Project / Overview
Small-model TTS
Fine-tuning a small text-to-speech model and exploring expressive speech.
Published September 13, 2026
I’ve fine-tuned a small text-to-speech model. This project is one starting point for my research into expressive speech and emotionally intelligent AI.
This page is a project overview. Training configuration, checkpoints, audio samples, and measured results are not published here yet.
From implementation to research
Building a small model raises questions I want to examine more carefully: what can be controlled, how that control should be represented, and what evidence would show that it works.
My generative TTS review outline sets out the reading that will help frame the next experiment.
Planned artifacts
- A technical note describing the model and fine-tuning setup.
- A record of the data preparation and training configuration.
- Audio samples with transcripts and labeled conditions.
- An evaluation protocol and reproducible experiment report.
Each artifact will be added when it is ready to share.