Audio model · dia
DI Dia 1.6B requirements
Text to speech (dialogue) · 1.6B params · fp16 · released Apr 2025. Realistic minimum to run: Nvidia GeForce RTX 3060 (12GB).
Run it
Dia 1.6B runs in PyTorch (CUDA) or nari-labs/dia at fp16. It needs a GPU (or Apple Silicon with enough unified memory).
Which devices can run Dia 1.6B?
Apple Silicon Macs
No mainstream local runtime for Dia 1.6B on Apple Silicon Macs yet.
RAM-only laptops
No mainstream local runtime for Dia 1.6B on RAM-only laptops yet.
iPhone & iPad
No mainstream local runtime for Dia 1.6B on iPhone & iPad yet.
Android
No mainstream local runtime for Dia 1.6B on Android yet.
NVIDIA GPUs
AMD GPUs
No mainstream local runtime for Dia 1.6B on AMD GPUs yet.
FAQ
How much memory does Dia 1.6B need?
At fp16 it consumes ~10 GB. It needs a GPU or Apple Silicon.
Can Dia 1.6B run on a phone or CPU?
Not practically. Dia 1.6B expects a GPU; CPU inference is too slow for real use.
Can I use Dia 1.6B commercially?
Yes. Dia 1.6B is licensed Apache-2.0, which permits commercial use.
Nari Labs' dialogue TTS (1.6B). The model card states ~10GB VRAM peak (steady ~7.4GB); an RTX 3060 12GB user reported OOM at peak. CPU support is listed as not-yet-added, and no Apple Silicon or AMD path is confirmed, so it is NVIDIA-only for now. Apache-2.0, commercial OK. Sources: HF model card, GitHub OOM report.
Sources
Memory is a sourced peak-usage anchor at fp16, validated 2026-06-15. See methodology.