FastSpeech 2: Fast and High-Quality End-to-End Text to Speech
read the original abstract
Non-autoregressive text to speech (TTS) models such as FastSpeech can synthesize speech significantly faster than previous autoregressive models with comparable quality. The training of FastSpeech model relies on an autoregressive teacher model for duration prediction (to provide more information as input) and knowledge distillation (to simplify the data distribution in output), which can ease the one-to-many mapping problem (i.e., multiple speech variations correspond to the same text) in TTS. However, FastSpeech has several disadvantages: 1) the teacher-student distillation pipeline is complicated and time-consuming, 2) the duration extracted from the teacher model is not accurate enough, and the target mel-spectrograms distilled from teacher model suffer from information loss due to data simplification, both of which limit the voice quality. In this paper, we propose FastSpeech 2, which addresses the issues in FastSpeech and better solves the one-to-many mapping problem in TTS by 1) directly training the model with ground-truth target instead of the simplified output from teacher, and 2) introducing more variation information of speech (e.g., pitch, energy and more accurate duration) as conditional inputs. Specifically, we extract duration, pitch and energy from speech waveform and directly take them as conditional inputs in training and use predicted values in inference. We further design FastSpeech 2s, which is the first attempt to directly generate speech waveform from text in parallel, enjoying the benefit of fully end-to-end inference. Experimental results show that 1) FastSpeech 2 achieves a 3x training speed-up over FastSpeech, and FastSpeech 2s enjoys even faster inference speed; 2) FastSpeech 2 and 2s outperform FastSpeech in voice quality, and FastSpeech 2 can even surpass autoregressive models. Audio samples are available at https://speechresearch.github.io/fastspeech2/.
This paper has not been read by Pith yet.
Forward citations
Cited by 18 Pith papers
-
FlexiSLM: A Dynamic and Controllable Frame Rate Spoken Language Model
FlexiSLM is the first spoken language model supporting dynamic and controllable frame rates on speech input and output, outperforming fixed-rate 7B models at high quality and enabling faster inference at lower rates l...
-
N\"ushuVoice: Reviving the Voice of Endangered N\"ushu with Pitch-Aware Text-to-Speech
NüshuVoice releases the first sentence-level Nüshu TTS dataset and shows that an F0-conditioned VITS model using five-level pitch notation outperforms baselines on spectral fidelity, pitch accuracy, and intelligibility.
-
HapticLDM: A Diffusion Model for Text-to-Vibrotactile Generation
HapticLDM is the first latent diffusion model that generates vibrotactile signals directly from text, using dynamic text curation and global denoising to improve realism and semantic alignment over autoregressive baselines.
-
MelShield: Robust Mel-Domain Audio Watermarking for Provenance Attribution of AI Generated Synthesized Speech
MelShield adds keyed low-energy spread-spectrum perturbations to Mel-spectrograms inside TTS pipelines before vocoding to enable robust extraction of user-specific attribution signals even after compression or noise.
-
AST: Adaptive, Seamless, and Training-Free Precise Speech Editing
AST enables seamless speech editing by latent recomposition on pre-trained TTS models plus adaptive weak fact guidance, plus a new dataset and WDTW metric, claiming 70% WER reduction and better temporal consistency wi...
-
MIDI-Informed Singing Accompaniment Generation in a Compositional Song Pipeline
MIDI-SAG generates consistent long-form singing accompaniments by feeding symbolic MIDI timing, chords, and structure labels into a compositional pipeline built from pre-trained modules.
-
Learning to Evade: Adaptive Attacks on Audio Watermarking
Introduces AWM adaptive attack using two-stage optimization and distribution estimation to bypass audio watermark detectors with low detection rates on voice datasets.
-
DisSpeech: Low-Resource Controllable Mandarin Stuttered Speech Synthesis for ASR Augmentation
DisSpeech synthesizes controllable stuttered Mandarin speech via discrete tokens and stuttering event labels to augment ASR datasets, improving recognition to 4.19% CER on stuttered tasks with minimal impact on fluent speech.
-
FineCombo-TTS: Collaborative and Precise Controllable Speech Synthesis Using Text Descriptions and Reference Speech
FineCombo-TTS learns a unified acoustic representation with a CFM-based Speech Variance Predictor for flexible precise TTS control from reference audio and text descriptions, supported by the new FineEdit paired dataset.
-
OLaPh: Optimal Language Phonemizer
Hybrid OLaPh framework outperforms prior G2P baselines on WikiPron while enabling synthetic data for an LLM that generalizes well on out-of-vocabulary terms.
-
Step-Audio 2 Technical Report
Step-Audio 2 integrates a latent audio encoder, reasoning-centric reinforcement learning, and discrete audio token generation into language modeling to deliver state-of-the-art performance on audio understanding and c...
-
JAM-Flow: Joint Audio-Motion Synthesis with Flow Matching
JAM-Flow introduces a unified flow-matching model with a Multi-Modal Diffusion Transformer that jointly synthesizes facial motion and speech from text, audio, or motion inputs.
-
Perceptual implications of automatic anonymization in pathological speech
Listeners detect automatic anonymization in pathological speech at 91-93% accuracy with a 30-point perceived quality drop, yet clinical severity ratings stay nearly unchanged for dysarthria, dysglossia, and dysphonia.
-
Seed-TTS: A Family of High-Quality Versatile Speech Generation Models
Seed-TTS models produce speech matching human naturalness and speaker similarity, with added controllability via self-distillation and reinforcement learning.
-
F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching
F5-TTS generates natural speech from text via flow matching on DiT with simple text padding, ConvNeXt refinement, and sway sampling, trained on 100K hours multilingual data.
-
MeloDISinger: Melody-Aware & Duration-Preserving Singing Voice Editing with Audio Infilling
Proposes MeloDISinger, a flow-matching SVE model with MeloDRP for melody-aware duration-preserving editing and audio infilling, claiming SOTA results.
-
Voice Mapping of Text-to-Speech Systems: A Metric-Based Approach for Voice Quality Assessment
Voice range indicates TTS model capability with VITS highest, Glow-TTS best at soft phonation, and CPPs of 7-8 dB marking natural quality while values over 10 dB sound robotic.
-
Intelligent Agents with Emotional Intelligence: Current Trends, Challenges, and Future Prospects
A holistic survey of affective computing for intelligent agents covering emotion understanding via multimodal data, affective cognition, emotional expression synthesis, key challenges, and future directions emphasizin...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.