APNext H3 Scene Pick

APNext H3 Scene Pick is the smallest node in the H3 family and one of the most useful, because it solves a practical pain: the Crossover Writer and Scenes Writer hand you scenes as a list, and whatever video node you plug into the other end runs once per element. Ten scenes, ten renders, one queue - which is exactly what you want for a final batch, and exactly the wrong thing when you're still iterating on scene three and don't want to burn a whole queue (and a chunk of your Claude Code quota) to look at it.

This node collapses the list down to one element. Feed it the scenes list, set index to the scene you care about, and it returns that single scene's prompt plus its duration. The index is 0-based and clamped, so index=3 past the end quietly gives you the last scene instead of erroring out - handy when a writer decides to trim the run to five scenes mid-project.

The three inputs

• scenes - the list output of an H3 Crossover Writer or Scenes Writer (really, any STRING list).
• index - which scene to pull, 0-based. Default 0.
• durations (optional) - the matching list; the picked scene's duration comes out the other side so your frame-count math stays in sync.

Outputs: scene (the prompt text), duration, the resolved index, and count (how many scenes were in the list).

How people actually use it

The classic pattern from the pack's own examples/h3/ files: fix the writer's seed so the scene list is reproducible, wire scenes/durations into Scene Pick, and put an incrementing primitive on index. Each queue run renders one scene; step index up and you walk through the story one clip at a time, fixing problems as you find them. Then, when everything's right, swap Scene Pick out for a direct wire to the video node and render the full batch.

You can also wire scene into the H3 Refiner to polish a single prompt, or into a prompt preview to eyeball it before committing. Pair it with H3 Scenes Join on the batch side and you've got both workflows - iterate one, deliver all of them stitched into a single file.

Installing

Part of dagthomas/comfyui_dagthomas ("SDXL Auto Prompter"). ComfyUI Manager → search comfyui_dagthomas → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/dagthomas/comfyui_dagthomas
cd comfyui_dagthomas
pip install -r requirements.txt

Restart, and find it under APNext → H3. No API key, no model download - it's pure plumbing.

One thing to keep in mind: Scene Pick selects from whatever the writer produced on the last run. Change the writer's seed or direction and re-run before you trust the pick, or you'll be rendering a stale scene and wondering where your edits went.