R42 Delivery Preview · First / Middle / Last
Preview a long H3 delivery without spawning 400 temp files
- delivery_frames
- preview_frames
- preview_report
You know the small annoyance: your long-video graph finishes, and to look at the result you wire the full IMAGE batch into a preview node. For a multi-segment H3 delivery that batch is hundreds of decoded frames, and ComfyUI dutifully writes all of them to temp files so it can show you a grid of thumbnails you will never scroll through.
IAMCCS_MiniMaxH3DeliveryPreviewStripR42 is the two-line fix. It takes the batch and returns three frames: the first, the middle and the last. That's it. That's the node.
Inputs and outputs
Required: delivery_frames (the IMAGE batch), current_segment, total_segments (both forced inputs), resolved_path (a STRING, forced input - the master file path when one exists) and delivery_report.
Outputs: preview_frames (three images) and preview_report, which also prints into the node's UI text area so you can read it on the canvas without opening a separate preview.
The three indexes are computed as a set - {0, count // 2, count - 1} - so a one-frame batch gives you one frame rather than three copies of it. An empty batch raises instead of silently passing nothing on.
The report is the useful half
R42 DELIVERY MASTER PREVIEW | segment=5/5 | source_frames=612 | shown=0,306,611 | path=/output/.../final.mp4 | <delivery report>
Two things it tells you that a thumbnail can't. First, the scope: if the current segment is the last one and a path resolved, it says DELIVERY MASTER; otherwise CURRENT CHUNK. So while a queued run is grinding through segments you can tell at a glance whether you're looking at an intermediate or the finished thing. Second, it preserves the segment provenance and the upstream delivery report in one string, which is the difference between "this looks wrong" and "this looks wrong and here's which of the eleven segments it came from".
IS_CHANGED returns NaN on purpose. Automatic segment queues reuse the same graph and the same node ids, so without that the cached preview would keep showing you the previous segment's frames.
Where it goes
It's a leaf node in practice. Put it after your delivery node and read the three frames plus the report, while your exporter or save node carries on doing its job from the full batch. It changes nothing about the render - the frames it selects are copies of references inside the existing batch, not a re-decode.
Install
Standard pack install:
cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes.git
Or Manager → search IAMCCS, then restart ComfyUI. No models, no optional dependencies.
Worth saying out loud because it looks like dead weight in a graph: this node exists specifically because previewing a long H3 delivery the obvious way is expensive. If you find yourself adding it to every workflow, that's not bloat - it's the correct instinct, and the alternative is temp-file churn on every segment.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| delivery_frames | IMAGE | — | |
| current_segment | INT | — | |
| total_segments | INT | — | |
| resolved_path | STRING | — | |
| delivery_report | STRING | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| preview_frames | IMAGE | — |
| preview_report | STRING | — |