ROIpad ← Back to Search

消费级显卡（比如5090/4090等）下的RTF统计

k2-fsa/OmniVoice

Status: Open

Opened: Apr 3, 2026

Comments: 5

原始Pytorch模型下大家大概是多少？

Python

View on GitHub ↗

Other Comments / Reviews

For RTF evaluation, with different GPUs, inference steps,...

by zhu-han Apr 4, 2026
344秒时长音频耗时51秒 RTF=0.15 测试方法： 1）自定义一个...

by cacard Apr 3, 2026
> [@cacard](https://github.com/cacard) what's your config...

by cacard Apr 3, 2026
@cacard what's your config? I only got RTF = 0.3 on 3090 ...

by rennyka-107 Apr 3, 2026
生成14秒音频平均1.12秒，RTF = 0.08，不错了。（on 24G VRAM...

by cacard Apr 3, 2026

Related Content

AI Insight OmniVoice's Real-Time Factor (RTF) performance on consumer-grade GPUs (e.g., 5090/4090). The user is inquiring about typical RTF statistics.

This inquiry into 'RTF statistics on consumer-grade GPUs' (e.g., 5090/4090) f...