update

main.py

@@ -7,12 +7,10 @@ import json
 import logging
 from pathlib import Path
 import platform
-import re
-import tempfile
 from typing import List
 import uuid
 
-from project_settings import project_path, log_directory, edge_tts_record_file, edge_tts_temp_directory
+from project_settings import project_path, log_directory, temp_directory, edge_tts_temp_directory
 import log
 
 log.setup(log_directory=log_directory)
@@ -21,6 +19,9 @@ import aiofiles
 import anyio
 import edge_tts
 import gradio as gr
+import librosa
+from scipy.io import wavfile
+import spacy
 
 from toolbox.os.command import Command
 
@@ -48,19 +49,57 @@ async def edge_tts_get_speakers() -> List[str]:
     return edge_tts_speakers_choices
 
 
-async def edge_tts_text_to_speech(text: str, speaker: str):
+async def edge_tts_text_to_speech(text: str, speaker: str, audio_dir: Path = edge_tts_temp_directory):
+    # tts
+    main_logger.info(f"edge tts; speaker: {speaker}; text: {text}")
     communicate = edge_tts.Communicate(text, speaker)
 
-    filename = edge_tts_temp_directory / "{}.wav".format(uuid.uuid4())
-    async with aiofiles.open(edge_tts_record_file.as_posix(), "a+", encoding="utf-8") as f:
-        await f.write(json.dumps({
-            "text": text,
-            "speaker": speaker,
-            "filename": filename.as_posix(),
-        }, ensure_ascii=False))
+    # save audio
+    audio_dir.mkdir(parents=True, exist_ok=True)
+    audio_file = audio_dir / "{}.wav".format(uuid.uuid4())
+    audio_file = audio_file.as_posix()
+    record_file = audio_dir / "edge_tts.jsonl"
+    try:
+        await communicate.save(audio_file)
+    except edge_tts.exceptions.NoAudioReceived:
+        audio_file = None
 
-    await communicate.save(filename)
-    return filename
+    # save record
+    async with aiofiles.open(record_file.as_posix(), "a+", encoding="utf-8") as f:
+        row = json.dumps({
+                "text": text,
+                "speaker": speaker,
+                "filename": audio_file,
+            },
+            ensure_ascii=False)
+        await f.write("{}\n".format(row))
+    return audio_file
+
+
+spacy_model = spacy.load("zh_core_web_sm")
+
+
+async def e_book_reading(txt_file: str, speaker: str):
+    txt_file = Path(txt_file)
+
+    audio_dir = temp_directory / "e_book_reading" / txt_file.stem / speaker
+
+    while True:
+        async with aiofiles.open(txt_file.as_posix(), "r", encoding="utf-8") as f:
+            data = await f.read()
+        doc = spacy_model(data)
+        for sentence in doc.sents:
+            text = sentence.text.strip()
+            if len(text) == 0:
+                continue
+            filename = await edge_tts_text_to_speech(text=text, speaker=speaker, audio_dir=audio_dir)
+
+            # sample_rate, signal = wavfile.read(filename)
+            signal, sample_rate = librosa.load(filename)
+            duration = len(signal) / sample_rate
+
+            yield filename
+            await asyncio.sleep(duration)
 
 
 def shell(cmd: str):
@@ -80,15 +119,6 @@ def main():
         gr.Markdown(value=title)
 
         with gr.Tabs():
-            with gr.TabItem("Ebook Reading"):
-                e_book_reading_file = gr.File(
-                    value=(project_path / "data/e_book/confucianism/the_analects.txt").as_posix(),
-                    label="txt"
-                )
-                e_book_reading_tts_engine = gr.Dropdown(
-                    choices=["Edge TTS"], value="Edge TTS", label="tts_engine"
-                )
-                e_book_reading_button = gr.Button(value="e_book_reading", variant="primary")
             with gr.TabItem("Edge TTS"):
                 edge_tts_text = gr.Textbox(value="学而时习之，不亦悦乎。", lines=4, max_lines=50, label="text")
                 edge_tts_speaker = gr.Dropdown(choices=edge_tts_speakers_choices, value="zh-CN-XiaoxiaoNeural", label="speakers")
@@ -96,30 +126,52 @@ def main():
                 edge_tts_audio = gr.Audio(type="filepath", label="audio", autoplay=True)
 
                 edge_tts_button = gr.Button(value="edge_tts", variant="primary")
-            edge_tts_button.click(
-                edge_tts_text_to_speech,
-                inputs=[
-                    edge_tts_text,
-                    edge_tts_speaker,
-                ],
-                outputs=[
-                    edge_tts_audio
-                ],
-            )
+                edge_tts_button.click(
+                    edge_tts_text_to_speech,
+                    inputs=[
+                        edge_tts_text,
+                        edge_tts_speaker,
+                    ],
+                    outputs=[
+                        edge_tts_audio
+                    ],
+                )
+
+            with gr.TabItem("Ebook Reading"):
+                e_book_reading_file = gr.File(
+                    value=(project_path / "data/e_book/confucianism/the_analects/the_analects.txt").as_posix(),
+                    label="txt"
+                )
+                e_book_reading_speaker = gr.Dropdown(choices=edge_tts_speakers_choices, value="zh-CN-XiaoxiaoNeural", label="speakers")
+
+                e_book_reading_audio = gr.Audio(type="filepath", label="audio", streaming=True)
+                e_book_reading_button = gr.Button(value="e_book_reading", variant="primary")
+
+                e_book_reading_button.click(
+                    e_book_reading,
+                    inputs=[
+                        e_book_reading_file,
+                        e_book_reading_speaker,
+                    ],
+                    outputs=[
+                        e_book_reading_audio
+                    ],
+                )
+
             with gr.TabItem("shell"):
                 shell_text = gr.Textbox(label="cmd")
                 shell_button = gr.Button("run")
                 shell_output = gr.Textbox(label="output")
 
-            shell_button.click(
-                shell,
-                inputs=[
-                    shell_text,
-                ],
-                outputs=[
-                    shell_output
-                ],
-            )
+                shell_button.click(
+                    shell,
+                    inputs=[
+                        shell_text,
+                    ],
+                    outputs=[
+                        shell_output
+                    ],
+                )
 
     launch = partial(
         blocks.queue().launch,

project_settings.py

@@ -18,8 +18,6 @@ temp_directory.mkdir(parents=True, exist_ok=True)
 edge_tts_temp_directory = temp_directory / "edge_tts"
 edge_tts_temp_directory.mkdir(parents=True, exist_ok=True)
 
-edge_tts_record_file = edge_tts_temp_directory / "edge_tts.jsonl"
-
 environment = EnvironmentManager(
     path=os.path.join(project_path, "dotenv"),
     env=os.environ.get("environment", "dev"),

requirements.txt

@@ -2,3 +2,5 @@ gradio==4.38.1
 python-dotenv==1.0.1
 spacy==3.7.5
 edge-tts==6.1.12
+scipy==1.14.0
+librosa==0.10.2.post1