NeuTTS-Nano (BF16 GGUF)

This repository contains the official bf16 GGUF quantization of the NeuTTS-Nano acoustic inference model. Optimized specifically for low-latency, high-efficiency edge execution using llama.cpp or compatible GGUF runtimes.

Model Details

• Architecture: NeuTTS-Nano
• Format: GGUF (outtype: bf16)
• File Size: 446 MB
• Precision: Brain Floating Point 16-bit (bf16)
• Primary Use Case: Real-time acoustic processing, voice generation, low-latency edge speech pipelines.

Deployment & Inference

To run this model locally via the CLI or spin up a local inference server using llama.cpp, utilize the following configurations.

CLI Benchmark

llama-cli -m neutts-nano-bf16.gguf -p "Your input text prompt here" --threads 4

Local Inference Server

llama-server -m neutts-nano-bf16.gguf --port 8080 --host 0.0.0.0

Language Variants

If you are looking for specific regional acoustic profiles, distinct variants are isolated in their own standalone repositories:

• Core/English: undefinelabs/neutts-nano
• French: undefinelabs/neutts-nano-french
• German: undefinelabs/neutts-nano-german
• Spanish: undefinelabs/neutts-nano-spanish