Best Open-Source Text-to-Speech Models

Open-source TTS has caught up enough that you can self-host a voice assistant or audiobook narrator without the cloud. These are the most capable models, ranked by community adoption.

  1. 1

    Kokoro 82M TTS

    0.082B params

    High quality 82M parameter TTS model. Excellent speech synthesis with multiple voice options. 86MB download.

    Min VRAM: 0.58GBQuant: ONNX-Q8F16Size: 0.08GBLicense: apache-2.0
  2. 2

    Piper TTS - Amy (English)

    0.02B params

    Lightweight TTS voice. High quality English speech synthesis. Default TTS model - runs on any iPhone. Only 63MB.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  3. 3

    Piper TTS - Lessac (English)

    0.02B params

    High quality English male voice. 63MB download. Runs on any device.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  4. 4

    Piper TTS - LibriTTS-R (English)

    0.02B params

    Medium quality English voice with natural prosody. 63MB download.

    Min VRAM: 0.57GBQuant: ONNXSize: 0.073GBLicense: mit
  5. 5

    Piper TTS - Spanish (MLS)

    0.02B params

    Spanish female voice. Natural prosody.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  6. 6

    Piper TTS - French (Siwis)

    0.02B params

    French female voice.

    Min VRAM: 0.53GBQuant: ONNXSize: 0.026GBLicense: mit
  7. 7

    Piper TTS - German (Thorsten)

    0.02B params

    German male voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  8. 8

    Piper TTS - Chinese (Huayan)

    0.02B params

    Chinese Mandarin voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  9. 9

    Piper TTS - Japanese (Kokoro)

    0.02B params

    Japanese voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  10. 10

    Piper TTS - Korean

    0.02B params

    Korean voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  11. 11

    Piper TTS - Russian (Irina)

    0.02B params

    Russian female voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  12. 12

    Piper TTS - Portuguese (Faber)

    0.02B params

    Portuguese voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit
  13. 13

    Piper TTS - Italian (Riccardo)

    0.02B params

    Italian male voice.

    Min VRAM: 0.53GBQuant: ONNXSize: 0.026GBLicense: mit
  14. 14

    Piper TTS - Arabic (Kareem)

    0.02B params

    Arabic voice.

    Min VRAM: 0.15GBQuant: ONNXSize: 0.063GBLicense: mit

Related

All TTS models