feat: local tts support

.gitignore

@@ -25,4 +25,8 @@ Thumbs.db
 # audio files
 audiobook_output/
 
-private_examples/
+# piper models
+*.onnx
+*.onnx.json
+
+private_examples/

README.md

@@ -16,6 +16,7 @@ If you're interested in hearing a sample of the audiobook generated by this tool
 - Python 3.6+ Or ***Docker***
 - For using Azure TTS, A Microsoft Azure account with access to the [Microsoft Cognitive Services Speech Services](https://portal.azure.com/#create/Microsoft.CognitiveServicesSpeechServices) is required.
 - For using OpenAI TTS, OpenAI [API Key](https://platform.openai.com/api-keys) is required.
+- For using Local TTS, a local TTS app is required, example [piper](https://github.com/rhasspy/piper).
 
 ## Audiobookshelf Integration
 
@@ -312,6 +313,28 @@ Here are some examples that demonstrate various option combinations:
    python3 epub_to_audiobook.py "path/to/book.epub" "path/to/output/folder" --tts openai --preview --output_text
    ```
 
+### Examples Using Local TTS
+
+1. **Local TTS with default settings**  
+   This command will convert an EPUB file to an audiobook using local [piper tts](https://github.com/rhasspy/piper).
+
+   ```sh
+   python3 epub_to_audiobook.py "path/to/book.epub" "path/to/output/folder" --tts local
+   ```
+2. **Local TTS with custom command**  
+
+   ```sh
+   python3 epub_to_audiobook.py "path/to/book.epub" "path/to/output/folder" --tts local --local_tts_cmd ./tts.sh 
+   ```
+
+   Example `tts.sh` powered by [coqui-ai/TTS](https://github.com/coqui-ai/TTS).
+
+   ```bash
+   #!/bin/bash
+
+   tts --text="$(cat)" --out_path "$@"
+   ```
+
 ## Troubleshooting
 
 ### ModuleNotFoundError: No module named 'importlib_metadata'

epub_to_audiobook.py

@@ -2,6 +2,7 @@
 import re
 import io
 import argparse
+import subprocess
 import html
 import ebooklib
 from ebooklib import epub
@@ -30,7 +31,7 @@
 
 TTS_AZURE = "azure"
 TTS_OPENAI = "openai"
-
+TTS_LOCAL = "local"
 
 @dataclasses.dataclass
 class AudioTags:
@@ -255,6 +256,51 @@ def text_to_speech(self, text: str, output_file: str, audio_tags: AudioTags):
 
             set_audio_tags(output_file, audio_tags)
 
+class LocalTTSProvider(TTSProvider):
+    def __init__(self, general_config: GeneralConfig, cmd, format, max_chars, break_ssml):
+        super().__init__(general_config)
+        self.cmd = cmd
+        self.format = format
+        self.max_chars = max_chars
+        self.break_ssml = break_ssml
+
+    def __str__(self) -> str:
+        return (super().__str__() + f", cmd={self.cmd}")
+
+    def text_to_speech(self, text: str, output_file: str, audio_tags: AudioTags):
+        text_chunks = split_text(text, self.max_chars, self.general_config.language)
+
+        with open(output_file, "wb") as outfile:
+            for i, chunk in enumerate(text_chunks, 1):
+                logger.debug(
+                    f"Processing chunk {i} of {len(text_chunks)}, length={len(chunk)}, text=[{chunk}]"
+                )
+                # replace MAGIC_BREAK_STRING with break SSML
+                chunk = chunk.replace(
+                    MAGIC_BREAK_STRING.strip(),
+                    self.break_ssml,
+                )  # strip in case leading bank is missing
+                logger.info(
+                    f"Processing chapter-{audio_tags.idx} <{audio_tags.title}>, chunk {i} of {len(text_chunks)}"
+                )
+
+                logger.debug(f"Text: [{chunk}], length={len(chunk)}")
+
+                chuck_file = f"{output_file}.{i}"
+                process = subprocess.Popen(f"{self.cmd} {chuck_file}", stdin=subprocess.PIPE, stdout=subprocess.PIPE, shell=True)
+                stdout, stderr = process.communicate(input=chunk.encode())
+                if stdout:
+                    logger.info(stdout.decode())
+                if stderr:
+                    logger.error(stderr.decode())
+
+                segment = io.FileIO(chuck_file)
+                segment.seek(0)
+                outfile.write(segment.read())
+                os.remove(chuck_file)
+
+        set_audio_tags(output_file, audio_tags)
+
 
 def sanitize_title(title: str) -> str:
     # replace MAGIC_BREAK_STRING with a blank space
@@ -429,6 +475,8 @@ def epub_to_audiobook(tts_provider: TTSProvider):
         audio_suffix = f"{tts_provider.format}"  # mp3, opus, aac, or flac
     elif isinstance(tts_provider, AzureTTSProvider):
         audio_suffix = "mp3"  # only mp3 is supported for Azure TTS for now
+    elif isinstance(tts_provider, LocalTTSProvider):
+        audio_suffix = tts_provider.format
     else:
         raise ValueError(f"Invalid TTS provider: {tts_provider.general_config.tts}")
 
@@ -473,7 +521,7 @@ def main():
     parser.add_argument("output_folder", help="Path to the output folder")
     parser.add_argument(
         "--tts",
-        choices=[TTS_AZURE, TTS_OPENAI],
+        choices=[TTS_AZURE, TTS_OPENAI, TTS_LOCAL],
         default=TTS_AZURE,
         help="Choose TTS provider (default: azure). azure: Azure Cognitive Services, openai: OpenAI TTS API. When using azure, environment variables MS_TTS_KEY and MS_TTS_REGION must be set. When using openai, environment variable OPENAI_API_KEY must be set.",
     )
@@ -556,6 +604,29 @@ def main():
         default="mp3",
         help="Available OpenAI output options: mp3, opus, aac, and flac. Check https://platform.openai.com/docs/guides/text-to-speech/supported-output-formats.",
     )
+    # Local TTS specific arguments
+    local_group = parser.add_argument_group("Local TTS Options")
+    local_group.add_argument(
+        "--local_tts_cmd",
+        default="piper --model en_US-lessac-medium --output_file",
+        help="command to use for local TTS, it should accept text via stdin and write the resulting audio to the filename passed as argument.",
+    )
+    local_group.add_argument(
+        "--local_tts_format",
+        default="wav",
+        help="The output format the local TTS service outputs (default: wav).",
+    )
+    local_group.add_argument(
+        "--local_tts_max_chars",
+        type=int,
+        default=10000,
+        help="The max chars the local TTS service could handle at once (default: 10000).",
+    )
+    local_group.add_argument(
+        "--local_tts_break_ssml",
+        default="   ",
+        help="The break SSML supported by the local TTS service (default: \"   \").",
+    )
 
     args = parser.parse_args()
 
@@ -574,6 +645,14 @@ def main():
         tts_provider = OpenAITTSProvider(
             general_config, args.openai_model, args.openai_voice, args.openai_format
         )
+    elif args.tts == TTS_LOCAL:
+        tts_provider = LocalTTSProvider(
+            general_config,
+            args.local_tts_cmd,
+            args.local_tts_format,
+            args.local_tts_max_chars,
+            args.local_tts_break_ssml,
+        )
     else:
         raise ValueError(f"Invalid TTS provider: {args.tts}")