✂️ VRGDG_PromptSplitterForManual
Preview one scene at a time, by hand
- text_output_1
- text_output_2
- text_output_3
- text_output_4
- text_output_5
- text_output_6
- text_output_7
- text_output_8
- text_output_9
- text_output_10
- text_output_11
- text_output_12
- text_output_13
- text_output_14
- text_output_15
- text_output_16
VRGDG_PromptSplitterForManual is the "point at a scene and look at it" node in VRGameDevGirl's splitter family. Same skeleton as the other index-based splitters - a JSON array in, an index to pick the scene, sixteen text outputs out - but its job is manual inspection: you dial in an index and read exactly what that scene's generation parameters are.
Where the ForFL and ForFMML variants are wired into the automated loop and driven by a stepper, this one is built for the moments when you want to look before you leap. The name is the clue: "manual." You're not rendering the whole song; you're checking scene 12 before committing hours of GPU time to it.
How it works
Two required inputs:
json_string- a JSON array of scene entries as text (defaults to[]). Element N is scene N.index- which scene to pull, 0 to 10000, zero-based.
Outputs: text_output_1 through text_output_16, all STRING - the selected scene unpacked across sixteen slots. The slot meanings aren't documented beyond the ports themselves; in the workflow they're whatever that section's video chain consumes.
The workflow fit
This is the debugging and QA surface for a music video render. The community's experience with this pack is that individual scenes need fiddling - people report that some scenes come out weak on character consistency or need reference images for locations and props. A node like this is how you isolate one scene's prompt, tweak it, and re-render just that chunk instead of the whole 30-scene batch. Set the index, read the outputs, edit the source map, re-check.
Gotchas
Because it's manual, the usual index mistakes apply hard: zero-based indexing (scene 1 is index: 0), and forgetting that the node reads from whatever JSON is in json_string - if you edited the prompt map upstream but didn't refresh this input, you're inspecting stale data. And unlike the automated variants, nothing advances this node's index for you; that's the point.
Installing it
With the pack:
cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl
restart, or ComfyUI Manager → search vrgamedev. No models, no heavy dependencies.
The honest take
If you're building your own workflow, this is arguably the most useful of the index splitters to have around, because previewing one scene before a long render is exactly the kind of thing ComfyUI makes tedious. Keep it on a tab next to your render queue and it'll earn its keep.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| json_string | STRING | [] | — |
| index | INT | 00–10000 | — |
Outputs (16)
| Name | Type | Description |
|---|---|---|
| text_output_1 | STRING | — |
| text_output_2 | STRING | — |
| text_output_3 | STRING | — |
| text_output_4 | STRING | — |
| text_output_5 | STRING | — |
| text_output_6 | STRING | — |
| text_output_7 | STRING | — |
| text_output_8 | STRING | — |
| text_output_9 | STRING | — |
| text_output_10 | STRING | — |
| text_output_11 | STRING | — |
| text_output_12 | STRING | — |
| text_output_13 | STRING | — |
| text_output_14 | STRING | — |
| text_output_15 | STRING | — |
| text_output_16 | STRING | — |