Face Fix - Load Anchors (Meta Batch)

Your enhanced anchor faces are sitting on disk in a known folder. Now you need them back in the graph, loaded as a batch, in the same deterministic order they were saved - and you want it to happen without manually pointing a file picker at a folder every run. That's the job of VRGDGFaceFixLoadAnchorsMetaBatch: it loads the prepared anchors through VideoHelperSuite's Meta Batch system, so loading behaves exactly like the tested workflow expects and you never touch the folder selector again.

How it works. Two inputs, both non-negotiable:

• face_fix_context - from Prepare Video and Anchors. This carries the automatically generated anchor-source folder, which is the entire trick: the node doesn't ask you where the anchors live because Prepare already decided, so a folder path you set once by hand becomes fully automatic.
• meta_batch - a VHS Batch Manager connection. The tooltip spells out the setup: set its frames_per_batch to Prepare's anchor_count, so all anchors follow the same managed loading behavior the tested Z-Image workflow uses. This is the pack leaning on VideoHelperSuite's battle-tested batching instead of reinventing it.

The node then outputs the anchors as a proper IMAGE batch with an accompanying mask, plus batch_frame_count and the face_fix_context carried forward. That batch is what feeds Z-Image for enhancement - and it's the exact batch whose output Store Enhanced Anchors will later compare against for count and order.

Why Meta Batch. Loading a folder full of images in ComfyUI is usually a "Load Images From Folder" node with a hand-picked path, and it breaks the moment your workflow runs twice and the folder moves. Meta Batch is VideoHelperSuite's way of managing batches across the graph - frames_per_batch, frame counts, predictable iteration - and wiring the Face Fix anchors into it means the whole "which anchors, how many, in what order" question is handled by the same machinery that loads the rest of your video, rather than a one-off loader.

The outputs. anchor_images (the IMAGE batch for Z-Image), mask, batch_frame_count, and face_fix_context (the handoff token the rest of the pipeline threads through every node).

Installing. In the vrgamedev pack:

cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl

or install via ComfyUI Manager - and you must also have VideoHelperSuite (VHS) installed, since this node's whole existence is a VHS integration. If VHS isn't present, the VHS_BatchManager port won't even appear.

Troubleshooting. The classic mistake is not setting frames_per_batch on the Batch Manager to match anchor_count - the loader and the manager disagree, and you get a truncated or padded batch that then fails the count check at Store Anchors. Also remember this node loads anchors; it doesn't know they're enhanced or not, so make sure you're loading the enhanced-anchor folder in the right workflow stage. If the context output from Prepare isn't flowing, verify Prepare ran successfully first - a failed detection pass means no anchor folder at all.