A Privacy-Preserving Multilingual Neural TTS Framework with Automated Artifact Correction and Email-Based Communication

Keerthana A L; Anand Biradar; S. Satheesh Kumar; Vishal M; Vishwanath Rajaput

doi:10.63363/aijfr.2025.v06i06.2514

A Privacy-Preserving Multilingual Neural TTS Framework with Automated Artifact Correction and Email-Based Communication

Author(s)	Ms. Keerthana A L, Mr. Anand Biradar, Mr. S. Satheesh Kumar, Mr. Vishal M, Prof. Vishwanath Rajaput
Country	India
Abstract	Recent advances in neural Text-to-Speech (TTS) have enabled highly naturalistic speech synthesis, yet state-of-the-art models still suffer from artifacts such as mispronunciations, skipped words, and unnatural prosody. These errors often stem from the model's inability to contextualize rare or complex phoneme sequences. This paper presents a novel, robust multilingual synthesis framework that directly addresses this challenge by building upon an automated artifact correction methodology. The core of our system integrates an internal correction algorithm that detects abnormal encoder context vectors by analyzing their deviation from a pre-computed "normal manifold" of training data, allowing for targeted correction without model retraining. We extend this foundation into a practical, end-to-end pipeline with two major contributions: (1) a multilingual synthesis capability that translates English text input into high-fidelity, intelligible speech in English, Tamil, and Hindi, and (2) a secure communication module for sharing the generated audio from one position to another. Our comprehensive evaluation demonstrates the system's effectiveness, achieving a 25.86% reduction in alignment errors, a subjective MOS of 4.6, and a strong comparative CMOS score of +1.34, indicating significant listener preference over the uncorrected baseline. Furthermore, the multilingual outputs achieve over 98% intelligibility, proving the system is a robust, high-quality, and practical solution for real-world communication.
Keywords	Neural Text-to-Speech (TTS), Artifact Correction, Multilingual Synthesis, English, Tamil, Hindi, Encoder Context Vectors, Normal Manifold, Secure Communication, Mean Opinion Score (MOS)
Field	Computer > Data / Information
Published In	Volume 6, Issue 6, November-December 2025
Published On	2025-12-15
DOI	https://doi.org/10.63363/aijfr.2025.v06i06.2514
Short DOI	https://doi.org/hbf935

View / Download PDF File

E-ISSN 3048-7641

doi

CrossRef DOI is assigned to each research paper published in our journal.

AIJFR DOI prefix is
10.63363/aijfr

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About AIJFR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9898-79-39-59	editor@aijfr.com

Advanced International Journal for Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

A Privacy-Preserving Multilingual Neural TTS Framework with Automated Artifact Correction and Email-Based Communication

Share this