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.
Python ai-video-generatorcontent-creationffmpeginstagram-reelsllmpythonshort-videosubtitlestext-to-speechtiktokvideo-automationvideo-workflowworkflow-automationyoutube-shorts
feat(tts): add Chatterbox TTS provider (self-hosted, OpenAI-compatible)
Paulo C. Benetti023922ea2775
diff --git a/app/config/config.py b/app/config/config.pyindex 72210eef..1ace0877 100644--- a/app/config/config.py+++ b/app/config/config.py@@ -153,6 +153,7 @@ def save_config():_cfg["azure"] = azure_cfg["siliconflow"] = siliconflow_cfg["elevenlabs"] = elevenlabs+ _cfg["chatterbox"] = chatterbox_cfg["ui"] = uif.write(toml.dumps(_cfg))@@ -164,6 +165,7 @@ proxy = _cfg.get("proxy", {})azure = _cfg.get("azure", {})siliconflow = _cfg.get("siliconflow", {})elevenlabs = _cfg.get("elevenlabs", {})+chatterbox = _cfg.get("chatterbox", {})ui = _cfg.get("ui",{diff --git a/app/services/voice.py b/app/services/voice.pyindex 63780a06..190fa546 100644--- a/app/services/voice.py+++ b/app/services/voice.py@@ -164,6 +164,29 @@ def get_elevenlabs_voices(api_key: str) -> list[str]:return []+def get_chatterbox_voices() -> list[str]:+ """Return the configured Chatterbox voices.++ Chatterbox is self-hosted, so there is no global voice catalog. Operators+ list the voice names exposed by their server via ``[chatterbox] voices``+ (a TOML array, or a comma-separated string). Each entry is normalised to+ the ``chatterbox:<name>`` format used by the TTS dispatcher.+ """+ voices = config.chatterbox.get("voices", []) or []+ if isinstance(voices, str):+ voices = [v.strip() for v in voices.split(",") if v.strip()]+ result = []+ for v in voices:+ v = str(v).strip()+ if not v:+ continue+ result.append(v if v.startswith("chatterbox:") else f"chatterbox:{v}")+ if not result:+ # keep the dropdown usable even before any voice is configured+ result = ["chatterbox:default-Female"]+ return result++_AZURE_VOICES_DATA_FILE = os.path.join(os.path.dirname(__file__), "data", "azure_voices.json")@@ -229,6 +252,10 @@ def is_elevenlabs_voice(voice_name: str) -> bool:return (voice_name or "").startswith("elevenlabs:")+def is_chatterbox_voice(voice_name: str) -> bool:+ return (voice_name or "").startswith("chatterbox:")++def is_no_voice(voice_name: str | None) -> bool:"""判断用户是否明确选择了“无配音”模式。@@ -398,6 +425,19 @@ def tts(else:logger.error(f"Invalid elevenlabs voice name format: {voice_name}")return None+ elif is_chatterbox_voice(voice_name):+ # 格式: chatterbox:<voice>,voice 可带显示用的 -Female/-Male 后缀+ parts = voice_name.split(":", 1)+ if len(parts) >= 2 and parts[1].strip():+ chatterbox_voice = parts[1].strip()+ if chatterbox_voice.endswith(("-Female", "-Male")):+ chatterbox_voice = chatterbox_voice.rsplit("-", 1)[0]+ return chatterbox_tts(+ text, chatterbox_voice, voice_file, voice_rate, voice_volume+ )+ else:+ logger.error(f"Invalid chatterbox voice name format: {voice_name}")+ return Nonereturn azure_tts_v1(text, voice_name, voice_rate, voice_file)@@ -1269,6 +1309,89 @@ def elevenlabs_tts(return None+def chatterbox_tts(+ text: str,+ voice: str,+ voice_file: str,+ voice_rate: float = 1.0,+ voice_volume: float = 1.0,+ model_id: str = "",+) -> Union[SubMaker, None]:+ """Generate speech with a self-hosted Chatterbox TTS server.++ Chatterbox (Resemble AI, MIT) is an open-source, locally hosted TTS model+ with zero-shot voice cloning — a self-hostable alternative to ElevenLabs.+ This talks to an OpenAI-compatible ``/audio/speech`` endpoint, so it works+ with the common community servers (e.g. devnen/Chatterbox-TTS-Server,+ travisvn/chatterbox-tts-api). Configure ``[chatterbox] base_url`` (and an+ optional ``api_key``).++ Like ElevenLabs, Chatterbox does not return word-level timestamps, so the+ subtitle path falls back to the full-text SubMaker. For tighter subtitle+ sync set ``subtitle_provider = "whisper"``.+ """+ text = (text or "").strip()+ if not text:+ logger.error("Chatterbox TTS text is empty")+ return None++ base_url = (config.chatterbox.get("base_url", "") or "").strip().rstrip("/")+ if not base_url:+ logger.error(+ "Chatterbox base_url is not set, please configure [chatterbox] base_url in config.toml"+ )+ return None++ api_key = config.chatterbox.get("api_key", "")+ if not model_id:+ model_id = config.chatterbox.get("model_id", "chatterbox") or "chatterbox"++ url = f"{base_url}/audio/speech"+ headers = {"Content-Type": "application/json"}+ if api_key:+ headers["Authorization"] = f"Bearer {api_key}"+ payload = {+ "model": model_id,+ "input": text,+ "voice": voice,+ "response_format": "mp3",+ # OpenAI speech API accepts speed 0.25-4.0; MoneyPrinterTurbo's rate is a+ # 1.0-centred multiplier, so it maps directly (clamped to the valid range).+ "speed": max(0.25, min(4.0, float(voice_rate or 1.0))),+ }++ for i in range(3):+ try:+ logger.info(f"start chatterbox tts, voice: {voice}, try: {i + 1}")+ ensure_file_path_exists(voice_file)++ response = requests.post(url, json=payload, headers=headers, timeout=120)+ if response.status_code != 200:+ logger.error(+ f"chatterbox tts failed with status {response.status_code}: {response.text[:200]}"+ )+ continue++ with open(voice_file, "wb") as f:+ f.write(response.content)++ audio_clip = AudioFileClip(voice_file)+ audio_duration = audio_clip.duration+ audio_clip.close()++ sub_maker = ensure_legacy_submaker_fields(SubMaker())+ logger.success(f"chatterbox tts succeeded: {voice_file}")+ return populate_legacy_submaker_with_full_text(+ sub_maker=sub_maker,+ text=text,+ audio_duration_seconds=audio_duration,+ )+ except Exception as e:+ logger.error(f"chatterbox tts failed: {str(e)}")++ return None++def _format_text(text: str) -> str:"""清理字幕对齐前的脚本文本。diff --git a/config.example.toml b/config.example.tomlindex 486cf5c1..f3bcb251 100644--- a/config.example.toml+++ b/config.example.toml@@ -397,3 +397,19 @@ upload_post_auto_upload = false# YouTube Shorts privacy status (options: "public", "unlisted", "private")upload_post_youtube_privacy_status = "public"++# Chatterbox TTS (self-hosted, OpenAI-compatible) ##########################+# Chatterbox (https://github.com/resemble-ai/chatterbox, MIT) is an open-source,+# locally hosted TTS model with zero-shot voice cloning — a self-hostable+# alternative to ElevenLabs. Run a server that exposes an OpenAI-compatible+# endpoint at POST {base_url}/audio/speech (e.g. devnen/Chatterbox-TTS-Server+# or travisvn/chatterbox-tts-api), then set base_url below. In the WebUI pick+# "Chatterbox TTS" as the TTS server, or via the API pass a voice_name like+# "chatterbox:<voice>". Subtitles use the full-text fallback (no word-level+# timestamps); set subtitle_provider = "whisper" for tighter sync.+[chatterbox]+# base_url = "http://localhost:4123/v1"+# api_key = ""+# model_id = "chatterbox"+# Voices exposed by your server; shown in the WebUI dropdown as chatterbox:<name>+# voices = ["Francisca-Female", "Antonio-Male"]diff --git a/webui/Main.py b/webui/Main.pyindex b9ed18d9..cddc26d1 100644--- a/webui/Main.py+++ b/webui/Main.py@@ -964,6 +964,7 @@ with middle_panel:("gemini-tts", "Google Gemini TTS"),("mimo-tts", "Xiaomi MiMo TTS"),("elevenlabs", "ElevenLabs TTS"),+ ("chatterbox", "Chatterbox TTS"),]# 获取保存的TTS服务器,默认为v1@@ -1016,6 +1017,9 @@ with middle_panel:saved_elevenlabs_api_key)filtered_voices = st.session_state[cache_key]+ elif selected_tts_server == "chatterbox":+ # 自托管 Chatterbox 服务的预置音色(来自 [chatterbox] voices 配置)+ filtered_voices = voice.get_chatterbox_voices()else:# 获取Azure的声音列表all_voices = voice.get_all_azure_voices(filter_locals=None)@@ -1038,6 +1042,9 @@ with middle_panel:if voice.is_elevenlabs_voice(v):parts = v.split(":", 2)return parts[2] if len(parts) >= 3 else v+ if voice.is_chatterbox_voice(v):+ name = v.split(":", 1)[1] if ":" in v else v+ return name.replace("-Female", "").replace("-Male", "")return (v.replace("Female", tr("Female")).replace("Male", tr("Male"))