Tag
The user shared an AI video and audio workflow: generate silent visuals with Seedance 2.5, then use Sonilo to create and fine-tune background music and sound effects. This addresses the difficulty of adjusting audio when AI-generated video and audio are bundled together, while also promoting Sonilo’s Halloween membership offer.
Google DeepMind 发布了 Genie 4 世界模型研究预览版，新增实时音频生成、更逼真的可交互世界，以及支持中途提示词编辑的实时操控功能。该模型可维持最长两分钟的场景一致性，并与 Gemini 深度集成，用于机器人仿真、游戏和影视原型制作等场景。
AMBIT是一个统一框架，通过将自回归场景规划与rectified-flow渲染耦合，实现第一阶Ambisonics（FOA）空间音频的生成与编辑，并利用FOA感知的连续VAE与on-policy自蒸馏来保持空间结构和声学保真度。
inclusionAI released Ming-3.0-TTS-0.5B, a dense 0.5B multilingual text-to-speech model built on Qwen2.5-0.5B supporting zero-shot voice cloning, voice design, paralinguistic inline tags, and free-form style control across 7 languages plus Chinese dialects.
RubatoCodec 提出一种基于固定语句事件预算的自适应语音令牌化方法，结合流式解码，在低事件率下显著提升语音重建质量与可懂度，并在匹配的文本到语音对比中取得更低 WER 与更高 UTMOS 分数。
作者认为顶级AI实验室近期的短期重点应是让模型具备真实音频能力并在“taste（品味/质量感）”上实现类似Opus 5.5的跃升，从而让人们能自娱自乐，把算力更多用于实际生产力而非处理复杂社交障碍。
Drama3 模型现已支持在 audio.cpp 中运行，在 RTX 5090 上全速运行时最快可达实时速度的 5 倍，或在低于 8 GB 显存下以约 2 倍实时速度运行，仍需进一步优化与修正。
Fish Audio 推出其最新 AI 语音模型 Drama 3，支持通过自然语言控制情感、节奏、耳语、笑声和停顿，可生成多角色场景、有声书与游戏对话，覆盖 83 种语言，并以社区许可证开放权重供自托管。
Fish Audio 发布了 Drama 3 预览版文本转语音模型，支持用自然语言、括号内标签、音效与局部标记来精确指导每一句台词的表演方式，并可通过 API 调用。
Sundar Pichai 分享了 Argon 生成的 3D 动态字体作品，全部通过代码生成（包括音乐），无需预录素材或视频编辑软件，每帧都由 3D 网格、粒子模拟等数学算法绘制。
Fish Audio 发布开源语音生成模型 Fish Audio Drama 3 的技术报告：基于 35B 参数混合专家骨干（每 token 激活 3B），支持多参考语音克隆、自然语言导演指令、多轮修订、多人场景配音、编辑式填充、长音频与流式合成，单张 H200 推理速度达 271 帧/秒。
An Ask HN discussion exploring the gap in commercialized AI models for text + reference audio -> audio generation, noting that while image and text modalities are commoditized, controllable audio generation still lags behind.
A tweet praising ElevenLabs' new Eleven v4 audio model, highlighting a demo ad where the AI voice reads a script listing everything AI voices supposedly 'can't do.'
Podonos is launching Onepin, a production-ready voice AI tool that checks each generated line for naturalness and pronunciation, correcting mispronunciations like brand names automatically so a finished voiceover comes out directly from a script.
Fish Audio 发布新一代文本转语音模型 Drama 3，采用 35B 参数 MoE 架构（每 token 激活 3B），支持最多 16 个说话人的多角色长文本生成、语音克隆与自然语言控制，并可在 SGLang-Omni 上流式部署。
A survey paper that unifies joint audio-visual generation, cross-modal generation, and coupled editing into a single formulation over distributions of audio-video pairs, and proposes a taxonomy with five design axes plus nine edit categories covering 28 edit types.
YuE2 is a unified AI model for symbolic and audio music generation that uses symbolic planning to achieve high-quality results, outperforming previous baselines and competing with proprietary systems like Suno.
Gemini Notebook announces that their AI hosts' sound has been updated using Gemini 3.8 TTS, indicating an improvement in text-to-speech capabilities.
Google introduces Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, new text-to-speech models that enable expressive and customizable audio generation for creators and developers.
HyperFlow introduces a new MiniMax-H3 variant with open-weight LoRA that enables 3× faster inference through data-free self-distillation while preserving video and stereo audio outputs.