# Direct GPU Image Sharing

Real-time OpenFX signal processing just became ultra-efficient on macOS:

* **Zero-latency GPU image sharing**: The OpenFX plugin now supports **direct GPU memory transfer**, removing costly CPU-GPU copies and enabling **instant signal handoff** to OmniScope.
* **Significantly reduced resource usage**: The plugin has been optimized to use **far less CPU and GPU**, especially when monitoring live Resolve timelines.
* For maximum efficiency, you can also **scale down the input resolution** in the plugin settings, ideal for heavy projects or low-power systems.

### How to enable Direct GPU Sharing

In OpenFX settings, enable "Direct GPU Sharing" option:<br>

<figure><img src="https://565437294-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MFqn5MuayPgQPMZ9pwf%2Fuploads%2F8VedeTDlv6JkGbFdKsff%2FCleanShot%202025-10-01%20at%2010.55.25%402x.png?alt=media&#x26;token=292fdf53-0761-4d5b-83db-946c6040ffc6" alt=""><figcaption></figcaption></figure>

Then, after couple of seconds, Nobe OmniScope should auto-detect new **GPU Share Source** named "**NobeOmniScope Metal Output**":

<figure><img src="https://565437294-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MFqn5MuayPgQPMZ9pwf%2Fuploads%2FQWVw5qQoPpvTvcg6pXLZ%2FCleanShot%202025-10-01%20at%2010.55.39%402x.png?alt=media&#x26;token=f7773c6e-c676-4cb3-96ee-53fb8ee45c0e" alt=""><figcaption></figcaption></figure>

After selecting this source the signal should now be shared directly through GPU with near zero latency.