Abstract
Current automated text-to-speech systems for long-form audio production suffer from inconsistent quality, limited emotional range, and inability to self-correct output deficiencies. Professional audiobook and radio drama production requires human narrators specifically because automated systems cannot achieve the performance quality, character consistency, and emotional nuance necessary for extended listening experiences. This creates a significant cost barrier for independent authors and small publishers, where professional narration costs £200-500 per finished hour.
This disclosure presents an automated multi-engine text-to-speech system that achieves professional broadcast quality through coordinated engine selection, performance enhancement techniques, and iterative quality refinement. The system employs two specialized synthesis engines: an actor-quality engine for narrative and dialogue, and an effects-capable engine for sound effects, music cues, and comedic timing. Intelligent routing algorithms analyze manuscript content to assign appropriate engines for each segment, maintaining character voice consistency across hours of narration.
The system implements self-assessment quality control where generated audio is analyzed against professional performance standards. Segments failing quality thresholds trigger automatic regeneration with adjusted parameters until acceptable output is achieved. An ethical voice cloning protocol prevents identity theft by requiring actors to record system-generated unique phrases rather than arbitrary text. Field testing demonstrates BBC Radio Drama-equivalent quality while reducing production costs by 90% compared to human narration.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Chiesa, David R P, "Automated Multi-Engine Text-to-Speech System with Self-Assessment Quality Control for Long-Form Audio Production", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9056