Posts

• Liepa to fairseq

  Convert the liepa2 corpus to fairseq

  Jan 13, 2023

• Extract audio from .pcm from NST Swedish Speech Synthesis

  Raw, 8? 16? byte header, big endian PCM 44.1k

  Jan 3, 2023

• Playing with SimpleLanguageModel from pynlpl

  I wanted to know if it could be used to write an ARPA LM. It cannot

  Dec 6, 2022

• Convert WebVTT to Elan

  For Whisper's output

  Nov 11, 2022

• Simple replacement for Kaldi's align-text

  Because life's too short to install Kaldi again

  Nov 10, 2022

• Phonetic transcription with HuggingFace

  wav2vec2 espeak phonetic model

  Oct 18, 2022

• Using IrishNLP's chunker with NLTK

  Convert chunks to a tree

  Oct 3, 2022

• Adapt `cmu_us_awb_arctic` to fairseq

  Writing the tsv/ltr files; from Kaggle

  May 7, 2022

• LJSpeech for ASR

  Resampled wav, more normalised text. From Kaggle

  May 4, 2022

• Split Liepa2

  Notebook to split the audio from the liepa2 corpus

  May 4, 2022

• Convert Waxholm to wav

  From Kaggle

  May 4, 2022

• Fix LJSpeech text

  LJSpeech comes with a normalised version, but it needs some extra work

  May 2, 2022

• Hyphenisation with pyphen

  Snippet to get Hunspell-based hyphenisation from pyphen

  Apr 26, 2022

• Prepping TIMIT for Fairseq (for PSST)

  For use with PSST

  Apr 4, 2022

• Getting timestamps on long audio

  With wav2vec2 and HuggingFace transformers

  Mar 8, 2022