nari-labs

/

Dia-1.6B

like 2.51k

Follow

Nari Labs 1.32k

I’m a huge fan of your Dia-1.6B TTS model. I'm currently running quantized LLMs like Gemma3 and Qwen2.5 on-device using the Google AI Edge Gallery, which supports local, offline inference in .task or .gguf format.

I’d love to be able to run Dia-1.6B in the same app, but the current model weights (e.g. .safetensors, .pth) are too large and not supported directly by Edge Gallery.