playreel/backend/tests/poster_alive/test_tts.py

54 lines
2.2 KiB
Python

"""tts 수동 확인용.
사용: uv run python -m tests.poster_alive.test_tts <포스터경로>
test_result/<이름>.narration.txt 가 있으면 재사용하고, 없으면 앞 단계부터 돌린다.
결과는 <이름>.tts.mp3(합쳐진 나레이션)와 <이름>.tts.txt(타임라인)로 떨어진다.
"""
import asyncio
import sys
from pathlib import Path
from answers.narration_answer import NarrationAnswer
from services.narration import generate_narration
from services.tts import synthesize
from tests.poster_alive.test_narration import load_regions
TEST_RESULT_DIR = Path(__file__).parent.parent / "test_result" / "poster_alive"
async def load_narration(poster_path: Path, output_base: Path) -> NarrationAnswer:
narration_path = output_base.with_suffix(".narration.txt")
if narration_path.exists():
return NarrationAnswer.model_validate_json(narration_path.read_text(encoding="utf-8"))
regions = await load_regions(poster_path, output_base.with_suffix(".detect.txt"))
answer = await generate_narration(poster_path, regions)
narration_path.write_text(answer.model_dump_json(indent=2), encoding="utf-8")
return answer
async def main() -> None:
poster_path = Path(sys.argv[1])
TEST_RESULT_DIR.mkdir(exist_ok=True)
output_base = TEST_RESULT_DIR / poster_path.stem
answer = await load_narration(poster_path, output_base)
audio = await synthesize(answer.narration, answer.voice)
mp3_path = output_base.with_suffix(".tts.mp3")
timeline_path = output_base.with_suffix(".tts.txt")
mp3_path.write_bytes(audio.mp3)
timeline_path.write_text(audio.timeline.model_dump_json(indent=2), encoding="utf-8")
print(f"engine={audio.timeline.engine} voice={audio.timeline.voice}")
for cue in audio.timeline.cues:
print(f" [{cue.index}] {cue.start:5.2f}s +{cue.duration:4.2f}s {cue.text}")
print(f"\n나레이션 종료 {audio.timeline.cues[-1].end:.2f}s "
f"+ 여백 {audio.timeline.tail}s → 영상 길이 {audio.timeline.total}s")
if audio.timeline.total > 8:
print("※ 8초를 넘는다 — 영상이 8초 고정이라 나레이션이 잘린다")
print(f"\n{mp3_path}\n{timeline_path}")
if __name__ == "__main__":
asyncio.run(main())