Earn 162,000 ($1,620.00)

Title: Train a German Text-to-Speech Model Using Parler-TTS

Description:

We are looking for an experienced machine learning engineer to train a text-to-speech (TTS) model in German using the Parler-TTS architecture. This model will utilize the German subset of the Multilingual LibriSpeech dataset, which must be annotated using DataSpeech (https://github.com/huggingface/dataspeech) for TTS purposes.

Objectives:

Data Preparation:

Download and prepare the German subset of the Multilingual LibriSpeech dataset.
Annotate the dataset with phonetic and prosodic features using DataSpeech.
Model Training:

Set up and train the Parler-TTS model following the instructions provided in the Parler-TTS training documentation.
Adjust the model and training parameters to be suitable for the German language, including phoneme set and acoustic features.
Testing and Validation:

Test the model to ensure it can generate clear and natural-sounding speech in German.
Conduct both subjective listening tests and objective quality evaluations.
Deliverables:

A trained German TTS model.
Complete documentation of the training process and model configurations.
Instructions on how to use the model to generate speech.
Note: The model is not for public use at this stage.

Requirements:

Experience with machine learning and neural network training, especially in the field of speech synthesis.
Access to necessary computational resources to train a deep learning model.
Ability to work within the specified timeline and deliver all required components.
Please submit a proposal outlining your experience, your approach to the project, and any initial thoughts on the challenges and resources needed