forked from Rockachopa/Timmy-time-dashboard
This commit is contained in:
1
src/content/narration/__init__.py
Normal file
1
src/content/narration/__init__.py
Normal file
@@ -0,0 +1 @@
|
||||
"""TTS narration generation for episode segments."""
|
||||
191
src/content/narration/narrator.py
Normal file
191
src/content/narration/narrator.py
Normal file
@@ -0,0 +1,191 @@
|
||||
"""TTS narration generation for episode segments.
|
||||
|
||||
Supports two backends (in priority order):
|
||||
1. Kokoro-82M via ``mlx_audio`` (Apple Silicon, offline, highest quality)
|
||||
2. Piper TTS via subprocess (cross-platform, offline, good quality)
|
||||
|
||||
Both are optional — if neither is available the module logs a warning and
|
||||
returns a failure result rather than crashing the pipeline.
|
||||
|
||||
Usage
|
||||
-----
|
||||
from content.narration.narrator import generate_narration
|
||||
|
||||
result = await generate_narration(
|
||||
text="Welcome to today's highlights episode.",
|
||||
output_path="/tmp/narration.wav",
|
||||
)
|
||||
if result.success:
|
||||
print(result.audio_path)
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import logging
|
||||
import shutil
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
|
||||
from config import settings
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
@dataclass
|
||||
class NarrationResult:
|
||||
"""Result of a TTS narration generation attempt."""
|
||||
|
||||
success: bool
|
||||
audio_path: str | None = None
|
||||
backend: str | None = None
|
||||
error: str | None = None
|
||||
|
||||
|
||||
def _kokoro_available() -> bool:
|
||||
"""Return True if mlx_audio (Kokoro-82M) can be imported."""
|
||||
try:
|
||||
import importlib.util
|
||||
|
||||
return importlib.util.find_spec("mlx_audio") is not None
|
||||
except Exception:
|
||||
return False
|
||||
|
||||
|
||||
def _piper_available() -> bool:
|
||||
"""Return True if the piper binary is on PATH."""
|
||||
return shutil.which("piper") is not None
|
||||
|
||||
|
||||
async def _generate_kokoro(text: str, output_path: str) -> NarrationResult:
|
||||
"""Generate audio with Kokoro-82M via mlx_audio (runs in thread)."""
|
||||
try:
|
||||
import mlx_audio # type: ignore[import]
|
||||
|
||||
def _synth() -> None:
|
||||
mlx_audio.tts(
|
||||
text,
|
||||
voice=settings.content_tts_voice,
|
||||
output=output_path,
|
||||
)
|
||||
|
||||
await asyncio.to_thread(_synth)
|
||||
return NarrationResult(success=True, audio_path=output_path, backend="kokoro")
|
||||
except Exception as exc:
|
||||
logger.warning("Kokoro TTS failed: %s", exc)
|
||||
return NarrationResult(success=False, backend="kokoro", error=str(exc))
|
||||
|
||||
|
||||
async def _generate_piper(text: str, output_path: str) -> NarrationResult:
|
||||
"""Generate audio with Piper TTS via subprocess."""
|
||||
model = settings.content_piper_model
|
||||
cmd = [
|
||||
"piper",
|
||||
"--model", model,
|
||||
"--output_file", output_path,
|
||||
]
|
||||
try:
|
||||
proc = await asyncio.create_subprocess_exec(
|
||||
*cmd,
|
||||
stdin=asyncio.subprocess.PIPE,
|
||||
stdout=asyncio.subprocess.PIPE,
|
||||
stderr=asyncio.subprocess.PIPE,
|
||||
)
|
||||
_, stderr = await asyncio.wait_for(
|
||||
proc.communicate(input=text.encode()),
|
||||
timeout=120,
|
||||
)
|
||||
if proc.returncode != 0:
|
||||
err = stderr.decode(errors="replace")[-400:]
|
||||
logger.warning("Piper TTS failed: %s", err)
|
||||
return NarrationResult(success=False, backend="piper", error=err)
|
||||
return NarrationResult(success=True, audio_path=output_path, backend="piper")
|
||||
except TimeoutError:
|
||||
return NarrationResult(success=False, backend="piper", error="piper timed out")
|
||||
except Exception as exc:
|
||||
logger.warning("Piper TTS error: %s", exc)
|
||||
return NarrationResult(success=False, backend="piper", error=str(exc))
|
||||
|
||||
|
||||
async def generate_narration(
|
||||
text: str,
|
||||
output_path: str,
|
||||
) -> NarrationResult:
|
||||
"""Generate TTS narration for the given text.
|
||||
|
||||
Tries Kokoro-82M first (Apple Silicon), falls back to Piper.
|
||||
Returns a failure result if neither backend is available.
|
||||
|
||||
Parameters
|
||||
----------
|
||||
text:
|
||||
The script text to synthesise.
|
||||
output_path:
|
||||
Destination path for the audio file (wav/mp3).
|
||||
|
||||
Returns
|
||||
-------
|
||||
NarrationResult
|
||||
Always returns a result; never raises.
|
||||
"""
|
||||
if not text.strip():
|
||||
return NarrationResult(success=False, error="empty narration text")
|
||||
|
||||
Path(output_path).parent.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
if _kokoro_available():
|
||||
result = await _generate_kokoro(text, output_path)
|
||||
if result.success:
|
||||
return result
|
||||
logger.warning("Kokoro failed, trying Piper")
|
||||
|
||||
if _piper_available():
|
||||
return await _generate_piper(text, output_path)
|
||||
|
||||
logger.warning("No TTS backend available (install mlx_audio or piper)")
|
||||
return NarrationResult(
|
||||
success=False,
|
||||
error="no TTS backend available — install mlx_audio or piper",
|
||||
)
|
||||
|
||||
|
||||
def build_episode_script(
|
||||
episode_title: str,
|
||||
highlights: list[dict],
|
||||
outro_text: str | None = None,
|
||||
) -> str:
|
||||
"""Build a narration script for a full episode.
|
||||
|
||||
Parameters
|
||||
----------
|
||||
episode_title:
|
||||
Human-readable episode title for the intro.
|
||||
highlights:
|
||||
List of highlight dicts. Each may have a ``description`` key
|
||||
used as the narration text for that clip.
|
||||
outro_text:
|
||||
Optional custom outro. Defaults to a generic subscribe prompt.
|
||||
|
||||
Returns
|
||||
-------
|
||||
str
|
||||
Full narration script with intro, per-highlight lines, and outro.
|
||||
"""
|
||||
lines: list[str] = [
|
||||
f"Welcome to {episode_title}.",
|
||||
"Here are today's top highlights.",
|
||||
"",
|
||||
]
|
||||
for i, h in enumerate(highlights, 1):
|
||||
desc = h.get("description") or h.get("title") or f"Highlight {i}"
|
||||
lines.append(f"Highlight {i}. {desc}.")
|
||||
lines.append("")
|
||||
|
||||
if outro_text:
|
||||
lines.append(outro_text)
|
||||
else:
|
||||
lines.append(
|
||||
"Thanks for watching. Like and subscribe to stay updated on future episodes."
|
||||
)
|
||||
|
||||
return "\n".join(lines)
|
||||
Reference in New Issue
Block a user