🎙️ SFlowTTS Inference Playground

ℹ️ About Me & The Model

Hi there! I am the developer behind SFlowTTS. This model is using a 44.1 kHz FSQ vocoder combined with a discrete Flow-Matching architecture. This 2-stage architecture has ~200M parameters. I trained the entire model within 7 days on 2x H100

Language

Speaker ID

Select a Story

Text to synthesize

Multiply Duration (Length Scale)

0.5 2

Flow Matching Temperature

0.1 2

Duration Noise Scale

0.1 2

Inference Steps

2 200

CFG Strength

0.1 10

Generated Audio

Generation St1ats