Indic Parler Nepali TTS - Khusi Production Checkpoint
This is a local fine-tuned Indic Parler-TTS checkpoint for Nepali TTS.
Important inference note: this checkpoint was trained with Khusi audio while preserving the original known-working Binisha-style text description label. For best results, use the Binisha wording in the voice description and provide Nepali text sentence-by-sentence or use the included local runtime that splits text automatically.
Recommended description:
Binisha voice is warm, clear and natural in Nepali, spoken by a female speaker. The speech sounds like audiobook narration with a close microphone and very clear audio, with almost no background noise. The delivery is natural and steady.
Long multi-sentence prompts may produce quiet/silent output. Split on Nepali danda (।) and synthesize chunks separately.
