Cine con IA · Prompt simple H3

Somebody on the internet wrote you a prompt. Maybe an LLM wrote it after you described the shot; maybe you copied a camera recipe or an entire six-section H3 prompt off a forum. You want to render it, not retype it into eight fields.

That's this node. It's a text box with two outputs, and it exists so the camera director has something to work on without forcing you into the modular scene builder.

What it does

You paste into texto. The node wraps your text in the same scene object the fancy scene node produces, tagging it as a raw prompt, and puts the text back out untouched. That's the whole function - no reformatting, no rewriting, no validation. Whatever you pasted is what goes to the model.

The two outputs are different things and that's the point:

• scene (CINECONIA_H3_SCENE) goes into Director de camara H3. The director detects the raw-prompt case and, instead of rebuilding a prompt from fields, edits yours in place: it looks for the detailed_description section and inserts the camera clause into it, after a [Shot 1] marker if one exists. If your text has no H3 sections at all, the clause is simply appended at the end. Your words survive either way, which is what you want when the text came from somewhere you trust.
• prompt (STRING) is the plain text, and it's the shortcut. If you don't want a camera clause - you're testing a model, checking a LoRA, reproducing someone's exact result - wire this straight into the pack's Escena node and skip the camera step entirely. One node, one wire, fewer places for something to go wrong.

When to reach for it

Use it for anything you didn't write yourself: LLM-drafted H3 prompts, verified camera recipes pasted whole, or prompts you're iterating on outside ComfyUI. The eight-field Escena por apartados H3 node is better when you're authoring the scene, because separate fields keep the character, the action and the setting from bleeding into each other - on current LLM-encoded video models, clean separation of parts is a real effect, not a formatting preference.

The simple node gives that up deliberately. It's the quick path, not the good one.

Inputs and outputs

One required input: texto, multiline, empty by default. Its tooltip explains both uses - paste a full prompt, or describe the scene in English, and then either connect scene to the camera director or use prompt directly. Outputs, as above: scene and prompt.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/chaLords/ComfyUI-Cine-con-IA.git

Restart and it's "Prompt simple H3" under Cine con IA/H3. No Python dependencies are installed by the pack, and a recent ComfyUI with native MiniMax H3 support is the real prerequisite. ComfyUI Manager may not surface it: the pack is registered as cine-con-ia in the Comfy Registry with its versions still pending review, so clone it or use the release ZIP. The UI, including this node's single tooltip, is in Spanish.

Two honest warnings

Nothing is validated, so a half-finished prompt renders exactly as badly as it reads. That's the trade for verbatim preservation, and it's why the pack puts the camera director between this node and the sampler rather than trusting the text.

And check which output you're using. If you wire prompt into the scene/conditioning path and also expect camera controls to do something, you'll wonder why the movement you picked never showed up - it only exists on the scene → director path. Wire prompt when you want the text as written; wire scene when you want it directed.