AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

DMOSpeech 2: Reinforcement Learning for Duration Prediction in Metric-Optimized Speech Synthesis

Yinghao Aaron Li, Xilin Jiang, Fei Tao, Cheng Niu, Kaifeng Xu, Juntong Song, Nima Mesgarani

AAAI 2026 · Image generation › Autoregressive and latent generation · approach: Neural text-to-speech synthesis · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
add GRPO reinforcement learning to the duration predictor (SIM and WER rewards) (ablation)DMOSpeech 2 student (4 steps) · Seed-TTS test-enWER: 3.750 -> 1.752Improved
add GRPO reinforcement learning to the duration predictor (ablation)DMOSpeech 2 student (4 steps) · Seed-TTS test-enSIM-o: 0.672 -> 0.698Improved
add GRPO reinforcement learning to the duration predictor (ablation)DMOSpeech 2 student (4 steps) · Seed-TTS test-zhCER: 2.000 -> 1.527Improved
add GRPO duration RL (subjective) (ablation)DMOSpeech 2 student · Seed-TTS test-en CMOSCMOS-N: -0.43Improved
replace 32-step F5-TTS teacher with 4-step distilled DMOSpeech 2F5-TTS · Seed-TTS RTFRTF: 0.1671 -> 0.0316Improved
add teacher-guided sampling (14 teacher steps + 2 student steps) (ablation)DMOSpeech 2 student · Pitch diversity CVf0CVf0: 0.4640 -> 0.5932Improved
add teacher-guided sampling (ablation)DMOSpeech 2 student · Seed-TTS test-enWER: 1.752 -> 1.738Improved
add teacher-guided sampling (ablation)DMOSpeech 2 student · Seed-TTS RTFRTF: 0.0316 -> 0.0941Hurt
add teacher-guided sampling (ablation)DMOSpeech 2 student · Seed-TTS test-zhCMOS-N: +0.45 (CMOS-N), +0.3 (CMOS-S)Mixed
distill teacher into few-step student with distribution matchingF5-TTS teacher · Pitch diversity CVf0CVf0: 0.6659 -> 0.4640Hurt
use DMOSpeech 2 instead of 8B LLM-based TTSLLaSA-8B comparison · Seed-TTS test-enWER: 3.994 -> 1.752Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy