scripts/tapi.py

import json, os, requests, warnings, wave
warnings.filterwarnings("ignore")



# synthesise speech
#   save 16khz mono wav file
#   and word-level timestamps
# return paths to wave and alignment files
def tiro(text,voice,save='./'):

    # endpoint working 2023
    url = 'https://tts.tiro.is/v0/speech'
    headers = {'Content-Type': 'application/json'}


    # synthesis
    payload_tts = {
    "Engine": "standard",
    "LanguageCode": "is-IS",
    "OutputFormat": "pcm",
    "SampleRate":"16000",
    "Text": text,
    "VoiceId": voice
    }
    
    # word time alignments
    payload_aln = {
    "Engine": "standard",
    "LanguageCode": "is-IS",
    "OutputFormat": "json",
    "SpeechMarkTypes": ["word"],
    "Text": text,
    "VoiceId": voice
    }

        
    tts_data = requests.post(url, headers=headers, json=payload_tts, verify=False)
    aln_data = requests.post(url, headers=headers, json=payload_aln, verify=False)

    
    #fname = save+text.replace(':','').replace('/','-')
    #wname = fname+'.wav'
    #aname = fname+'.json'
    wname = save+voice+'.wav'
    aname = save+voice+'.json'
    
    with wave.open(wname,'wb') as f:
        f.setnchannels(1)
        f.setframerate(16000)
        f.setsampwidth(2)
        f.writeframes(tts_data.content)

    with open(aname,'w') as f:
        f.write('{"alignments": [')
        f.write(aln_data.content.decode().replace('}\n{','},\n {'))
        f.write(']}')

    return(os.path.abspath(wname),os.path.abspath(aname))



        
#sentence = "Hæ hæ hæ hæ! Ég heiti Gervimaður Finnland, en þú?"
#voice = "Alfur"

#wf, af = tiro(sentence,voice)

#print(wf, af)