Speech Profiles [Part 2]

[josancamon19]

[josancamon19]

I want to be able to have a free recording of about 30 seconds or 1 minute, where the user can speak freely, and we show the transcript while is happening, and tell them when to stop, once they have certain amount of words/content.

We could put certain phrases for the user to read, so the speech profile is much more accurate.

• Speech profile should be required in onboarding.
• If they have this setup, they shouldn't be able to access this page again from settings once it's done, so it is less confusing.

[josancamon19]

Samples should be a single sample, instead of:

@router.post('/samples/upload')
def upload_sample(file: UploadFile, uid: str):
    print('upload_sample')
    path = f"_temp/{uid}"
    os.makedirs(path, exist_ok=True)
    file_path = f"{path}/{file.filename}"
    with open(file_path, 'wb') as f:
        f.write(file.file.read())
        aseg = AudioSegment.from_wav(file_path)
        print(f'Uploading sample audio {aseg.duration_seconds} secs and {aseg.frame_rate / 1000} khz')
        uploaded_url, count = upload_sample_storage(file_path, uid)
        print('upload_sample ~ file uploaded')
        if count >= 5:
            threading.Thread(target=_create_profile, args=(uid,)).start()
    # os.remove(file_path)
    return {"url": uploaded_url}

Ideally we store that in redis in base64 or something, so we don't use cloud storage anymore, and once read, simply load the bytes in pydub.

[josancamon19]

Done already.