app.py

# flake8: noqa
import random
import uuid
from dataclasses import dataclass
from pathlib import Path

import requests
import streamlit as st

from api import EMOTTS_API_ROUTE
from inference_pipeline import (
    CleanedTextIsEmptyStringError, SpeakerNotFoundError,
    inference_text_to_speech,
)
from src.constants import Emotion, Language, SupportedLanguages
from src.web.streamlit_utils import (
    hide_hamburger_menu, st_empty_block, st_header_centered,
)

BROKEN_SPEAKERS = ["0019"]
SELECTOR_TO_LANG = {
    "🇷🇺 Russian (ru-RU)": SupportedLanguages.russian,
    "🇺🇸 English (en-EN)": SupportedLanguages.english,
}

LOADING_PHRASES = [
    "👷‍♂️ Building language corpuses...",
    "😋 Dreaming about free GPU hosting...",
    "💫 Permuting tensors...",
]


@dataclass
class AppModes:
    api_connector: str = "api"
    standalone: str = "standalone"


APP_MODE = AppModes.api_connector
HOST = "localhost"
PORT = 8080
# EMOTTS_API_ENDPOINT = f"http://{HOST}:{PORT}{EMOTTS_API_ROUTE}"
EMOTTS_API_ENDPOINT = "https://api-emotts.appspot.com/tts/emo/v1"  #GCP


def run_in_api_connector_mode(language: Language, emotion: Emotion, input_text: str, speaker: str, audio_output_path: Path) -> None:
    loading_phrase = random.choice(LOADING_PHRASES)
    with st.spinner(loading_phrase):
        query = f"{EMOTTS_API_ENDPOINT}?lang={language.api_name}&emo={emotion.api_name}&speaker={speaker}&text={input_text}"
        r = requests.get(query, stream=True)
    if r.status_code == 200:
        with open(audio_output_path, 'wb') as f:
            for chunk in r:
                f.write(chunk)
    else:
        st.warning(f"Status: {r.status_code}")
        st.warning(f"Message: {r.text}")


def run_in_standalone_mode(language: Language, emotion: Emotion, input_text: str, speaker: str, audio_output_path: Path) -> None:
    loading_phrase = random.choice(LOADING_PHRASES)
    with st.spinner(loading_phrase):
        inference_text_to_speech(
            language=language,
            input_text=input_text,
            speaker=speaker,
            emotion=emotion,
            audio_output_path=audio_output_path,
        )


def layout_app() -> None:

    # header_text = "🥀 EMOtts dEMO 💔"
    header_text = "EmoTTS Project"
    st_header_centered(header_text)
    st_empty_block(2)

    language_name = st.selectbox(label="👅 Language", options=SELECTOR_TO_LANG.keys())
    language = SELECTOR_TO_LANG.get(language_name)

    with st.form(key="input_form"):
        col1, col2 = st.columns(2)
        with col1:
            speaker_selector = language.speaker_selector
            for broken_speaker in BROKEN_SPEAKERS:
                speaker_selector.pop(broken_speaker, None)
            speaker = st.selectbox(label="🗣 Speaker", options=speaker_selector.keys())
        with col2:
            emotion_name = st.select_slider(label="🎨 Emotion", options=language.emo_selector.keys())
            emotion = language.emo_selector.get(emotion_name)
        st_empty_block(2)
        input_text = st.text_area(label="📜 What should I say?", value=language.test_phrase, max_chars=140)
        st_empty_block(2)
        form_submit = st.form_submit_button("Synthesize speech")

    if form_submit:
        audio_output_path = Path(f"generated-{uuid.uuid4()}.wav")
        try:
            # Run inference pipeline
            if APP_MODE == AppModes.standalone:
                run_in_standalone_mode(
                    language=language,
                    input_text=input_text,
                    speaker=speaker,
                    emotion=emotion,
                    audio_output_path=audio_output_path,
                )
            elif APP_MODE == AppModes.api_connector:
                run_in_api_connector_mode(
                    language=language,
                    input_text=input_text,
                    speaker=speaker,
                    emotion=emotion,
                    audio_output_path=audio_output_path,
                )
            with open(audio_output_path, "rb") as audio_file:
                st.audio(audio_file.read())
            audio_output_path.unlink()
        except CleanedTextIsEmptyStringError:
            st.warning("😔 Looks like input text can not be pronounced")
            st.stop()
        except SpeakerNotFoundError:
            st.warning("🔍 This combination of speaker, language and emotion is not supported")
            st.stop()
        # except Exception:
        #     st.error("Oops! Forget about it and hit F5 🙈")
        #     st.stop()


def main() -> None:
    hide_hamburger_menu()
    layout_app()


if __name__ == "__main__":
    st.set_page_config(
        page_title="EmoTTS Project",
        page_icon="💬",
        layout="centered",
        initial_sidebar_state="auto",
    )
    main()