Models1 Oct 2026
Suno opens a beta of Speech, a model that voices text over its own music
Suno's new beta generates spoken word and its musical backing as one track, pitched at bedtime stories, pep talks and other personal audio.
News · Models
News
Short original briefs on AI in culture, each linking out to the original reporting rather than reproducing it.
Models30 Sep 2026
Utopai Studios releases a new PAI production platform and its own video model, Utopai X
Models29 Sep 2026
ElevenLabs' Eleven v4 clones a voice from 10 seconds and takes stage directions in 90-plus languages
The new model lets writers script laughs, accents and ambience inline and keeps a cloned voice across languages, while consent checks still rely on uploaders.
Latest
Models29 Sep 2026
Kuaishou's Kling unit previews Kling 4.0, with 30-second clips and keyframe control
Kling 4.0 doubles clip length to 30 seconds and adds multi-reference input and keyframes, as Kuaishou's AI video unit chases ByteDance's Seedance for studio and creator business.
Models26 Sep 2026
Google's new Gemini voice models clone a voice from 30 seconds, but only with the owner's spoken consent
Gemini 3.8 Flash TTS can copy a voice from a short sample only after its owner records matching consent, and the feature is switched off in several regions.
Models24 Sep 2026
Google rolls out Gemini 3.8 Flash TTS, voice models that take acting directions and copy an authorised voice
Google's new speech models take line-by-line acting direction, design voices from text and clone one from 30 seconds of audio if the speaker has consented.
GR
Models19 Sep 2026
Suno's paying users say v6 sounds worse than the models it replaced
Longtime subscribers describe Suno's licensed v6 output as muffled and lifeless and want the retired models back, a reaction that could cost the platform engagement.
Models15 Sep 2026
ElevenLabs ships Music v2.5 a day after its Universal deal, and its ownership promises don't quite match
ElevenLabs' new music model went live the day after its first major-label agreement, but its three statements on who owns the output say three slightly different things.
Models9 Sep 2026
Suno launches v6, its first models built with major-label partners
Suno retired its old models for v6, built with Warner, BMG and Believe, while still fighting suits from Universal, Sony, GEMA and a class of artists.
Models25 Aug 2026
ElevenLabs adds Composer, a section-by-section song editor, to ElevenMusic
ElevenMusic's new Composer lets users regenerate or restyle one part of a song while keeping the rest intact.
Models24 Aug 2026
Alibaba puts Wan3.0 in beta, pitching 30-second AI video for films and short dramas
Alibaba's new video model doubles clip length and accepts documents and web pages as input, aimed at uses from film production to creator content.
Models18 Aug 2026
Alibaba enters AI song generation with HappyShrimp 1.0 and a Chinese label partner
Alibaba's new prompt-to-song model reaches public beta with Taihe Music as a co-creation partner, putting a tech giant up against Suno and Udio.
Models1 Aug 2026
ByteDance releases Seedance 2.5, generating 30-second video clips with sound
The new version of ByteDance's video model makes longer clips with audio in one pass and takes dozens of reference files, shown off with an all-AI short film.
Models30 Jul 2026
Google releases Lyria 3.5 in Flow Music, saying nothing this time about training or watermarks
Google's new music model promises more expressive vocals, but its launch post dropped the training and copyright assurances that came with Lyria 3.
Models16 Jul 2026
Foundry releases SmartRoto, an AI plugin that animates artists' roto splines in Nuke
Foundry's $499-a-year SmartRoto leaves shape drawing to artists and uses a locally run model, trained on licensed footage, to keyframe splines up to four times faster.
Models28 May 2026
ElevenLabs releases Music v2, saying it was trained only on licensed data
ElevenLabs' new music model adds section-by-section regeneration, and the company says its training data was all licensed.
Models21 May 2026
Stability AI releases Stable Audio 3.0, music models it says are trained only on licensed data
The four-model family includes three open-weight models and an on-device version, with Stability pitching licensed training data as its difference.
Models28 Apr 2026
Anthropic pitches Claude to animation studios and backs the Blender fund
Anthropic added Claude connectors for Blender, Autodesk and Adobe pipelines, became a Blender Development Fund patron and partnered with art schools including RISD.
Models16 Apr 2026
ByteDance opens Seedance 2.0 to business customers in 100+ countries, but not the US
After the Hollywood backlash, ByteDance is selling Seedance 2.0 abroad with face and copyright filters, while keeping it out of the US.
Models15 Apr 2026
Utopai updates its PAI filmmaking model with three-minute 4K video and a continuity agent
The studio's PAI model now renders three-minute 4K clips and tracks continuity across shots, and Utopai says it is in use on Hollywood productions.
Models30 Mar 2026
Suno v5.5 lets paying users sing into their own tracks and tune the model to their catalogue
The update adds voice capture with a spoken check, personal models trained on a user's own songs and a taste profile, as Suno's fights with Universal and Sony continue.
Models27 Mar 2026
Suno 5.5 lets subscribers sing on their own AI songs and train personal models
Suno's March update adds verified voice uploads, personal fine-tuned models and taste learning, which it calls groundwork for label-backed models.
Models25 Mar 2026
Google releases Lyria 3 Pro, a music model for full three-minute songs
Google's new music model writes structured songs up to three minutes, is live across Gemini and its developer tools, and comes with only a vague account of its training data.
Models25 Mar 2026
OpenAI shuts down Sora, and its planned Disney partnership goes with it
The sudden closure of the video app and API voids a still-unsigned deal that would have put Disney characters in Sora.
