I've been testing F5 TTS AI for cloning a Spanish voice sample, but the output sounds robotic compared to demos I've seen. Wondering if this is a known limitation with non-English voices or if I'm missing something with my audio samples. Has anyone had better results with specific languages or sample lengths?
I haven't tested F5 TTS specifically, but voice cloning quality often depends heavily on sample audio clarity and length—try using 30-60 seconds of clean, professional speech without background noise. Have you experimented with different sample formats or checked if F5 has language-specific tuning parameters? Also worth asking their community directly about Spanish-specific benchmarks, as some TTS models do perform better on certain languages than others.