision encoders onto mobile GPUs, and adds a second local coding-agent integration. Main highlights: - Prompt batching for the LLM addon. Batch multiple prompts into one job and process them concurrently, with each answer returned the moment its generation finishes. - Native AMD GPU backend. A first-class HIP/ROCm backend in @qvac/vla-ggml, auto-selected over Vulkan with clean fallback when ROCm is absent. - A second local coding agent. OpenClaw joins OpenCode for local, cloud-free agent workflows. AGENTS - OpenCode plugin update (@qvac/opencode-plugin). Aligned with the current SDK, CLI, and AI SDK provider packages. A fresh install runs @OpenCode against managed local QVAC models out of the box, from the default qvac/qwen3.5-9b, with no manual qvac serve setup. - OpenClaw plugin (@qvac/openclaw-plugin). A second coding-agent integration alongside OpenCode. A fresh setup installs the plugin, creates a local qvac provider through onboarding, and runs a QVAC model through @OpenClaw's local service path. LANGUAGE MODELS - Prompt batching (LLM addon). Batch multiple prompts in one job and run them concurrently, each answer returns the moment its generation finishes, no waiting on the others. - Reasoning-context trimming on hybrid + recurrent models (@qvac/llm-llamacpp). remove_thinking_from_context now works beyond pure-attention models. Same JS API, no throw. VOICE AND SPEECH - Transcription (transcription-parakeet 0.9.0). More robust CPU fallback on GPU failure and a faster Vulkan backend on Pixel 9. - Text-to-speech features (tts-ggml 0.4.0). Adds LavaSR for noise removal and adjustable output frequency up to 48 kHz, plus Japanese via Chatterbox. - Text-to-speech fixes (tts-ggml 0.4.1). CPU fallback on GPU failure, a q8_0 KV crash fix on Metal with Chatterbox. VISION - Qwen3.5 vision encoder on GPU (Android). Image encoder moves onto the phone GPU, with a smarter tile-grid preprocessor and default image-token caps, for flagship Android: Vulkan on Mali (Pixel 9 Pro) and OpenCL on Adreno 830 (Galaxy S25). - Gemma-4 vision encoder on GPU (Android). Vision encoder runs on the phone GPU instead of CPU, same flagship Android targets. PLATFORM AND PERFORMANCE - AMD GPU backend (@qvac/vla-ggml). Native HIP/ROCm backend, auto-selected over Vulkan with clean fallback when ROCm is absent (Linux x64 only). Comes with ~23% faster than Vulkan, ~14% faster than PyTorch-ROCm, parity preserved. Unified code style. A cleaner, more consistent, easier-to-contribute codebase. Let's build. npm install @qvac/sdk

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: Huggingface: API: Live demo: Technical Report： HF Space: AtomGit： SGLang-Omni: vLLM: MLX-audio: Discord:

Voice agents, now on Vercel. Realtime, speech and transcription are now live on AI Gateway. Build with 𝚞𝚜𝚎𝚁𝚎𝚊𝚕𝚝𝚒𝚖𝚎, 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚂𝚙𝚎𝚎𝚌𝚑 & 𝚝𝚛𝚊𝚗𝚜𝚌𝚛𝚒𝚋𝚎 on AI SDK 7.

QVAC SDK 0.14.0 is live. This release makes the on-device stack faster on mobile, ships the developer-agent path, and takes local text-to-speech to 31 languages. Main highlights: - OpenCode and OpenClaw. The first official OpenCode plugin, plus a maintained OpenClaw compatibility path, both built on managed mode and qvac serve. Point a coding agent at a local model with far less setup and far fewer surprises. - Brain-computer interface transcription, on the SDK. Take recorded neural signal data and decode it into text, fully on-device, no cloud. Stream it in chunks through a simple API. In 0.14 it runs GPU-accelerated on iOS. - Text to Speech in 31 languages with our Supertonic3 upgrade. VOICE AND SPEECH - Supertonic3 multilingual TTS, 5 languages to 31. - Chatterbox and Supertonic now run on the Android GPU, with lower memory use (especially on iOS), quantized s3gen Chatterbox support, and a fix for Chatterbox occasionally emitting random speech. - Whisper transcription now runs on the iOS GPU. Parakeet runs on the Android GPU, with steadier real-time streaming. VISION AND OCR - VLM multi-tile batching: high-resolution Pan and Scan images are encoded in one pass instead of tile by tile, for faster vision throughput. - OCR on ggml (EasyOCR and DocTR) reaches full speed parity with the onnx path, across Metal, OpenCL, and Vulkan. PLATFORM AND RELIABILITY - Dynamic compute backends on Linux: one build picks the right backend at runtime, and opens the door to ROCm and CUDA support without per-backend builds. - Thinking tokens are kept out of the model context, so reasoning no longer fills the KV cache. SDK 0.14.0 is now leaner and faster to start. Let’s build.

3 patterns for multimodal RAG. Here's how they differ and when each one breaks down. Most RAG systems add multimodal support by converting everything to text first. Is your system natively multimodal, or just a conversion pipeline? The architecture choice shapes what you can query and what you lose. Shared vector space - Cross-modal search without format conversion - Requires large multimodal training datasets - Semantic drift is a real risk if training data is narrow Single grounded modality - Works with any existing text search setup - Spatial relationships in images don't survive conversion - Retrieval quality depends on captioning/transcription accuracy Separate retrieval pipelines - Best per-modality retrieval accuracy - Most complex to rank across modalities - Highest compute cost, independent search per modality Pick your pattern, clone the repo, and build it.