Create a custom voice from an authorized speech sample, then synthesize scripts for narration, spoken presentations, or product promotion. Cloud providers generate mp3 or wav audio using supplied credentials, without requiring a local GPU.
上传本人语音样本克隆专属音色，再用它合成口播、旁白或带货语音。当用户说“声音克隆、克隆/复刻我的声音、用我的声音配音、定制专属音色”时使用，需要用户自备云端 provider 凭证。使用公共现成音色时改用 tts-voiceover。
Review the instructions the agent will receive before you install.
用本人语音样本克隆音色，再合成任意文案。走云端 provider（用户自备 key），本地无需 GPU。 全部走
skills/shared/scripts/voice_clone.py。
不想克隆、用现成公共音色见 tts-voiceover（edge-tts，免费无需 key）； AI 生成音乐/BGM 见 ai-music；合成后与 BGM 混音见 audio-mix。
配置检查路径铁律：先
cd到运行时说明给出的 Easel 项目根，确认当前目录有.env和skills/shared/scripts/，再运行注册表、check、enroll或clone。不得改用 workspace 的共享脚本副本，也不得以env/printenv没显示变量为由判断VOICE_BASE_URL/Key 缺失。check支持时显式传--env-file .env。
选 provider 并在 .env 填 key，再 check 离线校验：
python skills/shared/scripts/voice_clone.py check --provider minimax --env-file .env
| provider | 服务 | .env 需配 |
|---|---|---|
dashscope | 阿里 CosyVoice 声音复刻 | DASHSCOPE_API_KEY（可选 DASHSCOPE_TTS_MODEL/DASHSCOPE_BASE_URL） |
minimax | MiniMax 语音克隆 | MINIMAX_API_KEY、MINIMAX_GROUP_ID（可选 MINIMAX_MODEL） |
fish-audio | Fish Audio | FISH_API_KEY（可选 FISH_BASE_URL） |
openai-compatible | OpenAI 兼容 /audio/speech | VOICE_API_KEY、VOICE_BASE_URL（预置 voice，非零样本克隆） |
gemini | Google Gemini TTS | GEMINI_API_KEY（可选 GEMINI_TTS_MODEL/GEMINI_VOICE/GEMINI_BASE_URL） |
⚠️ 各 provider 依公开 API 文档实现，端点/模型名可用 env 覆盖以适配实际参数。
执行前先跑 model_registry.py configured --group voice --env-file .env。只有一个可用时显式选择；多个可用且用户没点名时，列出 provider/模型询问本次使用哪个，不按 VOICE_PROVIDER 擅自选择。
| 字段 | 必填 | 说明 |
|---|---|---|
| 语音样本 | 克隆时必填 | 本人清晰无噪的语音（一般 10s-1min，具体看 provider 要求） |
| 文案 | 合成时必填 | 要用克隆音色说出来的文字 |
outputs/主题名/）脚本路径（相对项目根）：skills/shared/scripts/voice_clone.py（各子命令支持 -h）。
# minimax：上传样本文件
python skills/shared/scripts/voice_clone.py enroll --provider minimax \
--sample me.mp3 --name my_voice
# dashscope：用公网可访问的样本 URL
python skills/shared/scripts/voice_clone.py enroll --provider dashscope \
--sample-url https://.../me.wav --name myv
（fish-audio 用已有 model_id 或内联参考音频，openai-compatible 用预置 voice 名，无需 enroll。）
python skills/shared/scripts/voice_clone.py clone --provider minimax \
--voice-id my_voice --text "大家好，欢迎来到我的频道" --speed 1.0 \
-o outputs/主题名/vo.mp3
fish-audio 也可直接给参考音频：--sample ref.mp3 --sample-text "参考音频的文字"。
check 确认 key，再 enroll，再 clone。outputs/主题名/。主流声音克隆云服务：阿里 CosyVoice（声音复刻）、MiniMax（语音克隆 + T2A）、Fish Audio、 OpenAI 兼容 TTS。本地零样本克隆（GPT-SoVITS/CosyVoice 本地）需 GPU，故走云端 provider。
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Generate images from written descriptions, transform existing images with editing instructions, or create several variations. Choose image dimensions and quantity, then save the results using an OpenAI-compatible service or apimart with your own API key.
Generate background music for short videos, vlogs, and social posts from descriptions of mood, style, and instruments. Download MP3 audio, optionally include sung lyrics, and use shared scripts to trim tracks, adjust loudness, or add music to videos.
Create new videos from written scene descriptions or still images using external AI services. Set the aspect ratio and duration, animate a character image, and request generated audio when the selected model supports it.
Organize generated images, videos, and text in outputs/ by date, platform, and type. Add tags, search filenames and content, produce asset lists and reports, and preview archive moves before applying them to make past work easier to reuse.
Reduce background noise in audio recordings or video soundtracks with three processing levels. Keep the original file, create a cleaned version, and receive a report of the settings used and file details before and after processing.
Edit audio files by trimming sections, converting formats, matching loudness, or extracting sound from videos. You can also join recordings, add fades, and change playback speed while keeping the original pitch.