Text-to-speech

NeuTTS Air requirements, license & speed

By Neuphonic · released 2025-09-15 · ~552M (emb + active), ~360M active parameters

Hugging Face ↗ GitHub ↗ Compare

Commercial useYes
Min memoryNot stated
Runs on CPUYes
StreamingYes

Key facts

License
apache-2.0 source
Commercial use
Yes source
Air is Apache-2.0; newer Nano variants use NeuTTS Open License 1.0. HF repo gated
Parameters
~552M (emb + active), ~360M active source
Minimum memory
Not stated by source
Runs on CPU
Yes source
CPU benchmarks published (Ryzen 9HX 370, M4, Galaxy A25)
Streaming
Yes source
Streaming via GGUF backbones + llama-cpp-python
Voice cloning
Yes source
From ~3 s of audio
Languages
en source
Released
2025-09-15 source
Hugging Face repo creation date (proxy for release date)
Gated on HF
Yes accept terms on Hugging Face before downloading

Measured speed & memory

Benchmark pending. We haven't measured NeuTTS Air yet. Planned targets: CPU (8 vCPU), Apple Silicon Mac, RTX-class consumer GPU and a rented cloud GPU. Get notified.

Similar TTS models

Spotted an error? Report it. Every fact links to the model card or repo it came from.

New open speech models, tested. Weekly.

One email when a model ships: license, memory, and real speed on a Mac, a gaming GPU and a cloud GPU.