Transcription et synthèse vocale

# Le message vocal de l'exercice, fabriqué : la voix de synthèse du Mac lit message.txt.
say -v Thomas -f message.txt -o message.aiff
afconvert -f m4af -d aac message.aiff message.m4a     # un format que les API acceptent
afinfo message.m4a | grep -E "estimated duration"
ls -l message.m4a | awk '{print "taille :", $5, "octets"}'

from openai import OpenAI

client = OpenAI()
audio_file = open("message.m4a", "rb")

transcription = client.audio.transcriptions.create(
    model="gpt-transcribe", file=audio_file
)

print(transcription.text)

import os
from mistralai.client import Mistral

api_key = os.environ["MISTRAL_API_KEY"]
model = "voxtral-mini-latest"

client = Mistral(api_key=api_key)

with open("message.m4a", "rb") as f:
    transcription_response = client.audio.transcriptions.complete(
        model=model,
        file={
            "content": f,
            "file_name": "message.m4a",
        },
        ## language="en"
    )

import re, unicodedata

NOMBRES = {"quatre mille huit cent dix-sept": "4817", "trois": "3", "douze": "12"}

def mots(texte):
    """Minuscules, sans ponctuation, nombres écrits en chiffres : on compare les mots, pas la typographie."""
    t = texte.lower()
    for lettres, chiffres in NOMBRES.items():
        t = t.replace(lettres, chiffres)
    t = unicodedata.normalize("NFC", t)
    return re.findall(r"[\w'-]+", t)

def wer(reference, hypothese):
    """Taux d'erreur sur les mots : (substitutions + suppressions + insertions) / mots de la référence."""
    r, h = mots(reference), mots(hypothese)
    d = [[i + j if i * j == 0 else 0 for j in range(len(h) + 1)] for i in range(len(r) + 1)]
    for i in range(1, len(r) + 1):
        for j in range(1, len(h) + 1):
            d[i][j] = min(d[i - 1][j] + 1, d[i][j - 1] + 1, d[i - 1][j - 1] + (r[i - 1] != h[j - 1]))
    return d[len(r)][len(h)], len(r)

erreurs, total = wer(open("message.txt").read(), open("transcription.txt").read())
print(f"{erreurs} erreurs sur {total} mots : WER = {erreurs / total:.1%}")
ref, hyp = mots(open("message.txt").read()), mots(open("transcription.txt").read())
print("mots différents :", [(a, b) for a, b in zip(ref, hyp) if a != b])

from pathlib import Path
from openai import OpenAI

client = OpenAI()
speech_file_path = Path(__file__).parent / "speech.mp3"

with client.audio.speech.with_streaming_response.create(
    model="gpt-4o-mini-tts",
    voice="coral",
    input="Today is a wonderful day to build something people love!",
    instructions="Speak in a cheerful and positive tone.",
) as response:
    response.stream_to_file(speech_file_path)