Research map / Image generation / Autoregressive and latent generation
FastSpeech: Fast, Robust and Controllable Text to Speech
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| generate mel-spectrograms non-autoregressively with length regulator | Transformer TTS · LJSpeech | mel generation speedup: 269.40x | Helped |
| generate mel-spectrograms non-autoregressively | Transformer TTS · LJSpeech test | MOS: 3.88 -> 3.84 | No clear effect |
| use hard duration-based alignment instead of attention | Transformer TTS · 50 hard sentences | error rate: 34% -> 0% | Helped |
| use 1D convolution instead of FC layer in FFT block (ablation) | FastSpeech · LJSpeech | CMOS: -0.113 | Helped |
| add sequence-level knowledge distillation (ablation) | FastSpeech · LJSpeech | CMOS: -0.325 | Helped |
Closest papers by meaning
- FastSpeech 2: Fast and High-Quality End-to-End Text to Speech (ICLR 2021 · similarity 0.94)
- EfficientTTS: An Efficient and High-Quality Text-to-Speech Architecture (ICML 2021 · similarity 0.88)
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (ICLR 2024 · similarity 0.88)
- FastLongSpeech: Enhancing Large Speech-Language Models for Efficient Long-Speech Processing (NeurIPS 2025 · similarity 0.87)
- MaskGCT: Zero-Shot Text-to-Speech with Masked Generative Codec Transformer (ICLR 2025 · similarity 0.87)
- Long-Form Speech Generation with Spoken Language Models (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.