Capture Devices

A capture device brings live video and audio into the show from an external source. The source can be an SDI capture device, an HDMI capture device, a webcam, a video-over-IP stream, or a stream from another application on the same node.

Unlike a video file, a capture is always live. It has no in-time, no out-time, no looping, and no scrubbing. It plays whatever is arriving now, at the lowest latency the source allows.

Workflow Overview

The end-to-end workflow has four steps:

  1. Add a capture device in the Devices window. Set its name, width, and height.
  2. Add a capture source for every node that needs to play this capture. For hardware inputs (SDI, HDMI, ST 2110, webcam) this means one source per node, each bound to that node's physical input.
  3. Place cues on timelines that reference the capture device.
  4. Reuse the same device across cues and timelines. All cues that bind to one capture device share a single live stream on each node.

Do not point two capture devices at the same source. Each capture device opens its own connection to the source. On most hardware (SDI, HDMI, ST 2110, webcam) the second one fails because the input is already in use. On NDI, you get duplicate network traffic and double CPU load for no gain. Use one capture device per source, and bind as many cues to it as you need.

To change a source after adding it, select it in the Devices window and use Edit Capture Source (or Edit NDI Source for an NDI source).

How Capture Devices Work

A capture device is a named container with a resolution (width × height) that holds one or more capture sources. Each source is one of the input types listed in Capture Types below — for example a webcam, an NDI® stream, or an SDI input.

Cues bind to the capture device, not to a specific source inside it. At playback, each node picks the right source from the device's list automatically.

Capture devices that span across outputs from multiple nodes need a source for each node. Hardware inputs (SDI, HDMI, ST 2110, webcam, Spout) live on a specific node — each source is tied to the node where the input is connected.

Example: an SDI splitter sends one camera into node A and node B. The show has one capture device with two sources, one per node. A cue spanning both nodes plays the local feed on each.

NDI sources are not tied to a node — every node reaches the same NDI sender over the network. One NDI source in the capture device's list is enough for all rendering nodes.

Capture Resolution

A source delivers frames at its own native resolution. WATCHOUT allocates the capture texture at that size and the renderer scales it onto the cue's quad with bilinear sampling. The device's configured width × height does not constrain the source — it only tells cues what size to expect on the Stage.

Capture Types

TypeLatencyWhen to use
Generic CaptureLowWebcams, USB capture cards, and other Windows-recognized capture hardware
NDILowLive video from cameras, switchers, or graphics systems on the LAN
Deltacast VideoMasterVery low (GPU-direct)Professional broadcast capture over SDI, HDMI, or ST 2110
SpoutSub-frame (shared GPU texture)Real-time graphics from another application running on the same node and the same GPU as WATCHOUT

Generic Capture uses Windows MediaFoundation under the hood — the same framework that exposes webcams, USB capture cards, and HDMI capture sticks to other Windows applications.

Audio capture is only supported on NDI sources. The other types pass video only.

When the capture source's frame rate does not match the show's frame rate, WATCHOUT blends between consecutive captured frames by default. Frame blending adds one frame of latency. Spout is the exception — it does not use frame blending.

Placing Capture Cues on Timelines

Drag a capture device from the Devices window onto a timeline or onto the Stage. WATCHOUT creates a cue that references the device. The cue's geometry on the Stage defaults to the capture's width and height.

Capture cues:

  • Are always live. They have no in-time, no out-time, no loop, no playback controls.
  • Update in real time as the source produces frames.
  • Cannot scrub or hold a frame.

For everything else — placement, effects, blending, condition expressions — capture cues behave like any other cue. See Adding Media Cues.

One Device, Many Cues

A capture device's live stream is opened once on each node and shared across every cue that uses that device. Adding more cues that reference the same capture device does not add network load, decoding cost, or hardware access — every cue reads from the same source.

Don't confuse this with the multi-source pattern. Multi-source distributes one capture device across multiple nodes — one source per node. Multi-cue lets many cues on one node read from a single shared live stream.

Capture Source Properties at a Glance

Per-technology pages cover the details. The common shape is:

PropertyApplies toWhat it controls
NodeAll hardware capturesWhich node has the input
Stream / DeviceAllThe named source (NDI stream name, Spout sender, Deltacast input, MF device name)
FormatHardware capturesPixel format (YUV / RGB / Auto). Lists what the device reports as supported.
RangeHardware capturesSignal range override (Auto / Limited / Full)
Color SpaceAllHow to interpret the incoming signal's colorimetry
Sync ModeAllSynchronises the presentation of multiple sources from the same physical device. Off — each source picks its own latency independently. Maintain — sources with the same frame rate and transport type share latency estimation, so genlocked inputs (multiple ports on the same board) stay frame-aligned with each other.
Audio ChannelsNDI only0, 1, 2, 4, or 8. Locked at creation time.
Audio LatencyNDI only, when audio channels > 0Audio buffer size: 20 / 40 / 80 / 100 / 200 / 1000 ms
Multicast Address / Port / SPSST 2110 onlyDestination addressing for the IP flow
SDPST 2110 onlyPaste or load an SDP file instead of typing addresses

Pixel Formats

The Format dropdown shows what the connected device or transport reports as available. WATCHOUT can decode the following pixel formats from capture sources:

FormatWhat it isWhere it appears
NV128-bit 4:2:0Most webcams, USB capture cards, HDMI capture sticks
YUY28-bit 4:2:2USB capture cards in 4:2:2 mode
UYVY8-bit 4:2:2NDI streams, Deltacast SDI / HDMI at 8-bit
UYVA8-bit 4:2:2 + alphaNDI streams that carry a key / alpha channel
V21010-bit 4:2:2Deltacast SDI / HDMI / ST 2110 at 10-bit
RGB248-bit 4:4:4HDMI capture in computer-source mode

Auto (the default) lets WATCHOUT pick the device's native format. Use Auto unless you have a specific reason to override — for example, forcing 10-bit V210 on a Deltacast input that supports both 8-bit and 10-bit.

Higher bit depth (V210 over 8-bit formats) gives smoother gradients and is required for HDR content. Higher chroma sampling (4:4:4 over 4:2:2 over 4:2:0) preserves more colour detail at the cost of bandwidth. See Chroma Subsampling in the glossary for the underlying concept.

Monitoring During Playback

The Node Info view in the Nodes window shows live capture status during playback: which source is active on each node for each capture device, its observed resolution and frame rate, and whether frames are arriving cleanly. Use this in rehearsal to confirm every node is locked onto its expected source.

NDI® is a registered trademark of Vizrt NDI AB.