fix: limit chunk size to 250

1 files changed, 82 insertions, 0 deletions

diff --git a/converter/audio.py b/converter/audio.py
index 6e5a910..3e163be 100644
--- a/converter/audio.py
+++ b/converter/audio.py
@@ -5,6 +5,7 @@ import re
 import shutil
 import subprocess
 import traceback
+import wave
 from pathlib import Path
 from typing import Dict, List, NamedTuple, Optional, Tuple
 
@@ -50,6 +51,87 @@ def _concat_escape(path: str) -> str:
     return path.replace("'", "'\\''")
 
 
+def _concat_wav_files(sources: List[Path], destination: Path) -> bool:
+    """Concatenate WAV files with matching parameters using the wave module.
+
+    Returns False (touching nothing) when any input is not a readable WAV
+    or the parameters differ, so the caller can fall back to ffmpeg.
+    """
+    opened = []
+    try:
+        parameters = None
+        for source in sources:
+            wav_file = wave.open(str(source), "rb")
+            opened.append(wav_file)
+            current = (wav_file.getnchannels(), wav_file.getsampwidth(),
+                       wav_file.getframerate())
+            if parameters is None:
+                parameters = current
+            elif current != parameters:
+                return False
+        if parameters is None or min(parameters) < 1:
+            return False
+        with wave.open(str(destination), "wb") as output:
+            output.setnchannels(parameters[0])
+            output.setsampwidth(parameters[1])
+            output.setframerate(parameters[2])
+            for wav_file in opened:
+                output.writeframes(wav_file.readframes(wav_file.getnframes()))
+        return True
+    except (wave.Error, EOFError, OSError):
+        return False
+    finally:
+        for wav_file in opened:
+            try:
+                wav_file.close()
+            except Exception:
+                pass
+
+
+def _concat_with_ffmpeg(sources: List[Path], destination: Path) -> None:
+    """Concatenate audio files with ffmpeg's concat demuxer, re-encoding to
+    16-bit PCM WAV (handles inputs the wave module cannot)."""
+    if shutil.which("ffmpeg") is None:
+        raise RuntimeError(
+            "ffmpeg is required to concatenate audio parts in non-WAV formats "
+            "(install ffmpeg and try again)"
+        )
+    list_path = destination.with_name(destination.stem + "_parts.txt")
+    try:
+        with open(list_path, "w", encoding="utf-8") as list_file:
+            for source in sources:
+                list_file.write(f"file '{_concat_escape(str(source))}'\n")
+        command = [
+            "ffmpeg", "-y", "-hide_banner", "-loglevel", "error",
+            "-f", "concat", "-safe", "0", "-i", str(list_path),
+            "-c:a", "pcm_s16le", str(destination),
+        ]
+        proc = subprocess.run(command, capture_output=True, text=True)
+        if proc.returncode != 0:
+            raise RuntimeError(
+                f"ffmpeg failed to concatenate audio parts: {proc.stderr[-500:]}")
+    finally:
+        try:
+            list_path.unlink(missing_ok=True)
+        except OSError:
+            pass
+
+
+def concat_audio_files(sources: List[Path], destination: Path) -> None:
+    """Concatenate audio files into one file at ``destination``.
+
+    Joins the audio returned by several TTS sub-requests for a single
+    chunk. Uses the stdlib wave module when every input is a WAV with
+    matching parameters (lossless, no external tools); otherwise falls
+    back to ffmpeg's concat demuxer with re-encoding.
+    """
+    if not sources:
+        raise ValueError("No audio files to concatenate")
+    if _concat_wav_files(sources, destination):
+        return
+    _concat_with_ffmpeg(sources, destination)
+
+
 def _encode_args(output_format: str) -> List[str]:
     """Return ffmpeg output codec/bitrate args for the requested container."""
     if output_format == "m4b":