Notch Troubleshooting
Most problems with Notch blocks on Disguise media servers come down to a handful of causes: a runtime version mismatch, a missing playback license, a block that has not been discovered, or content that has not synced to the render nodes. This page lists the common symptoms and how to resolve them. For the full setup workflow, see RenderStream with Notch.
Block fails to load or crashes on load
Section titled “Block fails to load or crashes on load”The most frequent cause is a version mismatch between the Notch runtime that built the block and the runtime installed on the machine trying to play it.
- The Notch runtime version must match across every machine involved: Designer, Notch Host and all render nodes.
- A block exported from a newer version of Notch than the installed runtime will not load.
Confirm the same Notch version is installed everywhere, then re-export the block from a matching version if needed.
Block plays with a watermark or refuses to render
Section titled “Block plays with a watermark or refuses to render”A valid Notch playback license must be present on the server GPU. Without one, the block either fails to render or plays with a watermark.
- Check that the playback license is installed and active on the machine hosting the block.
- On a freshly imaged render node, confirm the license before assuming a content fault.
Block does not appear in the asset list
Section titled “Block does not appear in the asset list”Asset discovery only scans the RenderStream Projects location, so a block stored anywhere else will never appear.
- Place blocks under
C:\Users\USERNAME\Documents\Renderstream Projects, orD:\Renderstream Projectson an RX hardware system. - The discovery system scans all sub-directories, so a
Notchsub-folder is fine.
A block cannot be split across render nodes
Section titled “A block cannot be split across render nodes”A block with no exposed camera cannot be split across a cluster. If you assign it to multiple render nodes, the content is duplicated on each node rather than divided between them.
- The camera must be exposed in Notch Builder before the block is exported. The same applies to any parameters, lights or nulls you expect to control from Designer.
- Re-expose the camera in Notch Builder, re-export the block, then re-sync it to the cluster.
For the authoring steps, see the official Notch guidance on exposed parameters and cameras.
Content is out of date or missing on render nodes
Section titled “Content is out of date or missing on render nodes”If nodes render stale or blank content, the block has not synced across the cluster.
- Check the Source Machine in the asset editor points to the machine that holds the current block.
- Press Sync and confirm every entry in Sync tasks is marked completed before starting the workload.
Streams are not received or the workload will not start
Section titled “Streams are not received or the workload will not start”If the Workload status does not switch to Running, or Designer does not receive the expected streams:
- Confirm every entry in Sync tasks is marked completed before starting the workload. An incomplete sync will stall the start.
- Check that Notch Host is installed on every machine in the cluster, including render nodes. It ships with the main Disguise installation, but the versions must match across the network.
- Keep the cluster to a single Disguise product range. Mixing machines from different ranges, for example RX with VX, is untested and unsupported.
- Changes to video transport settings require restarting the workload before they take effect.
Poor performance or unexpectedly high GPU memory use
Section titled “Poor performance or unexpectedly high GPU memory use”- Creating multiple channel mappings against the same channel spawns a separate block instance for each mapping, so the scene renders multiple times. Reduce duplicate mappings where the extra instances are not needed.
- Heavy blocks can exhaust GPU memory. Check VRAM headroom on the render node and simplify the block if it is close to the limit.
Further help
Section titled “Further help”For issues specific to running ONNX AI models in Notch, see the troubleshooting section of ONNX AI Models in Notch.
For general Notch use inside Designer, see the Notch layer. Notch 2026.2 also introduces the Notch Manual MCP Server, which exposes the official Notch manual through Anthropic’s Model Context Protocol (MCP). You can connect an AI assistant such as Claude, ChatGPT or Cursor to it to query verified Notch documentation directly.