https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • h

    hecko

    10/02/2022, 6:13 PM
    oop
  • h

    hecko

    10/02/2022, 6:14 PM
    https://app.uberduck.ai/voice-model/36031100-5842-476e-95c4-a77bfebbb3fc
  • h

    hecko

    10/02/2022, 6:20 PM
    ok nice it worked
  • h

    hecko

    10/02/2022, 6:21 PM
    though it seems to be dependent on sample rate even though it said it would normalize it smhead
  • h

    hecko

    10/02/2022, 6:21 PM
    at 44khz it does this instead
  • l

    Lexi (delulu posts on the daily)

    10/02/2022, 6:22 PM
    wow so magical
  • h

    hecko

    10/02/2022, 6:23 PM
    cruise seals clip finds the right thing at 4th place, not bad
  • h

    hecko

    10/02/2022, 6:23 PM
    Copy code
    py
    # insert imports and definitions and such from above
    from numpy.linalg import norm
    
    test_embedding = classifier.encode_batch(torchaudio.load("untitled.wav")[0]).squeeze().tolist()
    
    with TinyDB(data_path, storage = CachingMiddleware(JSONStorage)) as db:
      voices = [{**i, "distance": norm(numpy.array(i["avg_embedding"]) - numpy.array(test_embedding))} for i in db.all() if type(i["avg_embedding"]) == list]
    
    print("\n".join([f"{i['name']} - {i['distance']}" for i in sorted(voices, key = lambda x: x["distance"])[:10]]))
  • h

    hecko

    10/02/2022, 6:24 PM
    could share the embeddings file but it's 150mb and as evidenced above it used the wrong sample rate so beh
  • h

    hecko

    10/02/2022, 6:25 PM
    if i were brave i'd turn this into a hosted thing
  • h

    hecko

    10/02/2022, 6:25 PM
    Which Uberduck Voice Are You?
  • l

    Lexi (delulu posts on the daily)

    10/02/2022, 6:28 PM
    wait what does it go locate from the bot cmd name or the model name or the
  • l

    Lexi (delulu posts on the daily)

    10/02/2022, 6:29 PM
    i forgot what it's called
  • h

    hecko

    10/02/2022, 6:29 PM
    it has all the metadata stored just in case
  • h

    hecko

    10/02/2022, 6:29 PM
    it's much smaller than the embeddings anyway because there's 5 per voice (1 per sample audio + average) and they're each hundreds floats stored as strings
  • h

    hecko

    10/02/2022, 6:30 PM
    very very crappy prototype
  • h

    hecko

    10/02/2022, 6:30 PM
    but hey it has successfully shown that there's potential
  • u

    {K EY1} (Kei)

    10/04/2022, 2:15 PM
    So i was sorta interested in radtts Has anyone here tested it much? I want to know how well the current pretrained model works with singing datasets and non-american accents Cuz if it doesn't work very well i'll train the talknet pretrained model i'm making on radtts too
  • u

    {K EY1} (Kei)

    10/04/2022, 2:18 PM
    I think you should It would be super interesting for people, i don't think they'd mind if there were to be bugs
  • h

    hecko

    10/04/2022, 3:21 PM
    i mean the main issue is that my vps is running centos 6 because i thought it'd be a good idea
  • u

    {K EY1} (Kei)

    10/04/2022, 4:02 PM
    Ohh
  • z

    zwf

    10/04/2022, 4:11 PM
    does it seem to accurately reflect voices that are perceptually similar?
  • h

    hecko

    10/04/2022, 4:24 PM
    i mean i barely tested it but seems so
  • h

    hecko

    10/04/2022, 4:25 PM
    here all the voices i recognize are feminine
  • h

    hecko

    10/04/2022, 4:26 PM
    y'know what here's the json https://drive.google.com/file/d/1-2Beep8jM3Jyq59sKd1dacxj5DjcL5aK/view
  • h

    hecko

    10/04/2022, 4:26 PM
    note that
    embeddings
    has 4 elements, vs
    avg_embedding
    which is what i actually use for the distance calculation
  • h

    hecko

    10/04/2022, 4:28 PM
    and one of the entries has
    null
    and
    nan
    respectively because the voice for it had no sample audio
  • c

    Couch

    10/04/2022, 4:32 PM
    monkaS also google got rid of that old wacky top bar from like 2012 in this page
  • m

    mega b

    10/05/2022, 2:01 AM
    New vocoder
  • m

    mega b

    10/05/2022, 2:01 AM
    https://google.github.io/df-conformer/wavefit/
1...986987988...1068Latest