Search

OpenMOSS-Team/MOSS-Audio-8B-Instruct · Hugging Face

MOSS-Audio is an open-source multi-capability audio understanding model from the OpenMOSS team and partners, offering four variants (Instruct/Thinking, 4B/8B) that unify transcription, speaker/emotion analysis, scene cues, music analysis, QA, and multi-hop reasoning, built with a three-part architecture and a deep cross-layer design to retain acoustic details.

TTFAだけ速い音声AIにしない：先行キューと「最初の意味ある音声」を分けて計測する #TypeScript - Qiita

A TypeScript demo models an audio-based QA flow: it streams an LLM response, plays a receipt audio while the answer is generated, then synthesizes and plays a semantic audio track, with careful cancellation handling and timing traces.

GitHub - leohuang2013/pyannote-audio_speaker-diarization_cpp: C++ version of pyannote audio speaker diarizaiton pipeline · GitHub

This repository ports pyannote-audio’s speaker diarization pipeline to C++, using ONNX models for segmentation and embeddings and C++ code for clustering. It includes build, run, and verification instructions, but reports slow performance and small output differences that can affect diarization results.

[2610.32749] Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes

A new real-world benchmark (RealSSA) and a causal AVTSE method (Falcon) deliver efficient, high-quality target speaker extraction, with substantial reductions in model size, compute, and latency while supporting multiple interference scenarios; code and demo are available.

[2611.06074] Which One Is Sounding? Binaural Sounding Instance Disambiguation for Audio-Visual Segmentation

BiSID-AVS introduces binaural audio cues to disambiguate sounding instances among same-class candidates in audio-visual segmentation, refining candidate regions into the true sounding instance and showing strong gains on synthetic and robotic datasets.

Stem splitter - splitting a song into AI tracks, price | Promptowy

Translated title. Original: Stem splitter - rozdzielenie piosenki na ścieżki AI, cena | Promptowy

The document explains how a stem splitter tool works, detailing that AI-driven separation produces vocal, drums, and bass from a mixed track for a per-minute price. It covers workflow steps, pricing, result validation, and alternatives for single-track needs.

Spotify on AI music: "People will try to abuse this

Translated title. Original: Spotify om AI-musiken: ”Folk kommer försöka missbruka det här”

Spotify is filtering out millions of AI-generated songs from user recommendations to curb spam, while still allowing users to generate remixes; the company notes AI use does not imply poor music and cautions that pay-for-profit abuse may occur.

ASCII.jp: JBL 7.1-channel soundbar drops from 121,000 yen to 76,000 yen! Enjoy sound from the back with detachable wireless rear speakers

Translated title. Original: ASCII.jp：JBLの7.1chサウンドバーが12万1000円→7万600円！ 着脱式ワイヤレスリアスピーカーで後方からも音を楽しめる

JBL BAR 800MK2 is a 7.1-channel soundbar with detachable wireless rear speakers and a subwoofer, supporting Dolby Atmos/DTS:X, room calibration, and streaming features; it is discounted on Amazon from 121,000 to 70,600 yen.

The profession AI won’t take away from music, according to Ricardo Confessori - Whiplash.Net

Translated title. Original: A profissão que a IA não vai tirar da música, segundo Ricardo Confessori - Whiplash.Net

Ricardo Confessori and producer Thiago Larenttes discuss AI’s role in music, arguing that producers remain needed and that full AI-generated songs can weaken a composer’s identity. Confessori’s album "Realm Breaker" tells a story about a scientist losing control of his creation, but was made through traditional songwriting and production.

[2610.38252] A Tutorial on Generative Models for Music Audio

This tutorial explains neural methods for music audio generation, covering codecs, autoencoders, autoregressive and diffusion models, sampling, and optimization. It connects training to playback tasks, discusses evaluation and legal considerations, and includes related audio modalities like speech and sound effects.