🎤 Fast Read

Neural TTS — Kokoro · Silero · Magpie (NVIDIA)

Kokoro
🏴️ EN
🇬🇧 GB
🇪🇸 ES
🇮🇹 IT
🇯🇵 JA
🇰🇷 KO
🇨🇳 ZH
🇮🇳 HI
🇧🇷 PT
Silero v5
🇷🇺 RU
🇺🇿 UZ
Magpie (NVIDIA — RTX 3060 Ti)
🇫🇷 FR
🇩🇪 DE
🇻🇳 VI
Speed 1.00
Kokoro ONNX

🎓 Expert Mode

Fine-tune voice parameters. Controls marked GPU only work with Magpie (French, German, Vietnamese).

📢 Speed All engines

How fast the voice talks. 1.0 is normal human pace. Scroll the slider or drag it.

0.5 slow 1.0 normal 1.5 fast 2.0 rush

🌡️ Temperature Magpie GPU

Controls emotion & expressiveness. Think of it like "dial up the feeling." Too high and the voice gets weird/unstable.

0.3 flat 0.5 calm 0.7 natural 1.0 expressive 1.3+ unstable

🎯 CFG (Stay-in-character) Magpie GPU

ON: Voice stays close to the speaker's original recorded style — consistent, predictable, good for long reads.

OFF: More free variation — sometimes more expressive, less consistent. Good for short creative snippets.

🎤 Voices All engines

Magpie has 5 speakers: Sofia & Aria (female), Jason, Leo, John (male). All speak FR/DE/VI.

Kokoro has 11 EN voices, regional ES/IT/JA/etc. Silero covers RU & UZ.

Magpie Controls — active
0.70
Expressiveness (0.7 = natural)