even less aggressive silence filtering and more data
Tensorboard data can be found here. This version used even less aggressive silence filtering (set to aggressiveness value to 0 this time) as well as data fro...
Posts by Tag
text-to-speech
less aggressive silence filtering
I modified the aggressiveness from 3 to 1. Training history can be found here for a run which used the LJSpeech mode I trained, and here for a model trained ...
webrtcvad and Shorter Sections
Cheryl and I spent some time taking a bit of a break from training models to try and figure out what it is. One big thing could be an issue with long sentenc...
Blizzard
I decided to try yet another dataset, this time the Blizzard dataset. The tensorboard results can be seen here. With this dataset, it is again successful at ...
Transfer Learning and Hand-Labelling
I suspect the most significant issue is the forced alignment being slightly off on two accounts: It might sometimes just be plain wrong It is common for...
2021-03-06
I tried to retrain again with the Karen Savage data after changing each line to be separated by each sentence. The goal here is to allow more empty space bet...
LJSpeech
To narrow down the problem further, I trained another model using the LJSpeech dataset but my own text preprocessing pipeline. Very annoyingly enough, the mo...
New Dataset
I investigated the previous preprocessing pipeline and there didn’t seem to be any issues. However, just in case, I decided to switch to using the mel spectr...
First Training Attempt
Preprocessing changes
Model Selection
The original plan was to train first with flowtron since it offers multi-voice training. However, the data for that would be more difficult to gather as it r...
Data Verification
I wanted to check the accuracy of aeneas, but definitely did not want to be manually looking at the json results and checking each word and its timestamp to ...
Data Cleaning
The main dataset we’ll use to start off with is The Adventures of Tom Sawyer read by John Greeman.
audio-samples
even less aggressive silence filtering and more data
Tensorboard data can be found here. This version used even less aggressive silence filtering (set to aggressiveness value to 0 this time) as well as data fro...
less aggressive silence filtering
I modified the aggressiveness from 3 to 1. Training history can be found here for a run which used the LJSpeech mode I trained, and here for a model trained ...
LJSpeech
To narrow down the problem further, I trained another model using the LJSpeech dataset but my own text preprocessing pipeline. Very annoyingly enough, the mo...
mechanical
Camera Issues
I got around to testing our webcam, and the results were dissapointing to say the least.
Camera stand design review
The ongoing pandemic situation has led to a reconsideration of the previous camera stand design from fall 2020. At this point we are almost certainly abandon...
OCR
Camera Issues
I got around to testing our webcam, and the results were dissapointing to say the least.