Nvidia Broadcast V1.0.0.25 Now
Unlike Chroma key (green screen), NVIDIA Broadcast’s semantic segmentation model identified human silhouettes without color requirements. In v1.0.0.25, the segmentation operated at ½ resolution to maintain real-time performance (up to 30 FPS). The output supported:
Notably, version 1.0.0.25 did not support background video replacement; that arrived in v1.1.x.
Three distinct effects were available per input device: Nvidia Broadcast V1.0.0.25
| Feature | Input Type | AI Model Type | Output Resolution* | |---------|------------|---------------|--------------------| | Noise Removal | Microphone | Recurrent Neural Network (RNN) | n/a (audio only) | | Room Echo Removal | Microphone | Convolutional RNN | n/a | | Virtual Background | Camera | Semantic segmentation (U-Net) | Up to 1080p@30fps | | Auto Frame | Camera | Face+pose detection | Cropped 720p/1080p | | Noise Removal | Speaker | RNN | n/a |
*Effective resolution after processing.
No software is perfect. Here are the common bugs associated with Nvidia Broadcast V1.0.0.25 and how to fix them:
Issue 1: "This application requires an Nvidia RTX GPU" Notably, version 1
Issue 2: Microphone stops working after PC sleeps
Issue 3: High CPU usage (20%+)

