Research map / Architectures and efficiency / RNNs and normalization
EfficientTTS: An Efficient and High-Quality Text-to-Speech Architecture
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 |
|---|---|---|---|
| Use EfficientTTS-CNN instead of Tacotron 2 | TTS · DataBaker | MOS: 4.20 ± 0.11 -> 4.41 ± 0.13 | Helped |
| Use EfficientTTS-CNN instead of Glow-TTS | TTS · LJ-Speech | MOS: 4.13 ± 0.18 -> 4.37 ± 0.13 | Helped |
| Use EfficientTTS-CNN instead of Tacotron 2 | TTS · DataBaker | Inference time, text-to-mel: 780 -> 6 ms | Helped |
| Use EfficientTTS-CNN instead of Glow-TTS | TTS · DataBaker | Training time: 120 -> 27 h | Helped |
| Use hard monotonic alignment instead of soft monotonic alignment (ablation) | EFTS-CNN · DataBaker | MSE loss: 0.33 -> 0.095 | Helped |
| Use hard monotonic alignment instead of soft monotonic alignment (ablation) | EFTS-CNN · DataBaker | Training steps to convergence: 450k -> 270k | Helped |
| Remove the monotonicity constraint (ablation) | EFTS-CNN · DataBaker | Convergence | Hurt |
| Add soft monotonic alignment to Tacotron 2 (ablation) | Tacotron 2 · 50-sentence robustness test | Error rate: 50% -> 14% | Helped |
| Add hard monotonic alignment to Tacotron 2 (ablation) | Tacotron 2 · 50-sentence robustness test | Error rate: 50% -> 8% | Helped |
| Replace Tacotron 2 with EFTS-CNN | TTS · 50-sentence robustness test | Error rate: 50% -> 4% | Helped |
| Use EFTS-Wav instead of Tacotron 2 | TTS · DataBaker | Inference time, text-to-wav: 824 -> 16 ms | Helped |
Closest papers by meaning
- DiTTo-TTS: Diffusion Transformers for Scalable Text-to-Speech without Domain-Specific Factors (ICLR 2025 · similarity 0.88)
- DELTA-TTS: Adapting Autoregressive Model into a Diffusion Language Model for Text-to-Speech (NeurIPS 2026 · similarity 0.88)
- MaskGCT: Zero-Shot Text-to-Speech with Masked Generative Codec Transformer (ICLR 2025 · similarity 0.88)
- CLaM-TTS: Improving Neural Codec Language Model for Zero-Shot Text-to-Speech (ICLR 2024 · similarity 0.87)
- Scaling Speech-Text Pre-training with Synthetic Interleaved Data (ICLR 2025 · similarity 0.87)
- EmergentTTS-Eval: Evaluating TTS Models on Complex Prosodic, Expressiveness, and Linguistic Challenges Using Model-as-a-Judge (NeurIPS 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.