Audio, voice and music
Models for speech, voices, sound and music, from transcription and text to speech to music generation. The Atlas has logged 177 launches and papers on it since 5 January 2023, 108 of them in 2026.
Google DeepMind, Alibaba Qwen (Tongyi) and OpenAI have the most.
Key launches
The 60 most significant of 177, newest first. Every launch in the Atlas
- Six new MAI models (including MAI Code 2) Microsoft · 17 November 2026
Microsoft unveiled six AI models at Ignite covering coding, cybersecurity, images and voice, and said it will license the models' weights for customers to customize. - MAI Code 2, MAI Code 2 Flash, MAI Cyber 2 Flash, MAI Image 2.7, MAI Realtime 1, Dragon Copilot Transcribe and Decision 1.1 Microsoft AI · 17 November 2026
Microsoft AI released seven new MAI models at Ignite 2026, covering coding, cyber defense, image, realtime voice, medical transcription and routing, priced well below rival models. - Suno Visuals Suno · 17 November 2026
Suno launched Visuals, which turns songs into lip-synced AI video clips of up to 15 seconds and generates images, using H3 Move, a model based on MiniMax's H3. - Visuals Suno · 17 November 2026
Suno launched Visuals, a feature that lets users generate 15-second video clips for their songs and join them into a full music video, built on a MiniMax model. - Tofu-1 Bland · 15 November 2026
Bland released Tofu-1, a full-duplex speech-to-speech model for phone calls, in a closed enterprise research preview. - Hy Realtime Tencent Hunyuan · 13 November 2026
Tencent Hunyuan released Hy Realtime, a real-time voice model that decides when to search the web and can sing songs on request. It is live in Yuanbao voice calls. - Full suite of application APIs (web search, file input, URL reading, OCR, ASR, TTS) DeepSeek · 13 November 2026
DeepSeek opened its app's web search, deep thinking search, file input, URL reading and safety checks inside its APIs, plus standalone OCR, ASR and TTS, so developers can rebuild… - MOSS-Live MOSI.AI · 13 November 2026
MOSI Intelligence and OpenMOSS released MOSS-Live, a full-duplex real-time voice system that ranks first on six full-duplex metrics against GPT-Live-1 and Doubao SeedDuplex. - Muse outbound business calls Meta · 12 November 2026
Meta's Muse consumer AI agent can now place phone calls to US businesses for paid subscribers, after a September test with early adopters. - Seedance 2.5 Flash ByteDance · 12 November 2026
ByteDance's Volcano Engine released Seedance 2.5 Flash, a lighter video model that it says is twice as fast as Seedance 2.5 and about 34% cheaper at 720p. - Gemini smartglasses Google · 12 November 2026
Google and Samsung launched Gemini smartglasses with Warby Parker and Gentle Monster, open for preorder from $395 and shipping December 10. - Android XR Intelligent Eyewear Samsung · 12 November 2026
Google and Samsung opened US and South Korea preorders for Gemini-powered Android XR smart glasses without a display, in Warby Parker and Gentle Monster frames from $395,… - Gemini Decision 1 Google · 11 November 2026
Google released Gemini Decision 1 and the Gemini Decisions API, which scores up to 1,000 closed-set questions per request in one forward pass at $0.10 per 1M input tokens. - ElevenAI ElevenLabs · 11 November 2026
ElevenLabs released ElevenAI, an experimental app for meeting notes, dictation and agent integration aimed at teams, in preview with early access signup. - Muse Ember Meta · 9 November 2026
Meta released Muse Ember, its smallest and cheapest Muse model for high-volume agent work, at $0.10 per million input tokens and $0.50 per million output tokens with a 1M-token… - Gemini Omni 1.2 Flash Google DeepMind · 5 November 2026
Google released Gemini Omni 1.2 Flash, a video generation and editing model with native 15-second clips and pricing 30% lower than Omni 1.1. - Woof Underdog · 5 November 2026
Underdog released Woof, a 31.7MB text-to-speech model that runs in a browser, and showed it in a browser game. The model can also generate music. - Audio Intelligence Apple · 3 November 2026
Apple began beta testing Audio Intelligence for Apple Watch Series 12 in iOS 27.3, a set of four opt-in AI features for sounds, songs, conversation catch-up and notes. - WAND-Vega Tencent Cloud · 30 October 2026
Tencent Cloud launched WAND-Vega, a multimodal application model, with a Real-time Intelligence service and a contact-center voice agent, at its Global Digital Ecosystem Summit… - Apriel Voice 27B Preview ServiceNow (Apriel) · 29 October 2026
ServiceNow released Apriel Voice 27B Preview, an open speech-language model built on Gemma 4 26B-A4B that handles speech, reasoning and tool calls in one model. - Nemotron 3.5 Transcribe NVIDIA · 28 October 2026
NVIDIA released Nemotron 3.5 Transcribe, an open-weight promptable model that transcribes 40 languages and handles recordings up to an hour long, alongside the Granary v2 dataset. - Prism Rime · 28 October 2026
Rime launched Prism, which it calls the first fine-tunable speech-to-speech model built for enterprises, with a verifiable text checkpoint before it speaks. - Genie 4 Google DeepMind · 27 October 2026
Google DeepMind announced Genie 4, a world model that generates interactive simulations with native audio, music and dialogue, now in research preview for trusted testers. - Apple Creator Studio AI assistant for Keynote, Pages, Numbers and Logic Pro Apple · 27 October 2026
Apple is adding a beta AI assistant built on Apple and OpenAI models to Keynote, Pages, Numbers and Logic Pro, and MCP support to Keynote, Pages and Numbers. - Hy-SimulMT Tencent · 27 October 2026
Tencent Hunyuan open-sourced Hy-SimulMT, a Chinese-English simultaneous interpretation model that clones the speaker's voice and lags about 3 seconds on long-form speech. - LFM3 Liquid AI · 27 October 2026
Liquid AI released LFM3, a family of six 2B-class open-weight models for text, vision, audio, omni and decision tasks, built on one shared backbone and 22T tokens of pre-training. - SparkAudio 1.0 iFLYTEK · 26 October 2026
iFLYTEK detailed SparkAudio 1.0, a speech foundation model with a 0.65B audio encoder and a 30B-A3B MoE language model, covering 99 languages and 202 dialects. - Dessa Block · 23 October 2026
Block announced Dessa, a speech-to-speech model trained from scratch on only human conversation, which can run locally on a MacBook Pro M4 Max or better. - Voxtral 3 Mistral AI · 20 October 2026
Mistral AI launched Voxtral 3, a speech-to-text model, a text-to-speech model and a LiveKit-based voice agent SDK for real-time voice agents. - Ming-3.0 speech models Ant Group · 20 October 2026
Ant Group open-sourced the Ming-3.0 speech family, five models for speech synthesis, audio understanding and conversation, and speech recognition, with the TTS models under the… - Eleven Realtime ElevenLabs · 20 October 2026
ElevenLabs launched Eleven Realtime, speech to speech voice agent models in two sizes, v1 Mini and v1 Max, with an API compatible with OpenAI's Realtime API. - MiDashengLM-Spatial Xiaomi · 16 October 2026
Xiaomi open-sourced MiDashengLM-Spatial, an audio-language model that understands what a sound is and where it comes from, with weights and code under Apache 2.0. - Prompt-to-voice for GPT-Live OpenAI · 16 October 2026
OpenAI released prompt-to-voice for all API customers, which creates a custom GPT-Live voice from a text description through the API or the voice Playground. - Bee Frontier and Bee Eclipse Amazon (Bee) · 15 October 2026
Amazon's Bee launched the $74.99 Bee Frontier wearable today and announced the $99.99 Bee Eclipse phone-back recorder, plus an $11.99 per month Bee Pro subscription. - Granite Switch 4.1 IBM · 15 October 2026
IBM released Granite-Switch-4.1-8B and -30B, which combine chat, a decision adapter, a Guardian safety adapter and fast speech recognition in one model served from one endpoint. - Decisions integration in GPT-Live OpenAI · 15 October 2026
OpenAI added native Decisions integration to GPT-Live sessions, so apps can act while the user is speaking or react to what Live says. - Grok Voice 3.0 SpaceXSI · 14 October 2026
SpaceXSI released Grok Voice 3.0, a speech-to-speech model that it says ranks first on the Artificial Analysis speech-to-speech index, responds in under a second and costs $0.08… - Eleven v4 (and v4 Turbo) ElevenLabs · 28 September 2026
Eleven v4 is ElevenLabs' most emotive TTS: inline delivery tags, consistent multi-speaker dialogue, 90+ languages and a ~150 ms Turbo variant. - UMG and ElevenLabs multi-year licensing agreement Universal Music Group / ElevenLabs · 10 September 2026
ElevenLabs signed its first major-label deal, a multi-year UMG licence and collaboration that starts with a fan platform for remixes and mashups. - Suno v6 (v6, v6-wild, v6-mini) Suno · 9 September 2026
Suno v6 is its first model built with the music industry (Warner Music Group, BMG, Believe), with section edits, mashups, sampling and text/audio/image/video prompts. - Wan3.0 (video) Alibaba (Qwen) · 24 August 2026
Hosted Wan3.0 makes native 30-second clips up to 1080p with audio and accepts documents, slides, spreadsheets and webpages as references. - Seedance 2.5 ByteDance · 31 July 2026
Seedance 2.5 makes 30-second single-pass clips with synced audio, up to 50 multimodal references, region-level edits and 4K, previewed 2026-06-23. - MiniMax H3 (Hailuo) MiniMax · 31 July 2026
MiniMax H3 is an open-weights audio-video model that makes 2K, 15-second clips with native stereo sound, priced below a third of mainstream models at 2K. - FLUX 3 (early access) Black Forest Labs · 23 July 2026
FLUX 3: one multimodal flow model for image, video with native audio, and robot-action prediction; open FLUX 3 Dev promised. - Gemini Omni (Omni Flash) Google DeepMind · 19 May 2026
Gemini Omni Flash makes video from mixed image, audio, video and text input and edits it conversationally; first of a planned family. - Gemini Omni Flash Google DeepMind · 19 May 2026
Gemini Omni Flash takes text, images, audio and video in one prompt and outputs editable video, folding the Veo line into Gemini itself. - Seedance 2.0 ByteDance · 12 February 2026
Seedance 2.0 jointly generates audio and video from text, image, audio and video references (9+3+3 per prompt), and Hollywood objected to it. - Kling 3.0 Kuaishou · February 2026
Kling 3.0 unifies video, image, audio and editing, with multi-shot storytelling, native audio, multilingual dialogue and up to 15 seconds per generation. - WMG and Suno settle; licensed models promised for 2026 Warner Music Group / Suno · 25 November 2025
Suno settled with Warner Music Group and agreed to launch licensed models in 2026; downloads move behind paid tiers, and Suno acquired Songkick from WMG. - ERNIE 5.0 Baidu · 13 November 2025
Natively autoregressive omni-modal model trained from scratch on text, image, video and audio with a 2.4T-parameter ultra-sparse MoE (about 3% active). - UMG and Udio settle, plan licensed 'walled garden' platform Universal Music Group / Udio · 29 October 2025
Universal Music Group and Udio settled their copyright suit with a payment and licences; Udio disabled downloads ahead of a licensed 2026 platform. - Sora 2 and the Sora app OpenAI · 30 September 2025
Sora 2 generates video with synchronized dialogue and sound, better physics, and "cameos" of real people, launched with a social Sora app. - Suno v5 Suno · 23 September 2025
Suno v5 delivers more natural vocals and a new composition architecture, described as 'best to date', for Pro and Premier first. - Eleven v3 (alpha, GA 2026-02-02) ElevenLabs · 3 June 2025
Eleven v3 adds audio tags ([whispers], [laughs]), multi-speaker dialogue and 70+ languages; the most expressive ElevenLabs TTS at launch. - Veo 3 Google DeepMind · 20 May 2025
Veo 3 generates video with synchronized dialogue, sound effects and ambient audio; Google calls it the first video model with native audio. - NotebookLM Audio Overviews Google · 11 September 2024
NotebookLM turns uploaded documents into a two-host podcast-style conversation, powered by Gemini 1.5. - Record labels sue Suno and Udio Universal, Sony, Warner (via RIAA) · June 2024
Major labels, coordinated by the RIAA, sued Suno (Massachusetts) and Udio (New York) for training on copyrighted recordings, seeking up to $150,000 per work. - GPT-4o OpenAI · 13 May 2024
One network handles text, audio and images end to end; GPT-4o costs half of GPT-4 Turbo and reaches free ChatGPT users. - Suno v3 Suno · 21 March 2024
Suno v3 is the first model Suno called 'radio-quality'. It makes two-minute songs with vocals from a text prompt and is free to all users. - VALL-E Microsoft Research · 5 January 2023
VALL-E treats text-to-speech as language modelling over neural-codec tokens and clones a voice from a 3-second sample, trained on 60K hours.
Most active labs
- Google DeepMind 29 launches and papers
- Alibaba Qwen (Tongyi) 20 launches and papers
- OpenAI 15 launches and papers
- ElevenLabs 10 launches and papers
- Suno 9 launches and papers
- ByteDance Seed 8 launches and papers
- Meta Superintelligence Labs (MSL) 5 launches and papers
- Microsoft AI (MAI) 5 launches and papers