Apppublic

uzbtrust/morfem-navoiy-tts

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes

App README

Morfem Navoiy TTS

Uzbek speech synthesis service for the Morfem application.

—Navoiy checkpoint: aisha-org/navoiy-tts at 912e6bb6663f11e5dec5de830ca06ea8dd4c985f
—Base model: FunAudioLLM/CosyVoice2-0.5B at eec1ae6c79877dbd9379285cf8789c9e0879293d
—CosyVoice runtime: commit 074ca6dc9e80a2f424f1f74b48bdd7d3fea531cc
—API: /synthesize(text, emotion, speed) -> audio

The model is loaded once through ZeroGPU's CUDA emulation and GPU inference is serialized to avoid memory spikes. The service uses CosyVoice's bundled reference prompt; it does not accept user voice uploads.

cosyvoice_runtime.zip is an inference-only bundle of the Apache-2.0 upstream runtime pinned to commit 074ca6d. Its SHA-256 checksum is verified before extraction.

No token is stored in this repository. If Hub authentication is ever required, add HF_TOKEN through the Space settings.