Verify Jellyfin NVENC with a deliberately transcoded sample

The useful distinctions

• Jellyfin documents NVENC/NVDEC acceleration on supported NVIDIA GPUs.
• Its Windows verification procedure deliberately lowers web-client resolution or bitrate to trigger video transcoding.
• Windows GPU engine views distinguish Video Decode and Video Encode activity.

Create a request that needs work

A GPU check needs a playback request that actually exercises the component under investigation. If the client receives the original video unchanged, an idle video encoder is unsurprising. Choose a small sample with known properties and deliberately request a lower representation. Keep the source and requested output in your notes so the observation has a clear meaning after the session ends.

Observe the relevant engines

General graphics activity can come from unrelated work, and a single percentage is a poor description of a conversion pipeline. Watch the named video engines during the same bounded request and keep the server’s conversion log as the technical record. The example is a worksheet for collecting that evidence; it does not claim that any real GPU, codec combination or particular driver has been tested here.

Keep conclusions narrow

A successful sample shows that this request used the configured path under those conditions. It does not certify every codec, subtitle combination or high-resolution workload. Likewise, an unsuccessful sample needs the actual FFmpeg error before a driver change is justified. Leave operating-system installation and hardware eligibility checks to their own documented procedures, then repeat the same sample after a specific correction so the comparison remains useful.

A focused way to proceed

1. Choose an owned local sample and note its source codec and dimensions.
2. Play it in Jellyfin Web and lower the selected resolution or bitrate to request video transcoding.
3. During that request, inspect Video Decode and Video Encode in the Windows GPU performance view.
4. Collect the corresponding FFmpeg log if the result does not match the intended acceleration path.

Worked example · synthetic

Sample: Workshop Colour Bars.mkv
Source: record codec and dimensions
Requested output: select a lower web-client quality
Evidence: Video Decode, Video Encode, matching FFmpeg log
Conclusion: limited to this request

Read the comparison

Idle encoding during direct playback is not evidence of an acceleration failure.

When to stop or ask for help

Escalate a failed forced transcode with the FFmpeg log, GPU model and installed versions; do not infer unsupported codecs are supported.