profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/alexpeattie/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Alex Peattie alexpeattie Peg (https://peg.co) London https://alexpeattie.com

alexpeattie/letsencrypt-fromscratch 83

A guide to creating a LetsEncrypt client from scratch in < 150 lines of Ruby

alexpeattie/prideify 45

Inspired by https://facebook.com/celebratepride - a super-lightweight library (< 100 lines of code) to “prideify” (add rainbow stripes to) any image using the <canvas> API.

alexpeattie/animate-textshadow 39

A lightweight jQuery plugin for animating the CSS text-shadow property

alexpeattie/alexpeattie.com 36

The code for alexpeattie.com - a static site built with Eleventy, hosted on Netlify

alexpeattie/xdog-sketch 25

Non-photorealistic, sketch style rendering in the browser with deeplearn.js + React 🎨

alexpeattie/nitlink 17

A nice, modern, nitpicky Ruby gem for parsing HTTP Link headers (per RFC 5988)

alexpeattie/windowed 15

Non-confusing, 'windowed slider' checkboxes and dropdowns with jQuery

alexpeattie/magic 13

Repository to go along with the talk "The magic of dimensionality reduction" given at ODSC London 2017

alexpeattie/vidbot 12

Vidbot connects YouTube push notifications with your Slack team, written in Node.js :robot:

startedtts-tutorial/survey

started time in 17 days

startedvara-tts/VARA-TTS

started time in 20 days

startedmindslab-ai/nuwave

started time in a month

startedchrisgrieser/pandoc_alfred

started time in a month

push eventalexpeattie/alexpeattie.com

Alex Peattie

commit sha deb25d6a63f506f2ad3f12deea58071b0c07abc0

More homepage tweaks

view details

push time in a month

startedhuawei-noah/Speech-Backbones

started time in a month

startedkeonlee9420/Daft-Exprt

started time in a month

startedmuesli/readme-scribe

started time in 2 months

startedKevinMIN95/StyleSpeech

started time in 2 months

startedDarshanDeshpande/audio-spectral-enhancement

started time in 2 months

startedAakashKumarNain/TF_JAX_tutorials

started time in 2 months

startedpatrick-kidger/equinox

started time in 2 months

startedmindslab-ai/univnet

started time in 2 months

startedkeonlee9420/Comprehensive-Tacotron2

started time in 2 months

startedyl4579/StarGANv2-VC

started time in 2 months

startedmindslab-ai/cotatron

started time in 2 months

startedWendison/VQMIVC

started time in 2 months

startedmindslab-ai/assem-vc

started time in 2 months

startedWendison/FCL-taco2

started time in 2 months

startedauspicious3000/AutoPST

started time in 2 months

startedkeonlee9420/VAENAR-TTS

started time in 2 months

startedWelkinYang/GradTTS

started time in 2 months

startedkeonlee9420/FastPitchFormant

started time in 2 months

startedthuhcsi/VAENAR-TTS

started time in 2 months

startedkeonlee9420/Expressive-FastSpeech2

started time in 2 months

startedrevsic/tf-mlptts

started time in 2 months

startedkeonlee9420/WaveGrad2

started time in 2 months

startedmindslab-ai/wavegrad2

started time in 2 months

issue commentjaywalnut310/vits

Anybody having luck fine tuning?

@alexpeattie Did you get an improvement from g/fm loss?

Hey @CookiePPP, sorry, I'm not 100% sure what you're asking? Do you mean did the g/fm loss improve as I continued training?

TaoTeCha

comment created time in 3 months

issue commentjaywalnut310/vits

Anybody having luck fine tuning?

Here you go:

Capture d’écran 2021-06-28 à 08 35 20

Capture d’écran 2021-06-28 à 08 35 57

It seems that the discriminator loss is improving steadily, while the mel loss is improving on the generator while other losses are climbing. However, generated samples seem to be getter slightly better.

The synthesized sound very good in terms of spectrogram inversion from only a few 10k steps (e.g. no tininess/reverb/warbling). And the prosody seems to be getting better as the training goes on. But for me by 180k it was already sounding better than Tacotron, so I suspect something might have gone wrong with your training.

TaoTeCha

comment created time in 3 months