Tag
NVIDIA released NemotronLabs RE-USE, a 9.6M-parameter multilingual universal speech enhancement model that restores speech quality across diverse degradations and sampling rates, outperforming larger open-source USE models while being ready for commercial use.
The author breaks down the engineering design of meeting notes in the Feishu (Lark) desktop app and uses it to build a Mac meeting notes tool (soon to be open-sourced), highlighting three core design elements: chunked audio uploads, parallel AI meeting notes tasks, and traceable charts.
Norah Studio is a creative studio app for exploring and transforming music on Mac, iPad, and iPhone, offering track isolation, tempo/key adjustment, chords, effects, harmonies, and audio/MIDI export, all running on-device.
VoxLock 是一个无需注册步骤的因果性目标说话人提取模型，通过微调 REAL-TSE 实现，在真实录音上达到接近商业级的语音识别精度（Krisp 基准 WER 4.95%），且 CPU 占用和延迟极低，模型与代码均已开源。
Alkove is an open-source (Apache-2.0) on-device voice front-end made of five tiny models (7.5 MB total) — a typed wake-word system, voiceprint-based speaker filtering, and personal VAD — running on a custom Rust inference engine with npm, pip, Android, LiveKit, and Pipecat packages, aimed at reducing false wake-ups and cross-talk before STT in local voice assistants.
Maybloc launches as a new state-of-the-art speech-to-text product for multi-speaker audio, claiming the lowest diarization error across several benchmarks, supporting 45 languages and up to 25 speakers, with add-ons like subtitles, PII redaction, summaries, and translation.
Audio Lip Sync Pro is a Blender tool that generates high-quality lip-sync facial animations directly from audio clips, running fully locally with no account or network access required. Unity and Maya releases are planned.
DuET is a dual-encoder transformer that jointly performs MIDI-to-score transcription and beat tracking to improve timing and duration accuracy in expressive piano performance transcription, achieving large MUSTER error reductions on the Gangqinpu dataset.
作者讲述了如何通过黑盒测试与信号分析，逆向工程一款商业空间音频处理程序，并在 Linux 上使用 PipeWire 复现其效果，涵盖了 HRTF、房间响应与心理声学的实践探索。
The paper introduces PawCT, the first end-to-end neural framework for part-aware choral transcription that identifies active SATB vocal parts and transcribes each into a separate note-level track, achieving substantial F1 improvements over prior choral AMT baselines on the YouChorale dataset.
A practitioner shares how their ASR pipeline failed on code-switching calls because language detection ran once at the start, mangling the Hindi portion of English-Hindi utterances; they now run continuous detection and route unstable chunks to a review queue, exposing how benchmark numbers diverged from real-world caller behavior.
Patchcord is a macOS menu bar app that lets users run any Audio Unit plugin on their microphone and route the processed audio as a virtual mic to Zoom, Teams, Meet, or Discord, with per-mic EQ and a live spectrum display.
This article explains how to achieve high-quality digital audio playback on the ZX Spectrum's 1-bit beeper using assembly language and specific timing techniques, verified on both period hardware and modern clones.
The article explores how voices in one's head, such as auditory hallucinations, are being reframed as beneficial features in technological and psychological contexts.
The paper proposes optimizations to the Viterbi algorithm using the Hirschberg algorithm and constrained random walk, reducing memory usage from 140 GB to 5 MB and improving speed, enabling forced alignment to run on end-user devices for better scalability in speech processing.
Foleyfy is an iPhone app that converts short recordings into organized sound-effect packs for games, featuring on-device processing and currently in development.
FRAUDSkill is a structured frozen-weight adaptation framework for audio anti-fraud detection that optimizes external skill programs without modifying the underlying audio-language model, achieving higher accuracy and reduced invalid outputs.
A tutorial for Septabee, a free digital audio workstation under active development, showcasing its unique interface, high-quality sample-accurate audio processing, and advanced features like a custom programming language for synthesis.
Soundmap is a web tool that visualizes sound in real-time using a webcam and a connected smartphone, turning audio into visual displays.
Desert Ant Labs, a European AI lab, launches a suite of small, specialized on-device models for audio, vision, and text, offering fast inference and privacy benefits by running locally on devices.