Back to AI News
arxiv

WaveGen: End-to-End Text-to-Waveform Diffusion Model with Internal Spectral Trajectory Learning

WaveGen is a fully end-to-end diffusion-based text-to-waveform model that eliminates external components like neural codecs, vocoders, duration predictors, and self-supervised embeddings. It uses a novel Spectral Forcing architecture with disentangled diffusion heads and a multi-scale spectral v-loss to jointly learn spectral and waveform trajectories. The model scales from 0.1B to 0.6B parameters and achieves state-of-the-art results on the Emilia dataset, with a WER of 1.47 and SIM of 0.75 on Seed-en. Source code and checkpoints are publicly available.

AI Brief

This introduces a significant architectural simplification for TTS, achieving strong results without external modules, which could streamline future systems.

arXiv:2611.12527v1 Announce Type: cross Abstract: This work proposes WaveGen, a fully end-to-end diffusion-based text-to-waveform framework based on Spectral Forcing, which jointly learns an internal spectral trajectory and the target waveform trajectory within a single model, enabling direct end-to-end text-to-waveform generation. Unlike recent TTS pipelines, WaveGen does not rely on separately trained or externally decoded intermediate representations, such as neural audio codecs, Mel-spectrogram vocoders, or VAE-based latent models. In addition, WaveGen performs internal text-speech alignment within a single model, eliminating the need for external alignment modules such as duration predictors. To preserve a truly end-to-end formulation, our framework further avoids dependence on self-supervised representations, such as BERT-based text embeddings and Wav2Vec 2.0-based semantic representations, as well as semantic distillation methods for training optimization. To achieve this, we carefully design Spectral Forcing, a model architecture with disentangled diffusion heads, and a training objective based on a multi-scale spectral $v$-loss. We further validate the scalability of end-to-end TTS models by scaling the model size from 0.1B to 0.6B parameters. Using a single architecture, open-source data, and one-stage training, without external modules such as duration predictor or REPA, WaveGen demonstrates fully end-to-end waveform generation achieving state-of-the-art performance on the Emilia training dataset. In particular, WaveGen achieves a WER of 1.47 and a SIM of 0.75 on the Seed-en benchmark. Source code and checkpoints are available at https://github.com/sh-lee-prml/WaveGen
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!