Skip to main content
Search

# Text-to-Speech

13 related articles

Subscribe to updates
  1. arXiv

    WaveGen folds zero-shot TTS into one model: text goes straight to waveform, hitting WER 1.47 and SIM 0.75 on Seed-en

    WaveGen pulls zero-shot speech synthesis back from multi-stage pipelines into a single diffusion model: text produces raw waveform directly, without a neural audio codec, Mel vocoder, VAE latent or duration predictor. Spectral Forcing makes the same model learn a low-resolution spectral trajectory alongside the target waveform trajectory, giving text-speech alignment a structured guide in high-resolution waveform space. Trained in one stage on open Emilia data, the 0.6B-parameter WaveGen reaches WER 1.47 and speaker similarity SIM 0.75 on the Seed-en benchmark.

  2. arXiv

    Antalia-2 Mini ships a 7.6M-parameter Turkish voice trained only on a teacher about 300 times larger

    Antalia-2 Mini is a 7,620,931-parameter single-voice Turkish text-to-speech model whose training audio was generated entirely by VoxCPM2, a teacher roughly 300 times larger, cloning one synthetic reference clip, with about 1,790 hours kept after automatic checks. On the 495 sentences of Freya-TR-Eval it reaches 1.09% WER and 0.220% CER, a paired bootstrap cannot separate it from the 8.6M EMA Lightning, the authors tuned settings on those sentences, and no human listening test exists.

  3. arXiv

    RubatoCodec fixes each utterance's event budget so content no longer sets sequence length, and it decodes in a stream

    Variable-frame-rate speech codecs allocate tokens by content, but threshold-based allocation lets content set the token count, so at a fixed setting the sequence length a speech generator must produce varies across utterances and datasets. RubatoCodec adds an exact utterance-level event budget and streaming decoding to X-Codec-2, using changes in codec features and waveform spectra to cut speech into variable-duration events, each represented by discrete codes plus an explicit duration. Controlled experiments show adaptive allocation beats fixed-rate coding in reconstruction fidelity and intelligibility at low event rates; in a matched text-to-speech comparison the representation yields lower WER and higher UTMOS and UTMOSv2 with comparable speaker similarity.

  4. Microsoft releases VibeVoice-ASR-Streaming: transcription emits speaker, time and content per chunk as speech arrives

    Microsoft released VibeVoice-ASR-Streaming, which turns the "who said what" transcription that previously required processing a whole audio segment into continuous output as speech arrives. The earlier VibeVoice-ASR handled 60-minute long-form audio in a single pass, producing structured transcriptions with speaker, timestamps and content within a 64K token length. The new model emits text once per chunk and supports customized hotwords and 10 languages.

  5. Audiobooks stop reading in one voice: Audible fixes a voice profile per character and lifts dialogue differentiation by 3.7 points

    Audible's technology team applies controllable text-to-speech to audiobooks: a large language model first builds a fixed voice profile for each character and writes natural-language stage directions such as "whispering nervously" for each line, and a speech model then synthesizes from them. Character attribution reached 97% and 94% accuracy on 25 English and Spanish mystery, sci-fi and romance books; in a MUSHRA listening test, dialogue differentiation improved by 3.7 points, and eight of ten voices showed statistically significant listening gains across English and Spanish tests.

  6. Mistral Blog

    Voxtral 3 folds listening, reasoning and speaking into one stream, finishing each voice turn in under a second

    Mistral released Voxtral 3, which puts speech recognition, reasoning, tool calling and speech synthesis into a single real-time streaming pipeline, alongside a voice agent SDK. Previously, making an agent listen accurately, reason, speak naturally and act across systems in multiple steps remained difficult. Its Small transcription model beats Whisper, Scribe v2 and Gemini 3.5 Transcribe on multilingual transcription and diarization in English, French and Spanish, and realtime transcription runs below 500 ms latency.

  7. Salesforce's voice model speaks only after it writes: a smaller model aims at enterprise conversation you can inspect

    Salesforce AI Research has built Launa-Omni, a voice model for real-time enterprise conversation that keeps speech and language in one audio-adapted text model and uses a separate in-house text-to-speech model to speak the reply. Unlike pipelines that chain speech recognition, a language model and speech generation, speech is not transcribed before reasoning, so pace, hesitation and emphasis survive the first handoff. The model is significantly smaller than several larger voice models and reports competitive results on spoken instruction following, multi-turn dialogue and customer-service tasks; a simplified version already powers Slack's voice dictation.

  8. ElevenLabs Blog

    UK Parliament hears its first AI-voiced speech: Baroness Campbell delivers the Health Bill debate in her own voice

    For the first time, a member of the UK Parliament addressed a house using an AI-reconstructed voice. Baroness Campbell had relied on an assistant to read her speeches since 2012, as her speaking ability declined with spinal muscular atrophy. She found ElevenLabs' voice cloning herself, uploaded old recordings to rebuild the tone, rhythm and pitch of her voice, then used text-to-speech to generate the speech, played through an assistive device during the Health Bill's Second Reading.

  9. Baseten

    Voice agents stop waiting for the full sentence: Gradium hits 214 ms to first audio and 2–3x faster cold starts

    Gradium TTS pushes time-to-first-audio below typical human speech response, reaching a median TTFA of 214 ms on the Coval benchmark as of September 2026 with an interquartile range tight enough for developers to set reliable latency budgets for the rest of the pipeline. It uses Kyutai's Delayed Streams Modeling, which treats text and audio as two parallel streams on a shared clock so audio can start before the upstream model finishes the sentence. On Baseten it ships as a dedicated deployment whose weights are pre-mirrored and cached, cutting cold starts by 2–3x.

  10. arXiv

    vLLM-Omni unifies speech, diffusion, world-model and robot-policy serving under one orchestrator

    The vLLM-Omni team presents an orchestrator-centered multi-stage serving runtime that splits omni, TTS, diffusion, world-model and VLA workloads into stages bound to individual engines and resource budgets, moves hidden states, codec codes, latents and paged KV through OmniConnector, and adds session control for long-lived interaction; it reports H100/H200 measurements for Qwen3-Omni speech serving, a TTS portfolio, image and video diffusion, and MiniCPM-o duplex health metrics.