Merge pull request #167 from aedocw/chapter-breaks-in-text

Chapter breaks in text

README.md

@@ -2,6 +2,8 @@ This script takes an epub (or text file) and reads it to an m4b audiobook file,
 
 I recognize this is not very user friendly, but I wanted to share in case folks thought it was useful. If there are a few more people than myself that find this is useful I will keep working on turning it into something that could be used by someone without dev experience.
 
+**NOTE:** Latest release adds a new workflow allowing you to export the epub to text, make any necessary modifications, then read the book as a text file. Any line beginning with "# " is considered a chapter break, and will be automatically inserted during export, named "# Part 1", etc. If you replace "Part 1" with whatever you want that section to be called it will be labeled that way in the audiobook metadata.
+
 **NOTE:** DeepSpeed support for XTTS has been added! If deepspeed is installed and you have a compatible GPU, it will be detected and used. For XTTS, this will yeild a 3x-4x speed improvement! Install deepspeed with `pip install deepspeed`.
 
 **NOTE:** The Coqui team released their curated XTTS voice models recently, and they sound great. A recent update here
@@ -14,10 +16,6 @@ Example usage: `epub2tts my-book.epub --engine xtts --speaker "Damien Black"`
 
 Example usage: `epub2tts my-book.epub --start 4 --end 20 --xtts shadow-1.wav,shadow-2.wav,shadow-3.wav`
 
-**NOTE:** Now with [OpenAI TTS](https://platform.openai.com/docs/guides/text-to-speech) support! It's not free, but the average cost for a few books I tested was around $7. If you use `--openai <API key>` flag epub2tts will provide a cost estimate and prompt you to approve before continuing.
-
-**NOTE:** HUGE thanks to a recent PR from [wonka929](https://github.com/wonka929), epub2tts now recognizes when a CUDA GPU is available and will use it automatically. In a brief test I did, the speedup was incredible!
-
 ## USAGE:
 Usage: 
 
@@ -35,6 +33,8 @@ To skip reading any links, add: `--skiplinks`
 
 Using `--scan` will list excerpts of each chapter, then exit. This is helpful for finding which chapter to start and end on if you want to skip bibliography, TOC, etc.
 
+Using `--export txt` will export the entire book to text file. This will honor `--start` and `--end` arguments as well.
+
 To specify which chapter to start on (ex 3): `--start 3`
 
 To specify which chapter to end on (ex 20): `--end 20`

epub2tts.py

@@ -58,6 +58,7 @@ def __init__(
         self.output_filename = self.bookname + ".m4b"
         self.chapters = []
         self.chapters_to_read = []
+        self.section_names = []
         if source.endswith(".epub"):
             self.book = epub.read_epub(source)
             self.sourcetype = "epub"
@@ -110,7 +111,10 @@ def generate_metadata(self, files, title, author):
                 file.write("TIMEBASE=1/1000\n")
                 file.write("START=" + str(start_time) + "\n")
                 file.write("END=" + str(start_time + duration) + "\n")
-                file.write("title=Part " + str(chap) + "\n")
+                if len(self.section_names) > 0:
+                    file.write(f"title={self.section_names[chap-1]}\n")
+                else:
+                    file.write("title=Part " + str(chap) + "\n")
                 chap += 1
                 start_time += duration
 
@@ -210,13 +214,24 @@ def get_chapters_text(self):
             text = file.read()
         text = self.prep_text(text)
         max_len = 50000
-        while len(text) > max_len:
-            pos = text.rfind(" ", 0, max_len)  # find the last space within the limit
-            self.chapters_to_read.append(text[:pos])
-            print(f"Part: {len(self.chapters_to_read)}")
-            print(str(self.chapters_to_read[-1])[:256])
-            text = text[pos + 1 :]  # +1 to avoid starting the next chapter with a space
-        self.chapters_to_read.append(text)
+        lines_with_hashtag = [line for line in text.splitlines() if line.startswith("# ")]
+        if lines_with_hashtag:
+            for line in lines_with_hashtag:
+                self.section_names.append(line.lstrip("# ").strip())
+            sections = re.split(r"#\s+", text.strip())[1:]
+            for section in sections:
+                if self.sayparts == False:
+                    lines = section.splitlines()
+                    section = "\n".join(lines[1:])
+                self.chapters_to_read.append(section.strip())
+        else:
+            while len(text) > max_len:
+                pos = text.rfind(" ", 0, max_len)  # find the last space within the limit
+                self.chapters_to_read.append(text[:pos])
+                print(f"Part: {len(self.chapters_to_read)}")
+                print(str(self.chapters_to_read[-1])[:256])
+                text = text[pos + 1 :]  # +1 to avoid starting the next chapter with a space
+            self.chapters_to_read.append(text)
         self.end = len(self.chapters_to_read)
 
     def read_chunk_xtts(self, sentences, wav_file_path):
@@ -408,7 +423,7 @@ def read_book(self, voice_samples, engine, openai, model_name, speaker, bitrate)
                 print(f"{outputflac} exists, skipping to next chapter")
             else:
                 tempfiles = []
-                if self.sayparts:
+                if self.sayparts and len(self.section_names) == 0:
                     chapter = "Part " + str(partnum + 1) + ". " + self.chapters_to_read[i]
                 else:
                     chapter = self.chapters_to_read[i]

setup.py

@@ -10,7 +10,7 @@
     author_email="doc@aedo.net",
     url="https://github.com/aedocw/epub2tts",
     license="Apache License, Version 2.0",
-    version="2.3.3",
+    version="2.3.4",
     packages=find_packages(),
     install_requires=requirements,
     py_modules=["epub2tts"],