Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix(minimax): align audio containers (#1639)

Rudy Mizrahi Celekli3b43ea443eef
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 0bdfc64b..a907336d 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -71,7 +71,7 @@ jobs:
- name: Run native revision-rendering tests
run: >-
uv run --no-sync python -X utf8 -m coverage run --append -m pytest -q
- test/services/test_video_project.py
+ test/services/test_video_project.py test/services/test_minimax_output_container.py
- name: Report test coverage
run: uv run --no-sync python -m coverage report
diff --git a/app/services/voice.py b/app/services/voice.py
index ed74737b..e71e3710 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -2322,6 +2322,29 @@ def _write_validated_minimax_audio(audio_bytes: bytes, voice_file: str) -> float
output.write(audio_bytes)
_validate_remote_tts_audio(temp_path)
+ # MiniMax's configured response container can differ from the caller's
+ # artifact extension (the task pipeline normally requests audio.mp3).
+ # Keep matching payloads byte-identical; transcode supported mismatches
+ # before publication so MIME/extension consumers see the promised codec.
+ source_format = (
+ "wav" if audio_bytes[:4] == b"RIFF" and audio_bytes[8:12] == b"WAVE"
+ else "flac" if audio_bytes.startswith(b"fLaC")
+ else "mp3" if audio_bytes.startswith(b"ID3") or (
+ len(audio_bytes) >= 2 and audio_bytes[0] == 0xFF
+ and audio_bytes[1] & 0xE0 == 0xE0
+ )
+ else None
+ )
+ output_format = output_suffix.lstrip(".").lower()
+ if source_format and output_format in {"mp3", "wav", "flac"} and source_format != output_format:
+ from pydub import AudioSegment
+
+ _configure_pydub_ffmpeg(AudioSegment)
+ segment = AudioSegment.from_file(io.BytesIO(audio_bytes), format=source_format)
+ exported = segment.export(temp_path, format=output_format)
+ if exported is not None:
+ exported.close()
+ _validate_remote_tts_audio(temp_path)
audio_clip = AudioFileClip(temp_path)
try:
audio_duration = float(audio_clip.duration)
diff --git a/test/services/test_minimax_output_container.py b/test/services/test_minimax_output_container.py
new file mode 100644
index 00000000..741bdbc2
--- /dev/null
+++ b/test/services/test_minimax_output_container.py
@@ -0,0 +1,45 @@
+"""Native provider-payload decoding; no paid synthesis requests."""
+import json
+import shutil
+import subprocess
+
+import pytest
+
+from app.services.voice import _write_validated_minimax_audio
+
+
+pytestmark = pytest.mark.skipif(
+ not shutil.which("ffmpeg") or not shutil.which("ffprobe"),
+ reason="native FFmpeg/FFprobe required; exercised in the post-install CI pass",
+)
+
+
[email protected]("source_format,output_format,codec", [
+ ("wav", "mp3", "mp3"),
+ ("flac", "mp3", "mp3"),
+ ("wav", "flac", "flac"),
+])
+def test_minimax_audio_matches_output_container(tmp_path, source_format, output_format, codec):
+ source = tmp_path / ("provider." + source_format)
+ subprocess.run([
+ "ffmpeg", "-v", "error", "-f", "lavfi", "-i",
+ "sine=frequency=440:duration=0.3", str(source),
+ ], check=True)
+ output = tmp_path / ("audio." + output_format)
+ duration = _write_validated_minimax_audio(source.read_bytes(), str(output))
+ assert 0.25 < duration < 0.5
+ probe = json.loads(subprocess.check_output([
+ "ffprobe", "-v", "error", "-show_streams", "-of", "json", str(output),
+ ]))
+ assert probe["streams"][0]["codec_name"] == codec
+
+
+def test_minimax_matching_wav_keeps_original_bytes(tmp_path):
+ source = tmp_path / "provider.wav"
+ subprocess.run([
+ "ffmpeg", "-v", "error", "-f", "lavfi", "-i",
+ "sine=frequency=440:duration=0.3", str(source),
+ ], check=True)
+ output = tmp_path / "audio.wav"
+ _write_validated_minimax_audio(source.read_bytes(), str(output))
+ assert output.read_bytes() == source.read_bytes()