OBS Studio 32.2.2 NVENC and Cloud OBS Upgrade Checklist

The direct answer

OBS Studio 32.2.2 is the current OBS release on September 3, 2026, and it is not a casual update for producer-driven shows. The official download page lists version 32.2.2 as released on August 14, 2026. The GitHub release notes add two immediate operational warnings: OBS 32.2.x updates the NVIDIA SDK to version 13, which raises the minimum supported NVIDIA driver to 570, and macOS 12 is no longer supported in 32.2.x.

The hotfix itself is useful. OBS says 32.2.2 fixes plugins not loading properly on the first start after updating on Windows. The 32.2.1 hotfix fixed game capture failing after an update if the previous hook was still in use. Underneath those hotfixes, the 32.2 line also adds dynamic bitrate support to multitrack video, an SDR-to-HDR compose filter, better FPS selector UX, and several frontend and plugin-facing changes that matter to anyone running remote production or automation.

If you use StreamableRun with Cloud Hosted OBS, the safe move is to upgrade in a rehearsal window and treat 32.2.2 like a live system change. Check the GPU driver first, keep macOS 12 machines on 32.1.2, retest plugins and WebSocket controls, then run the real Cloud OBS scene package through a private destination before touching the public stream.

Check the driver and operating system before you open a scene

The NVIDIA requirement is the first gate, not the last. OBS explicitly says the 32.2 release updates to NVIDIA SDK 13 and requires driver version 570 or newer. If your workflow depends on NVENC for local contribution, local recording, or backup encoding, do not wait until the encoder dropdown acts weird. Verify the driver before the update and write down which machine is responsible for which encode job.

The macOS warning is just as important. The 32.2.x line no longer supports macOS 12 because of the Qt update. OBS says macOS 12 users should stay on 32.1.2. That means mixed fleets need a simple rule. Do not upgrade every laptop just because one Windows box is ready. Producer machines, backup laptops, and emergency local OBS boxes should all be checked against the OS they actually run.

This sounds obvious, but real stream teams get burned by support drift. The main operator updates their machine and forgets the emergency backup laptop in the backpack. Then the backup cannot load the same plugins or scenes during a real failure. Write the platform map before you do the install.

• Confirm NVIDIA driver 570+ on every Windows machine that may use NVENC.
• Keep macOS 12 systems on OBS 32.1.2.
• Check the backup laptop, not just the main production machine.
• Write down which machine owns local encode, local record, and emergency fallback.
• Do not trust auto-update prompts as your deployment plan.

What changed in the 32.2 line that producers should care about

The 32.2 line is worth attention even if your scene collection looks stable. OBS lists new features including dynamic bitrate support for multitrack video, an SDR-into-HDR compose filter, a redesigned add-source dialog, missing-file support for filters, and plugin-facing API additions such as custom icons for source types. Those are not just nice-to-have changes. They affect how quickly an operator can identify sources, recover broken assets, and keep color and bitrate behavior predictable.

OBS also notes a fix for audio mixer state getting out of sync when settings are changed through WebSocket or plugins. That matters for remote teams more than solo desk streamers. If a producer uses Companion, a web panel, Stream Deck macros, or custom tooling, a mixer-state mismatch is the kind of bug that creates confident but wrong decisions during a noisy live recovery.

The practical conclusion is not that everyone should fear the release. It is that teams using remote control, multiple scenes, or browser-source-heavy shows should give 32.2.2 a real staging run instead of a lunch-break update.

Retest plugins, scripts, and WebSocket control together

OBS's developer guide is still the cleanest summary of extension risk. OBS can be extended through plugins, scripts, and WebSocket, and each one can break differently. Plugins may need rebuilds or may fail to load. Scripts may survive but expose unexpected source behavior. WebSocket panels may still connect while the changed scene, source, or audio behavior underneath them makes the button result wrong.

The 32.2.2 hotfix specifically mentions Windows plugin loading on first start after an update. That should tell you exactly what to test: first boot after install, first scene load, and first time the tools that matter get exercised. Do not just confirm that OBS launches. Confirm that the specific plugins, docks, panels, and remote buttons your team relies on are actually available and act the way the runbook expects.

For StreamableRun users, the important split is this: local OBS tools can still be flaky while Cloud Hosted OBS stays stable. That means the producer should rehearse how to move the public show to a safe scene if a local plugin or script does something dumb. The upgrade plan should assume the cloud layer is the buffer between software churn and the viewer.

• Test first launch after update on Windows before show day.
• Open every plugin dock or panel your team depends on.
• Run the actual scene-switch and mute macros, not just a connection test.
• Confirm WebSocket controls against the live scene collection.
• Keep a manual fallback plan if a plugin path fails.

Browser sources and missing assets need their own pass

The 32.2 line adds missing-file support for filters, which is useful, but it is not a reason to stop testing overlays. Cloud OBS and local OBS shows often rely on browser sources, media sources, HTML overlays, alerts, clips, and sponsor graphics that only reveal problems when the exact scene gets triggered. A silent fallback scene with the wrong asset is still a failure.

Open the real scenes. Trigger the alert source. Load the lower third. Move to the BRB scene. Hide and reveal the browser source that carries the graphic. The useful question is not whether the overlay exists in a list. It is whether the operator can rely on it during a transition or a recovery. If a filter or source points at a missing file, find that out before the show hits the paid segment.

This is one place where StreamableRun can reduce pain. Keep the viewer-facing fallback scenes and destination layer separate from the local OBS assets that may have changed. If a local scene behaves badly after the update, the cloud producer can stay on a safe scene while the local operator repairs the asset path.

Run the private Cloud OBS drill, not just a desktop preview

A desktop preview is not an upgrade test. Use a named StreamableRun ingest, send it into Cloud Hosted OBS, and start a private destination. Then rehearse the specific jobs that break when OBS changes under you: scene switching, NVENC contribution, audio mutes, plugin-driven actions, browser-source overlays, backup scene activation, and return to main after a reconnect.

Make somebody watch from the viewer side. If your workflow uses local OBS for contribution and StreamableRun for the public production layer, the viewer preview tells you whether the whole chain stayed sane. A clean local canvas does not prove that the private destination got the right audio or that the fallback returned correctly after a source issue.

Also test the boring edges. Restart OBS once. Reconnect the source once. Open the same scene after the restart. Use the exact backup scene you plan to use in public. Upgrades break boring repeatability more often than dramatic one-shot features.

• Use a private destination first.
• Restart OBS once after the update and repeat the core actions.
• Test scene switching from both the local UI and the remote control path.
• Confirm the viewer preview, not only the local preview.
• Practice a safe cutaway if the updated local OBS path stops behaving.

Producer handoff rules for 32.2.2

A producer should not have to reverse-engineer the upgrade state during a show. Put the facts in the handoff note: OBS version, whether the machine is on NVIDIA driver 570 or newer, which plugins were validated, which controls were retested, and whether the machine is allowed to be the main contribution source or only a backup.

Keep the handoff short and useful. A remote producer needs to know what they can trust. If a scene-switch panel was not retested, say that. If the backup laptop is stuck on macOS 12 and OBS 32.1.2 by design, say that too. Clarity beats pretending every machine is identical.

When the runbook is clear, StreamableRun becomes the operational safety net instead of a mystery box. The producer knows when to rely on Cloud OBS, when to request a local restart, and when to stay on fallback until the viewer preview proves the updated path is clean.

Where StreamableRun fits

StreamableRun's public features page is useful here because it centers the operational pieces that matter during an OBS upgrade: remote OBS access, stream continuity, multiple destinations, and the ability to switch between desktop and IRL sources without ending the stream. Those are exactly the properties that let you test local OBS changes without turning every mistake into a public outage.

If the 32.2.2 update exposes a plugin or local encoding problem, the cloud producer can keep the stream alive with fallback scenes or another ingest while the local machine gets repaired. That is much better than letting the update happen directly on the only box that viewers depend on.

OBS 32.2.2 is a useful release. It just deserves grown-up handling. Update the driver, respect the macOS 12 boundary, retest the extension stack, and use StreamableRun as the buffer between local software churn and the public show.