Node Types

GPU Config

A GPU Config node is one device used in one particular way. It names a detected device and holds everything about how that device is to be used: which hardware context, which encoders and decoders, which filters, and any settings those have of their own.

In Expert Mode the same device may appear more than once, each time as a separate node with its own configuration - which is what makes it possible to try a card one way first and a different way as a fallback. Each node carries the short name the graph shows it under, a letter for the vendor and a number, such as N10 for an Nvidia device.

Opening a node's configuration gives you:

Hardware Context - which context type to use, where the device offers more than one. An Nvidia card typically offers NvEnc and Vulkan; an Intel card can offer Quick Sync, Vulkan and VAAPI. The choice matters: contexts differ in performance, in which filters they support, and in maturity. Where a device offers only one, this is not shown. See Hardware Acceleration for what each vendor offers.

Encoding - which encoder to use for H.264, H.265 and AV1 output. Each has its own parameters, reachable from the button beside it, and an info button describing what the encoder reports about itself.

Filtering - which filter to use for deinterlacing, scaling, tone mapping and overlay, each with its own parameters. Outside Expert Mode only tone mapping is offered here.

Decoding - which decoder to use per source codec: H.264, H.265, AV1, MPEG-2, H.263, MPEG-4, VC-1, VP8 and VP9. Every one of these lists the software decoder for that codec alongside the hardware ones, so decoding in software while encoding on the GPU is a normal thing to configure.

Options - vendor-specific behaviour, where the context has any. Currently only the Nvidia NvEnc context does; see Nvidia.

Fallback point - whether evaluation ends successfully here. Only offered in Expert Mode, since it is the only mode where entries can be arranged, and a fallback point has to be the last entry in its group. See Fallback Paths.

Every encoder and decoder dropdown also carries a Skip entry. Selecting it opts this node out of handling that media type entirely, which is how you express "use this card, but not for AV1".

Anything you do not change stays on the device's automatic choice rather than being frozen at whatever was current when you configured it - see Delta Storage.

Software Config

A Software Config node is transcoding on the CPU, represented as a node so that it can take part in the graph like any other entry. In a default configuration it is the last entry, marked as the fallback point - which is what Emby's built-in software fallback amounts to. In Expert Mode neither of those is fixed.

It is configured the same way a GPU node is, with two differences. There is no Hardware Context to choose, software being the only one, and there are no vendor options. What remains is Encoding, Decoding and Filtering, populated with software codecs and filters - scaling, Yadif and Bwdif for deinterlacing, Tonemap and Supertonemap for tone mapping, and overlay.

Its decoding section differs from a GPU node's in one respect worth knowing: it offers every software decoder present, rather than a subset chosen for compatibility with the encoder. Software encoders and decoders are always mutually compatible, so there is nothing to narrow down.

Expert Mode allows more than one software node, which is what lets you configure software transcoding differently in different circumstances - one configuration for ordinary playback, another as a last resort after several hardware attempts have failed. In Basic and Advanced Mode there is exactly one, it sits at the end of the graph, and it cannot be moved or unmarked as the fallback point.

Condition Group

A condition group is merely a container which allows to apply the same conditions to a group of nodes. That group of nodes will be shown inside the container with a left-to-right order for visual clarity. Functionally, a Condition Group doesn't provide any extra value and is equivalent to having the same nodes in top-down order with the same set of conditions applied. In this regard, a Condition Group is rather a kind of visual/syntactical sugar for better clarity and simplified configuration, especially when considering that nodes inside a Condition Group can have their own conditions atop.

Load Balancing Group

With a Load Balancing Group, you can distribute transcoding load across multiple GPUs.

Details for setting up Load Balancing are described in Load Balancing.

Failure Point

You can insert a Failure Point wherever you want to stop any further transcoding attempts to be made. By default, a Failure Point hits in both passes (see Graph Evaluation). If you want a Failure Point to be effective only in fallback cases (PASS 2), you can set a condition on it for ErrorCount > 0.