Accepted at Interspeech 2026

Stabilizing Instruction Supervision for Instruct-TTS via
Controllable Diversification and Drift Filtering

Yizhong Geng, Kecan Mao, Qifei Li, Cong Wang, Yingming Gao, Ruimin Wang, Chunfeng Wang, Hao Li, and Ya Li

Beijing University of Posts and Telecommunications & Li Auto

40.4% -> 15.4%semantic drift after constrained rewriting
56.4%average instruction-following accuracy
4.16 / 4.16human NMOS / CMOS for the full recipe

Paper Summary

Instruct-TTS systems commonly expand structured style labels into natural-language training instructions through LLM rewriting, yet we find that over 40% of unconstrained rewrites contain semantic drift that corrupts supervision signals and weakens generalization. We formalize this problem as instruction supervision instability and propose a data-centric stabilization recipe that jointly improves coverage and fidelity through three complementary mechanisms: controllable instruction diversification for systematic linguistic expansion, LLM-based drift filtering for semantic quality assurance, and attribute-aligned supervision that grounds prosody control in parameterized acoustic perturbations. On the Chinese split of InstructTTSEval, our recipe raises average instruction-following accuracy from 34.5% without fine-tuning and 51.0% with naive fine-tuning to 56.4%, while constrained rewriting reduces drift from 40.4% to 15.4%. Ablations confirm all three mechanisms are complementary, and the drift taxonomy may extend to instruction-driven generation beyond TTS.

Key Contribution: A data-centric stabilization recipe with controllable diversification, drift filtering, and attribute-aligned supervision — raising instruction-following accuracy to 56.4% and reducing semantic drift from 40.4% to 15.4%.
VenueInterspeech 2026 accepted paper
FocusReliable instruction supervision for controllable TTS
DemoChinese InstructTTSEval samples across APS, DSD, and RP

Method Overview

Method Overview

Figure 1: Overview of the proposed stabilization recipe for Instruct-TTS instruction supervision.

Citation

@inproceedings{geng2026stabilizing,
  title = {Stabilizing Instruction Supervision for Instruct-TTS via Controllable Diversification and Drift Filtering},
  author = {Geng, Yizhong and Mao, Kecan and Li, Qifei and Wang, Cong and Gao, Yingming and Wang, Ruimin and Wang, Chunfeng and Li, Hao and Li, Ya},
  booktitle = {Proceedings of Interspeech 2026},
  year = {2026}
}