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(tts): stop MiniMax retries after accepted synthesis

Rudy Celeklie445e48d8a5d
diff --git a/app/services/voice.py b/app/services/voice.py
index 40cc6a92..4d138237 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -2237,25 +2237,34 @@ def minimax_tts(text: str, voice_id: str, voice_rate: float, voice_file: str, vo
}
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
for attempt in range(3):
+ received_success = False
try:
logger.info(f"start MiniMax TTS, model: {model}, voice: {voice_id}, try: {attempt + 1}")
response = requests.post(url, json=payload, headers=headers, timeout=120)
if response.status_code != 200:
logger.error(f"MiniMax TTS failed with status {response.status_code}: {response.text[:200]}")
continue
+ received_success = True
body = response.json()
data = body.get("data") or {}
base_resp = body.get("base_resp") or {}
- if base_resp.get("status_code") != 0 or data.get("status") != 2:
+ status_code = base_resp.get("status_code")
+ if not isinstance(status_code, int) or isinstance(status_code, bool):
+ logger.error("MiniMax returned an unknown acceptance status; stop paid retries")
+ return None
+ if status_code != 0:
logger.error(f"MiniMax TTS returned an unsuccessful response: status_code={base_resp.get('status_code')}, audio_status={data.get('status')}")
continue
+ if data.get("status") != 2:
+ logger.error("MiniMax accepted the request but returned incomplete audio")
+ return None
audio_hex = data.get("audio")
if not isinstance(audio_hex, str) or not audio_hex:
logger.error("MiniMax TTS returned empty audio data")
- continue
+ return None
if len(audio_hex) > _MINIMAX_TTS_MAX_AUDIO_HEX_CHARS:
logger.error("MiniMax TTS returned audio data exceeding the supported size")
- continue
+ return None
audio_duration = _write_validated_minimax_audio(bytes.fromhex(audio_hex), voice_file)
logger.success(f"MiniMax TTS succeeded: {voice_file}")
return populate_legacy_submaker_with_full_text(
@@ -2269,8 +2278,12 @@ def minimax_tts(text: str, voice_id: str, voice_rate: float, voice_file: str, vo
f"stop paid retries: {type(exc).__name__}"
)
return None
- except (OSError, ValueError) as exc:
+ except (OSError, ValueError, TypeError, AttributeError) as exc:
logger.error(f"MiniMax TTS failed: {str(exc)}")
+ if received_success:
+ # The request may already be billed, even if JSON/audio parsing
+ # or local file publication failed. Never regenerate it here.
+ return None
return None
diff --git a/test/services/test_minimax_success_retry.py b/test/services/test_minimax_success_retry.py
new file mode 100644
index 00000000..7f8c585b
--- /dev/null
+++ b/test/services/test_minimax_success_retry.py
@@ -0,0 +1,42 @@
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import Mock, patch
+import requests
+from app.services import voice
+
+
+class TestMinimaxAcceptedResponseRetry(unittest.TestCase):
+ def test_successful_paid_response_is_never_resubmitted_for_invalid_audio(self):
+ for audio, error in [("not-hex", None), ("01", ValueError("decode failed")), ("01", OSError("disk full")), ("", None)]:
+ with self.subTest(audio=audio, error=error), tempfile.TemporaryDirectory() as tmp:
+ response = Mock(status_code=200)
+ response.json.return_value = {"base_resp": {"status_code": 0}, "data": {"status": 2, "audio": audio}}
+ output = Path(tmp) / "narration.mp3"
+ output.write_bytes(b"previous narration")
+ with patch.dict(voice.config.minimax_tts, {"api_key": "test-key"}), patch.object(voice.requests, "post", return_value=response) as post, patch.object(voice, "_write_validated_minimax_audio", side_effect=error, return_value=1):
+ result = voice.minimax_tts("Hello.", "voice-id", 1, str(output))
+ self.assertIsNone(result)
+ self.assertEqual(post.call_count, 1, "paid synthesis was resubmitted")
+ self.assertEqual(output.read_bytes(), b"previous narration")
+
+ def test_connect_timeout_before_acceptance_can_retry(self):
+ response = Mock(status_code=200)
+ response.json.return_value = {"base_resp": {"status_code": 0}, "data": {"status": 2, "audio": "01"}}
+ with patch.dict(voice.config.minimax_tts, {"api_key": "test-key"}), patch.object(voice.requests, "post", side_effect=[requests.ConnectTimeout("not connected"), response]) as post, patch.object(voice, "_write_validated_minimax_audio", return_value=1):
+ result = voice.minimax_tts("Hello.", "voice-id", 1, "unused.mp3")
+ self.assertIsNotNone(result)
+ self.assertEqual(post.call_count, 2)
+
+ def test_unreadable_or_incomplete_success_body_is_not_resubmitted(self):
+ for body in (None, [], {}, {"base_resp": {}}, {"data": {"status": 2, "audio": "01"}}, {"base_resp": {"status_code": 0}, "data": {"status": 1}}):
+ with self.subTest(body=body):
+ response = Mock(status_code=200)
+ if body is None:
+ response.json.side_effect = ValueError("invalid JSON")
+ else:
+ response.json.return_value = body
+ with patch.dict(voice.config.minimax_tts, {"api_key": "test-key"}), patch.object(voice.requests, "post", return_value=response) as post:
+ result = voice.minimax_tts("Hello.", "voice-id", 1, "unused.mp3")
+ self.assertIsNone(result)
+ post.assert_called_once()