home.social

#neuralnetworks — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #neuralnetworks, aggregated by home.social.

  1. Oliver Geer @[email protected] ·

    #askFedi Do you know of speech-to-text live transcription models that:
    1. perform OK for lightly technical English speech, in near-real-time,
    2. have available (ideally libre/open-source) training data, libre/open-source training code, libre/open-source inference code, and/or libre/open-source weights (ideally some amount of freedom to use/study/modify/distribute; e.g. the Open Source AI Definition)

    Glanced at:
    april-asr: probably best candidate so far; somewhat below quality expectations on my laptop (tested with flathub.org/en/apps/net.sapple; quality/performance is configurable; maybe I should run it on a more powerful server and check if it's better?)
    OpenAI Whisper: no training data/code
    Vosk official model: not sure about training data/code, not sure about weights
    Mozilla DeepSpeech: Archived
    CoquiSTT: Looks abandoned

    #Transcription #NeuralNetworks #NN #MachineLearning #ML #SpeechToText #STT #FOSS #FLOSS #OpenSource #FreeSoftware #LibreSoftware #OSAID #OpenSourceAI #CommonVoice

    #askfedi #transcription #neuralnetworks #nn #machinelearning #ml
  2. Oliver Geer @[email protected] ·

    #askFedi Do you know of speech-to-text live transcription models that:
    1. perform OK for lightly technical English speech, in near-real-time,
    2. have available (ideally libre/open-source) training data, libre/open-source training code, libre/open-source inference code, and/or libre/open-source weights (ideally some amount of freedom to use/study/modify/distribute; e.g. the Open Source AI Definition)

    Glanced at:
    april-asr: probably best candidate so far; somewhat below quality expectations on my laptop (tested with flathub.org/en/apps/net.sapple; quality/performance is configurable; maybe I should run it on a more powerful server and check if it's better?)
    OpenAI Whisper: no training data/code
    Vosk official model: not sure about training data/code, not sure about weights
    Mozilla DeepSpeech: Archived
    CoquiSTT: Looks abandoned

    #Transcription #NeuralNetworks #NN #MachineLearning #ML #SpeechToText #STT #FOSS #FLOSS #OpenSource #FreeSoftware #LibreSoftware #OSAID #OpenSourceAI #CommonVoice

    #askfedi #transcription #neuralnetworks #nn #machinelearning #ml
  3. Oliver Geer @[email protected] ·

    #askFedi Do you know of speech-to-text live transcription models that:
    1. perform OK for lightly technical English speech, in near-real-time,
    2. have available (ideally libre/open-source) training data, libre/open-source training code, libre/open-source inference code, and/or libre/open-source weights (ideally some amount of freedom to use/study/modify/distribute; e.g. the Open Source AI Definition)

    Glanced at:
    april-asr: probably best candidate so far; somewhat below quality expectations on my laptop (tested with flathub.org/en/apps/net.sapple; quality/performance is configurable; maybe I should run it on a more powerful server and check if it's better?)
    OpenAI Whisper: no training data/code
    Vosk official model: not sure about training data/code, not sure about weights
    Mozilla DeepSpeech: Archived
    CoquiSTT: Looks abandoned

    #Transcription #NeuralNetworks #NN #MachineLearning #ML #SpeechToText #STT #FOSS #FLOSS #OpenSource #FreeSoftware #LibreSoftware #OSAID #OpenSourceAI #CommonVoice

    #askfedi #transcription #neuralnetworks #nn #machinelearning #ml
  4. Oliver Geer @[email protected] ·

    #askFedi Do you know of speech-to-text live transcription models that:
    1. perform OK for lightly technical English speech, in near-real-time,
    2. have available (ideally libre/open-source) training data, libre/open-source training code, libre/open-source inference code, and/or libre/open-source weights (ideally some amount of freedom to use/study/modify/distribute; e.g. the Open Source AI Definition)

    Glanced at:
    april-asr: probably best candidate so far; somewhat below quality expectations on my laptop (tested with flathub.org/en/apps/net.sapple; quality/performance is configurable; maybe I should run it on a more powerful server and check if it's better?)
    OpenAI Whisper: no training data/code
    Vosk official model: not sure about training data/code, not sure about weights
    Mozilla DeepSpeech: Archived
    CoquiSTT: Looks abandoned

    #Transcription #NeuralNetworks #NN #MachineLearning #ML #SpeechToText #STT #FOSS #FLOSS #OpenSource #FreeSoftware #LibreSoftware #OSAID #OpenSourceAI #CommonVoice

    #commonvoice #opensourceai #osaid #libresoftware #freesoftware #opensource
  5. Oliver Geer @[email protected] ·

    #askFedi Do you know of speech-to-text live transcription models that:
    1. perform OK for lightly technical English speech, in near-real-time,
    2. have available (ideally libre/open-source) training data, libre/open-source training code, libre/open-source inference code, and/or libre/open-source weights (ideally some amount of freedom to use/study/modify/distribute; e.g. the Open Source AI Definition)

    Glanced at:
    april-asr: probably best candidate so far; somewhat below quality expectations on my laptop (tested with flathub.org/en/apps/net.sapple; quality/performance is configurable; maybe I should run it on a more powerful server and check if it's better?)
    OpenAI Whisper: no training data/code
    Vosk official model: not sure about training data/code, not sure about weights
    Mozilla DeepSpeech: Archived
    CoquiSTT: Looks abandoned

    #Transcription #NeuralNetworks #NN #MachineLearning #ML #SpeechToText #STT #FOSS #FLOSS #OpenSource #FreeSoftware #LibreSoftware #OSAID #OpenSourceAI #CommonVoice

    #askfedi #transcription #neuralnetworks #nn #machinelearning #ml