ByteDance just launched Seed-Audio 1.0, an all-in-one audio model capable of generating multi-character dialogue, emotional voice acting, sound effects, and background music in a single pass. It uses audio references to keep voices consistent across long podcasts or shows, and is currently rolling out via API.
Rednote just open-sourced dots.tts, a 2B fully continuous end-to-end autoregressive speech system with zero discrete tokens in the pipeline. It achieves top speaker similarity and sub-1% word error rates on Seed-TTS benchmarks, supporting zero-shot cloning and speech editing under an Apache-2.0 license.
Boson AI just released Higgs TTS 3, an open-weights speech model built specifically for real-time conversational agents. Developers can inject inline tags directly into the text stream to tweak emotions, pitch, pacing, and pauses across 100+ languages. It basically turns audio generation into real-time acting direction.
Miso just open-sourced Miso One, an 8B text-to-speech model claimed to be the most emotive voice model in the world. It nails human-like emotional delivery with a lightning-fast 110-millisecond latency—faster than human reaction time. The weights are publicly available now, with an API coming soon for real-time voice apps.
ByteDance just unveiled WavTTS, a zero-shot speech generation model built on Diffusion Transformers and flow matching. Instead of relying on compressed latent representations, it directly models raw audio waveforms to bypass information loss. Why it matters: it clones voices from just a few seconds of audio, proving that end-to-end waveform modeling can rival mainstream latent-space models.
Brilliant just launched Koji, a "graphical tutor" designed to build understanding rather than dependency. Instead of spitting out solutions, Koji monitors your screen, spots where you hesitate, and draws directly on interactive diagrams with guiding questions. It's now live across most math and coding courses.
Baidu just open-sourced NAVA, a 6.3B-parameter model that jointly generates 720p video and stereo audio from a single prompt instead of relying on post-production dubbing. It supports reference-audio voice control and reaches state-of-the-art lip-sync benchmarks with up to 5x fewer parameters than rival open baselines.
ElevenLabs just launched Dubbing v2, a new model built to carry the emotional nuance and performance of original audio into translated languages. Instead of flat voiceovers, dubbed media can now retain the speaker's authentic energy and delivery across every target language.
Spotify just introduced Personal Podcasts, an AI feature letting Premium listeners generate short, custom audio episodes from prompts, PDFs, or links. The audio blends general knowledge with your Spotify listening history for tailored daily briefings or topic explainers. It hits U.S. subscribers next month, while a real-time podcast Q&A tool is already live in three countries.
Alibaba just released Qwen 3.5 Live Translate, a simultaneous interpretation model that factors in live video context to disambiguate words and parse on-screen specs. It covers 60 languages, clones your voice in real time, and cuts speech latency down to 2.8 seconds. A free web demo is live now.
Researchers just dropped Mega ASR, a lightweight speech recognition model built specifically for noisy, distorted real-world audio. Trained on 2.6 million challenging samples covering reverberation, clipping, and severe background interference, it cuts error rates by over 30% against leading baselines. The entire model sits under 5GB, meaning developers can easily run and fine-tune it locally on consumer GPUs.
Meta just unveiled WavFlow, an audio model that generates 44.1kHz sound directly in raw waveform space without using a VAE. It can add synced Foley and music to silent videos, though Meta's policies mean the public repo only includes training scripts rather than the full production checkpoint.
Amazon just rolled out Alexa Podcasts for U.S. Alexa+ subscribers, turning any curiosity into a fully produced audio episode within minutes. Users outline a topic conversationally, adjust length and direction, and let AI host voices deliver stories sourced from over 200 media outlets like Reuters and AP. It basically turns Alexa into a personal, on-demand audio tutor.
Researchers at Zhejiang University just unveiled AudioHijack, a technique that embeds imperceptible malicious commands into everyday audio to hijack generative voice assistants. By slightly tweaking sound waveforms, attacks achieve up to a 96% success rate in forcing models to leak data or run unauthorized tasks, transferring even to commercial services from Microsoft and Mistral.
Microsoft just gave Edge a major AI upgrade on desktop and mobile, letting Copilot reason across all your open tabs to compare details and answer questions. The update adds long-term memory, hands-free Voice and Vision, Journeys topic cards, and a tool to turn tabs into podcasts. Meanwhile, Microsoft is retiring its standalone Copilot Mode.
Resemble AI just launched DramaBox, a directable speech model that treats prompts like movie scripts. Text inside quotes becomes dialogue, while everything outside directs the performance—from venomous whispers to mid-sentence giggles and deep sighs. It supports voice cloning, bakes in deepfake-proof watermarks, and is available for local download if you have 24GB of VRAM.
A new 2-billion-parameter text-to-speech model just dropped, offering zero-shot voice cloning from just a few seconds of reference audio. It captures subtle vocal nuances like whispers and stutters while enabling cross-lingual speech on consumer GPUs under an Apache 2.0 license.
ScenemaAI just open-sourced Scenema Audio, an expressive speech model extracted from the LTX 2.3 audiovisual transformer. It lets you direct emotional shifts and ambient background noise using stage-direction action tags, alongside zero-shot voice cloning. The model fits in 16GB VRAM and supports 13 languages.
Thinking Machines just announced a research preview of native interaction models. Instead of turn-based chatting with external scaffolding, it processes audio, video, and text in continuous 200ms micro-turns paired with an asynchronous background reasoning model. This lets the AI interrupt, listen, speak, and search simultaneously without awkward pauses.
xAI just rolled out Grok Voice mode for Apple CarPlay, letting drivers chat with the AI hands-free right from their dashboard. Enabled by iOS 26.4, the integration follows Apple's strict distraction guidelines by keeping replies audio-only without on-screen text. After living mostly inside Teslas, Grok is now joining ChatGPT and Perplexity across millions of everyday cars.
