Text-to-Speech & Speech-to-Text
Text-to-speech and Speech-to-text Machine Learning platforms for English and Vietnamese.
Project Overview
The CodeLink text-to-speech converts speech to text and back smoothly in both English and Vietnamese with a natural output available in a range of voices and tones.
CLIENT
CodeLink
Team Model
Front-end Developer
Back-end Developer
DevOps
AI & ML Lead
Quality Assurance Engineer
PLATFORM
Web
TECHNOLOGY
NextJS
NestJS
Flutter
Python
AI/ML
RabbitMQ
Kubernetes
AI (Pytorch)
Tensorflow
TensorflowJS
ReactJS
Google Cloud Platform (GCP)
Firebase Crashlytics
Redis
Challenge
How can we leverage Machine Learning to create a fluid and natural text-to-speech and speech-to-text model for Vietnamese?
Request
CodeLink tasked the internal product team with creating a model and standards for text-to-speech and speech-to-text between English and Vietnamese.
Engagement Model
We worked as an autonomous team to create the text-to-speech and speech-to-text model for Vietnamese.
Engagement Length and Scale
CodeLink internal teams worked for 4 months to develop the proprietary text-to-speech and speech-to-text AI model.
Project Outcome
To find the best approach, we tested several TTS and STT models. Because there were no quality Vietnamese datasets available in the market, we decided to start with text-to-speech technology in Vietnamese. We created an annotation app using ReactJS, Firebase, and GCP for the annotation process and coordinated with local voice talent and text labelers to create the data. To optimize the core back-end, we built a Kubernetes cluster for inference services. After training the model, we deployed the system.
