Download
1 variant available
fp32 SafeTensor
Dainiaus_vokalas_step700.safetensors
Full precision, largest file • 203.04 MB
Verified: 10 minutes ago
A synthetic male vocal model, trained using 5 different male voices. It is a versatile, trilingual LoRa model capable of singing in Lithuanian, Russian, and English. Tested styles include pop, disco fox, rock, rock 'n' roll, Russian rock, country rock, and country Americana. Trained for 800 steps.
Keywords embedded in the LoRa include [dainius_vocal], "solo male vocal," "deep baritone," and "dramatic tone." Other descriptors also work, such as "clear tone," "crisp vocal," "crisp tone," "smoky vocal," "raspy vocal," and "rough vocal." (It is recommended to include "male" in these prompts—e.g., "raspy male vocal").
Sample video: rock music in three languages.

