Uncloudy voice assistant: Difference between revisions

From titipi
Jump to navigation Jump to search
No edit summary
 
(5 intermediate revisions by the same user not shown)
Line 1: Line 1:
==== How to access automatic subtitling in an interview, conferencing, gathering, conversation in a counter cloud way? ====
== How to access automatic subtitling in an interview, conferencing, gathering, conversation in a counter cloud way? ==
For example in the case of a counter cloud video conferencing like BBB and what happens when somebody complains about the non-seamlessness.  
For example in the case of a counter cloud video conferencing like BBB and what happens when somebody complains about the non-seamlessness.  


Line 5: Line 5:
Some thoughts on machine listening:
Some thoughts on machine listening:


The process of machine listening has been supported and developed technically through legal, state and military cases. Older threads:
The process of machine listening has been supported and developed technically through legal, state and military cases and developments. Clouds and voice databases are also rooted in this.
 
 
We can unpack the processes of speech recognition in interviewing, transcribing, listening, training, listening to patterns, predicting and follow hybrid ways of automated subtitling, that include computing and not.
 
=== Some algorithms ===
 
==== Speech recognition tools ====
 
* [https://alphacephei.com/vosk/install Vosk]
* [https://otter.ai/ otter.ai]
* [https://gitlab.com/nglk/speaking_with_the_machine/-/blob/master/ttssr2_transcribe.py Speech_recognition with Pocketsphinx]
* [https://pypi.org/project/pocketsphinx/ Livespeech with Pocketsphinx]
from pocketsphinx import LiveSpeech
for phrase in LiveSpeech():
    # how to pause the mic occasionally so the live speech print appears often
    # check timer thread
    print(phrase)
* [https://github.com/lowerquality/gentle/pulse Gentle]
 
==== Poetic experiments ====
* [https://gitlab.com/nglk/gossip-gaps/-/blob/master/gossip-gaps.py Gossip Booth]: recognise words and keep only the ones with no meaning, like vocal expressions.
* Transcribe only phonemes with coqui STT and CMU Sphinx
* [https://gitlab.com/nglk/radioactive Radioactive Monstrosities]: voice upload and vocal transformation on a browser with the use of Web Audio API
* [https://github.com/jreus/chorusworkshop Remnants of future voices]: cloning voice scripts with an interface
* [https://gitlab.com/nglk/para-dictionary-wordmord WordMord Para-dictionary]: replace words of a text/transcription with a collective made dictionary
* [https://github.com/aparrish/nonsense-verse-pycon-2020 Pincelate]: Nonsense verse (and other creative uses of spelling) with Python and machine learning by Allison Parrish
 
 
 
 
 
Older threads:


* https://pzwiki.wdka.nl/mediadesign/User:Angeliki/Ttssr-Speech_Recognition_Iterations
* https://pzwiki.wdka.nl/mediadesign/User:Angeliki/Ttssr-Speech_Recognition_Iterations
Line 12: Line 44:
* scripts with WordMord
* scripts with WordMord
* workshop and references with Jon
* workshop and references with Jon
* Radioactive
* Radioactive  





Latest revision as of 14:18, 21 March 2023

How to access automatic subtitling in an interview, conferencing, gathering, conversation in a counter cloud way?

For example in the case of a counter cloud video conferencing like BBB and what happens when somebody complains about the non-seamlessness.


Some thoughts on machine listening:

The process of machine listening has been supported and developed technically through legal, state and military cases and developments. Clouds and voice databases are also rooted in this.


We can unpack the processes of speech recognition in interviewing, transcribing, listening, training, listening to patterns, predicting and follow hybrid ways of automated subtitling, that include computing and not.

Some algorithms

Speech recognition tools

from pocketsphinx import LiveSpeech
for phrase in LiveSpeech(): 
   # how to pause the mic occasionally so the live speech print appears often
   # check timer thread
   print(phrase)

Poetic experiments

  • Gossip Booth: recognise words and keep only the ones with no meaning, like vocal expressions.
  • Transcribe only phonemes with coqui STT and CMU Sphinx
  • Radioactive Monstrosities: voice upload and vocal transformation on a browser with the use of Web Audio API
  • Remnants of future voices: cloning voice scripts with an interface
  • WordMord Para-dictionary: replace words of a text/transcription with a collective made dictionary
  • Pincelate: Nonsense verse (and other creative uses of spelling) with Python and machine learning by Allison Parrish



Older threads:


references: