Text to Speech
Free online text-to-speech tool: natural main engine (CosyVoice 3) + emotional voice engine (IndexTTS-2), supports multiple voices, adjustable speed/pitch/volume, multilingual, emotional styles, voice cloning, and multi-speaker dialogue. Export MP3/WAV and SRT subtitles. GPU self-deployed for privacy protection.
风格/情感对「声音克隆」和「自定义参考音色」效果最佳；预置命名音色以音色本身为准。
FAQ
Powered by self-deployed open-source dual engines CosyVoice 3 (natural main) and IndexTTS-2 (emotional voice), completely free. Offers daily free synthesis quota with no login required.
"Natural Main" (CosyVoice 3) offers more natural, fluent speech and faster speed, with multiple preset voices—ideal for daily reading, customer service, and announcements; "Emotional Voice" (IndexTTS-2) delivers stronger, more dramatic expression—perfect for audiobooks, character voice acting, and video narration. Switch between them with one click at the top of the page.
Yes. Select "Voice Cloning", upload a clear 5–15 second audio sample, enter corresponding text, and generate speech in your voice. Both engines support voice cloning. Reference audio is used only for this session and deleted immediately after, never stored.
Supports Chinese, English, Japanese, Korean, Cantonese, and cross-language synthesis. The main engine allows natural language instructions in the "Style" field, such as "speak happily" or "in Sichuan dialect"; switching to the "Emotional Voice" mode lets you input emotions like "happy", "sad", "angry", or "surprised" directly for stronger emotional expression.
GPU deployment ensures fast processing; first use or idle periods require initial model loading (a few seconds for main engine, 20–40 seconds for emotional engine). Long texts are synthesized sentence by sentence and streamed back—listen to completed parts while waiting.
Yes. Enable "Multi-Speaker", input each character followed by their lines in separate lines, then assign different voices to each speaker.