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.

feat: add MiniMax text-to-speech support

octo-patchfafa572ef52b
diff --git a/app/config/config.py b/app/config/config.py
index 28af63af..4a22ebf3 100644
--- a/app/config/config.py
+++ b/app/config/config.py
@@ -463,6 +463,7 @@ def save_config():
config_to_save["app"] = dict(app)
config_to_save["azure"] = dict(azure)
config_to_save["siliconflow"] = dict(siliconflow)
+ config_to_save["minimax_tts"] = dict(minimax_tts)
config_to_save["elevenlabs"] = dict(elevenlabs)
config_to_save["chatterbox"] = dict(chatterbox)
config_to_save["ui"] = dict(ui)
@@ -517,6 +518,7 @@ whisper = _cfg.get("whisper", {})
proxy = _cfg.get("proxy", {})
azure = _SynchronizedConfig(_cfg.get("azure", {}))
siliconflow = _SynchronizedConfig(_cfg.get("siliconflow", {}))
+minimax_tts = _SynchronizedConfig(_cfg.get("minimax_tts", {}))
elevenlabs = _SynchronizedConfig(_cfg.get("elevenlabs", {}))
chatterbox = _SynchronizedConfig(_cfg.get("chatterbox", {}))
ui = _SynchronizedConfig(
diff --git a/app/services/voice.py b/app/services/voice.py
index 3ef5fb85..e7b1841d 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -29,6 +29,14 @@ from app.utils import utils
_DEFAULT_EDGE_TTS_TIMEOUT_SECONDS = 30.0
_MIMO_DEFAULT_BASE_URL = "https://api.xiaomimimo.com/v1"
_MIMO_DEFAULT_TTS_MODEL = "mimo-v2.5-tts"
+MINIMAX_TTS_GLOBAL_URL = "https://api.minimax.io/v1/t2a_v2"
+MINIMAX_TTS_CN_URL = "https://api.minimaxi.com/v1/t2a_v2"
+MINIMAX_TTS_DEFAULT_MODEL = "speech-2.8-hd"
+MINIMAX_TTS_DEFAULT_VOICE = "English_expressive_narrator"
+MINIMAX_TTS_MODELS = (
+ "speech-2.8-hd", "speech-2.8-turbo", "speech-2.6-hd", "speech-2.6-turbo",
+ "speech-02-hd", "speech-02-turbo", "speech-01-hd", "speech-01-turbo",
+)
NO_VOICE_NAME = "no-voice"
# `none` 是 PR #981 里曾使用过的无配音标识。这里短期兼容这个值,避免
# 已经手动调用过该分支的 API 用户升级后立即失效;WebUI 和新代码统一使用
@@ -139,6 +147,12 @@ def get_mimo_voices() -> list[str]:
return [f"mimo:{voice}-{gender}" for voice, gender in voices_with_gender]
+def get_minimax_voices() -> list[str]:
+ """Return the configured MiniMax voice for the TTS selector."""
+ voice_id = str(config.minimax_tts.get("voice_id", MINIMAX_TTS_DEFAULT_VOICE) or MINIMAX_TTS_DEFAULT_VOICE).strip()
+ return [f"minimax:{voice_id}"]
+
+
def get_elevenlabs_voices(api_key: str) -> list[str]:
if not api_key:
return []
@@ -248,6 +262,10 @@ def is_mimo_voice(voice_name: str):
return voice_name.startswith("mimo:")
+def is_minimax_voice(voice_name: str | None) -> bool:
+ return (voice_name or "").startswith("minimax:")
+
+
def is_elevenlabs_voice(voice_name: str) -> bool:
return (voice_name or "").startswith("elevenlabs:")
@@ -433,6 +451,12 @@ def tts(
else:
logger.error(f"Invalid mimo voice name format: {voice_name}")
return None
+ elif is_minimax_voice(voice_name):
+ voice_id = voice_name.split(":", 1)[1].strip()
+ if voice_id:
+ return minimax_tts(text, voice_id, voice_rate, voice_file, voice_volume)
+ logger.error(f"Invalid MiniMax voice name format: {voice_name}")
+ return None
elif is_elevenlabs_voice(voice_name):
# 格式: elevenlabs:{voice_id}:{name}
parts = voice_name.split(":")
@@ -1287,6 +1311,85 @@ def mimo_tts(
return None
+def _resolve_minimax_tts_url(configured_url: str) -> str:
+ configured_url = (configured_url or "").strip().rstrip("/")
+ if not configured_url:
+ return MINIMAX_TTS_GLOBAL_URL
+ if configured_url in {MINIMAX_TTS_GLOBAL_URL, MINIMAX_TTS_CN_URL}:
+ return configured_url
+ if configured_url.endswith("/v1"):
+ return f"{configured_url}/t2a_v2"
+ return configured_url
+
+
+def minimax_tts(text: str, voice_id: str, voice_rate: float, voice_file: str, voice_volume: float = 1.0) -> Union[SubMaker, None]:
+ """Generate speech with the synchronous MiniMax T2A HTTP API."""
+ text, voice_id = (text or "").strip(), (voice_id or "").strip()
+ if not text or not voice_id:
+ logger.error("MiniMax TTS requires text and a voice ID")
+ return None
+ settings = config.minimax_tts
+ api_key = str(settings.get("api_key", "") or config.app.get("minimax_api_key", "") or "").strip()
+ if not api_key:
+ logger.error("MiniMax TTS API key is not set")
+ return None
+ url = _resolve_minimax_tts_url(str(settings.get("base_url", "") or ""))
+ model = str(settings.get("model_id", MINIMAX_TTS_DEFAULT_MODEL) or MINIMAX_TTS_DEFAULT_MODEL).strip()
+ if model not in MINIMAX_TTS_MODELS:
+ logger.error(f"Unsupported MiniMax TTS model: {model}")
+ return None
+ try:
+ speed = max(0.5, min(2.0, float(voice_rate or 1.0)))
+ volume = max(0.0, min(10.0, float(voice_volume or 1.0)))
+ pitch = max(-12, min(12, int(settings.get("pitch", 0) or 0)))
+ sample_rate = int(settings.get("sample_rate", 32000) or 32000)
+ bitrate = int(settings.get("bitrate", 128000) or 128000)
+ channel = int(settings.get("channel", 1) or 1)
+ except (TypeError, ValueError) as exc:
+ logger.error(f"Invalid MiniMax TTS audio setting: {str(exc)}")
+ return None
+ audio_format = str(settings.get("audio_format", "mp3") or "mp3").strip()
+ if audio_format not in {"mp3", "wav", "flac", "pcm"}:
+ logger.error(f"Unsupported MiniMax TTS audio format: {audio_format}")
+ return None
+ payload = {
+ "model": model, "text": text, "stream": False, "language_boost": "auto", "output_format": "hex",
+ "voice_setting": {"voice_id": voice_id, "speed": speed, "vol": volume, "pitch": pitch},
+ "audio_setting": {"sample_rate": sample_rate, "bitrate": bitrate, "format": audio_format, "channel": channel},
+ }
+ headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+ for attempt in range(3):
+ 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
+ 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:
+ logger.error(f"MiniMax TTS returned an unsuccessful response: status_code={base_resp.get('status_code')}, audio_status={data.get('status')}")
+ continue
+ audio_hex = data.get("audio")
+ if not isinstance(audio_hex, str) or not audio_hex:
+ logger.error("MiniMax TTS returned empty audio data")
+ continue
+ ensure_file_path_exists(voice_file)
+ with open(voice_file, "wb") as output:
+ output.write(bytes.fromhex(audio_hex))
+ audio_clip = AudioFileClip(voice_file)
+ audio_duration = audio_clip.duration
+ audio_clip.close()
+ logger.success(f"MiniMax TTS succeeded: {voice_file}")
+ return populate_legacy_submaker_with_full_text(
+ ensure_legacy_submaker_fields(SubMaker()), text, audio_duration
+ )
+ except (OSError, ValueError, requests.RequestException) as exc:
+ logger.error(f"MiniMax TTS failed: {str(exc)}")
+ return None
+
+
def elevenlabs_tts(
text: str,
voice_id: str,
diff --git a/config.example.toml b/config.example.toml
index 2be87616..afac1962 100644
--- a/config.example.toml
+++ b/config.example.toml
@@ -314,6 +314,21 @@ speech_region = ""
# SiliconFlow TTS API key: https://siliconflow.cn/
api_key = ""
+[minimax_tts]
+# API key: https://platform.minimax.io/user-center/basic-information/interface-key
+# Global endpoint: https://api.minimax.io/v1/t2a_v2
+# China endpoint: https://api.minimaxi.com/v1/t2a_v2
+# The MiniMax text-generation API key is used as a fallback when this is empty.
+api_key = ""
+base_url = "https://api.minimax.io/v1/t2a_v2"
+model_id = "speech-2.8-hd"
+voice_id = "English_expressive_narrator"
+sample_rate = 32000
+bitrate = 128000
+audio_format = "mp3"
+channel = 1
+pitch = 0
+
[elevenlabs]
# API key: https://elevenlabs.io/app/settings/api-keys
# Favorite voices in the ElevenLabs voice library are shown in the WebUI.
diff --git a/test/services/test_voice.py b/test/services/test_voice.py
index 5e4c8a09..37c0a604 100644
--- a/test/services/test_voice.py
+++ b/test/services/test_voice.py
@@ -587,6 +587,58 @@ class TestVoiceService(unittest.TestCase):
self.assertEqual(getattr(sub_maker, "subs", []), ["小米语音合成测试", "第二句话"])
self.assertEqual(len(getattr(sub_maker, "offset", [])), 2)
+ def test_minimax_tts_uses_regional_endpoint_and_hex_audio(self):
+ class _Response:
+ status_code, text = 200, ""
+
+ @staticmethod
+ def json():
+ return {"data": {"audio": b"audio".hex(), "status": 2}, "base_resp": {"status_code": 0}}
+
+ class _Clip:
+ duration = 2.5
+
+ def close(self):
+ pass
+
+ captured = {}
+
+ def _post(url, json=None, headers=None, timeout=None):
+ captured.update(url=url, json=json, headers=headers, timeout=timeout)
+ return _Response()
+
+ settings = {
+ "api_key": "test-key", "base_url": vs.MINIMAX_TTS_CN_URL,
+ "model_id": "speech-2.8-turbo", "voice_id": "male-qn-qingse",
+ "sample_rate": 32000, "bitrate": 128000, "audio_format": "mp3", "channel": 1,
+ }
+ with tempfile.TemporaryDirectory() as tmp_dir, patch.object(
+ vs.config, "minimax_tts", settings
+ ), patch.object(vs.requests, "post", side_effect=_post), patch.object(
+ vs, "AudioFileClip", return_value=_Clip()
+ ):
+ voice_file = str(Path(tmp_dir) / "minimax.mp3")
+ result = vs.minimax_tts("Speech test.", "male-qn-qingse", 1.2, voice_file, 1.5)
+ self.assertEqual(Path(voice_file).read_bytes(), b"audio")
+
+ self.assertIsNotNone(result)
+ self.assertEqual(captured["url"], "https://api.minimaxi.com/v1/t2a_v2")
+ self.assertEqual(captured["headers"]["Authorization"], "Bearer test-key")
+ self.assertEqual(captured["json"]["model"], "speech-2.8-turbo")
+ self.assertEqual(captured["json"]["text"], "Speech test.")
+ self.assertEqual(captured["json"]["voice_setting"]["voice_id"], "male-qn-qingse")
+ self.assertEqual(captured["json"]["audio_setting"]["format"], "mp3")
+
+ def test_minimax_voice_helpers_and_dispatch(self):
+ with patch.object(vs.config, "minimax_tts", {"voice_id": "narrator"}):
+ self.assertEqual(vs.get_minimax_voices(), ["minimax:narrator"])
+ self.assertTrue(vs.is_minimax_voice("minimax:narrator"))
+ sentinel = object()
+ with patch.object(vs, "minimax_tts", return_value=sentinel) as implementation:
+ result = vs.tts("test", "minimax:narrator", 1.0, "voice.mp3", 1.0)
+ self.assertIs(result, sentinel)
+ implementation.assert_called_once_with("test", "narrator", 1.0, "voice.mp3", 1.0)
+
def test_chatterbox_voice_helpers(self):
"""is_chatterbox_voice / get_chatterbox_voices basics and normalisation."""
self.assertTrue(vs.is_chatterbox_voice("chatterbox:default-Female"))
diff --git a/webui/Main.py b/webui/Main.py
index f386944e..3f20e75a 100644
--- a/webui/Main.py
+++ b/webui/Main.py
@@ -126,6 +126,7 @@ _RUNTIME_CONFIG_SECTIONS = {
"azure": config.azure,
"chatterbox": config.chatterbox,
"elevenlabs": config.elevenlabs,
+ "minimax_tts": config.minimax_tts,
"siliconflow": config.siliconflow,
"ui": config.ui,
}
@@ -951,6 +952,8 @@ def _infer_tts_server_from_voice(voice_name):
return "gemini-tts"
if voice.is_mimo_voice(voice_name):
return "mimo-tts"
+ if voice.is_minimax_voice(voice_name):
+ return "minimax-tts"
if voice.is_elevenlabs_voice(voice_name):
return "elevenlabs"
if voice.is_chatterbox_voice(voice_name):
@@ -2663,6 +2666,16 @@ def _get_voice_preview_provider_signature(tts_server: str) -> dict:
}
if tts_server == "mimo-tts":
return {"credential": _credential_signature(config.app.get("mimo_api_key", ""))}
+ if tts_server == "minimax-tts":
+ return {
+ "base_url": config.minimax_tts.get("base_url", ""),
+ "model_id": config.minimax_tts.get("model_id", ""),
+ "voice_id": config.minimax_tts.get("voice_id", ""),
+ "credential": _credential_signature(
+ config.minimax_tts.get("api_key", "")
+ or config.app.get("minimax_api_key", "")
+ ),
+ }
if tts_server == "elevenlabs":
return {
"model_id": config.elevenlabs.get("model_id", ""),
@@ -3247,6 +3260,7 @@ def _render_audio_settings(panel, params):
("siliconflow", "SiliconFlow TTS"),
("gemini-tts", "Google Gemini TTS"),
("mimo-tts", "Xiaomi MiMo TTS"),
+ ("minimax-tts", "MiniMax TTS"),
("elevenlabs", "ElevenLabs TTS"),
("chatterbox", "Chatterbox TTS"),
]
@@ -3295,6 +3309,8 @@ def _render_audio_settings(panel, params):
elif selected_tts_server == "mimo-tts":
# 获取 Xiaomi MiMo TTS 的预置音色列表
filtered_voices = voice.get_mimo_voices()
+ elif selected_tts_server == "minimax-tts":
+ filtered_voices = voice.get_minimax_voices()
elif selected_tts_server == "elevenlabs":
# 音色列表位于 Key 输入框之前渲染,必须先统一恢复重连状态并读取
# 配置/环境变量,否则页面会用空 Key 加载并缓存空音色列表。
@@ -3333,6 +3349,8 @@ def _render_audio_settings(panel, params):
if voice.is_chatterbox_voice(v):
name = v.split(":", 1)[1] if ":" in v else v
return name.replace("-Female", "").replace("-Male", "")
+ if voice.is_minimax_voice(v):
+ return v.split(":", 1)[1]
return (
v.replace("Female", tr("Female"))
.replace("Male", tr("Male"))
@@ -3454,6 +3472,59 @@ def _render_audio_settings(panel, params):
_set_runtime_config("app", "mimo_api_key", mimo_api_key)
+ if tts_mode_enabled and (
+ selected_tts_server == "minimax-tts"
+ or (voice_name and voice.is_minimax_voice(voice_name))
+ ):
+ minimax_tts_api_key = st.text_input(
+ tr("MiniMax TTS API Key"),
+ value=config.minimax_tts.get("api_key", ""),
+ type="password",
+ key="minimax_tts_api_key_input",
+ )
+ _set_runtime_config("minimax_tts", "api_key", minimax_tts_api_key)
+
+ minimax_tts_endpoints = [
+ voice.MINIMAX_TTS_GLOBAL_URL,
+ voice.MINIMAX_TTS_CN_URL,
+ ]
+ configured_endpoint = config.minimax_tts.get(
+ "base_url", voice.MINIMAX_TTS_GLOBAL_URL
+ )
+ if configured_endpoint not in minimax_tts_endpoints:
+ configured_endpoint = voice.MINIMAX_TTS_GLOBAL_URL
+ minimax_tts_base_url = stable_selectbox(
+ tr("MiniMax TTS Endpoint"),
+ options=minimax_tts_endpoints,
+ default_value=configured_endpoint,
+ key="minimax_tts_endpoint_select",
+ )
+ _set_runtime_config("minimax_tts", "base_url", minimax_tts_base_url)
+
+ configured_model = config.minimax_tts.get(
+ "model_id", voice.MINIMAX_TTS_DEFAULT_MODEL
+ )
+ if configured_model not in voice.MINIMAX_TTS_MODELS:
+ configured_model = voice.MINIMAX_TTS_DEFAULT_MODEL
+ minimax_tts_model = stable_selectbox(
+ tr("MiniMax TTS Model"),
+ options=list(voice.MINIMAX_TTS_MODELS),
+ default_value=configured_model,
+ key="minimax_tts_model_select",
+ )
+ _set_runtime_config("minimax_tts", "model_id", minimax_tts_model)
+
+ minimax_tts_voice_id = st.text_input(
+ tr("MiniMax TTS Voice ID"),
+ value=config.minimax_tts.get(
+ "voice_id", voice.MINIMAX_TTS_DEFAULT_VOICE
+ ),
+ key="minimax_tts_voice_id_input",
+ )
+ _set_runtime_config(
+ "minimax_tts", "voice_id", minimax_tts_voice_id.strip()
+ )
+
# ElevenLabs API key section
if tts_mode_enabled and (
selected_tts_server == "elevenlabs"