MakHamTTS Paradee ONNX

Standalone, single-voice Thai ONNX text-to-speech model derived from Kokoro-82M and Paradee, featuring the makham_female acoustic timbre.

license: apache-2.0

Artifacts in this Directory

• paradee_thai/makham_female_thai_int8.onnx: 82.7 MB weight-only quantized ONNX model (recommended).
• paradee_thai/makham_female_thai.onnx: 325.5 MB FP32 full-precision ONNX model.
• paradee_thai/config.json: 115-token vocab configuration for 5-tone Thai phonemes.

Model Interface

• Inputs:
  • input_ids: [1, tokens] (int64)
  • speed: [1] (float32)
• Outputs:
  • waveform: [1, samples] (float32, 24 kHz)

No external ref_s or .npy voicepack is required.

See https://github.com/wannaphong/MakHamTTS for running TTS model.

fastthaig2p: https://github.com/awslabs/FastThaiG2P/