This comparison workflow pits three guidance modes of the MiniMax H3 Max model against each other: pure text-to-video, first-frame-to-video (animate from a single image), and first-to-last-frame-to-video (FLF2V). All three paths use the same base model but differ in how much visual constraint you provide. The text-only path is driven by your prompt (via MinimaxHailuo03ContextIRNode), the first-frame path anchors the opening composition, and the FLF2V path uses both the first and last frames (MinimaxHailuo03FirstLastFrameNode) to tightly direct motion, continuity, and lighting from start to finish. The graph also includes an optional Prompt Enhancer stage and an optional 2K upscale branch controlled by ComfySwitchNode and PrimitiveBoolean flags, so you can compare raw vs enhanced prompts and base vs upscaled output in one place.
Methodologically, you run multiple passes with a shared prompt and seed, changing only the guidance mode and/or feature flags. SaveVideo captures each render, while SaveText logs the active settings (prompt, flags, and mode) so you can verify that only a single variable changed between runs. If you enable MinimaxHailuo03RegenerateNode, you can produce controlled alternates—holding the same inputs constant—so differences are attributable to sampling rather than setup drift. Compare the results for prompt faithfulness, motion between keyframes, lighting stability, composition cleanliness, and edge fidelity. For convenience, try the included example assets (dream_face_first.png and dream_face_last.png) to see how first-only vs FLF2V affects transitions.
API
Usa este flujo de trabajo desde código
Cada flujo de trabajo de Comfy es un grafo JSON. El siguiente payload es este flujo de trabajo, exactamente como lo ejecuta ComfyUI: descárgalo desde la URL, guárdalo en control de versiones o cárgalo en ComfyUI y ejecútalo nodo por nodo.
Ejecuta desde TypeScript o Python con el Comfy SDK. El mismo código funciona en Comfy Cloud o en un ComfyUI propio: solo cambia la URL base.
// Install (beta)
npm i @comfyorg/sdk
// Run this workflow (TypeScript)
import { Comfy } from "@comfyorg/sdk";
const client = new Comfy({ apiKey: "comfyui-..." });
const wf = await client.workflows.fromFile("workflow_api.json");
const job = await client.run(wf);
await job.getOutputs("<output-node-id>")[0].toFile("output.png");El SDK utiliza un flujo de trabajo en formato API: abre este flujo de trabajo en ComfyUI y usa Archivo → Exportar flujo de trabajo (API).
El acceso a la API de Comfy Cloud requiere un plan con clave API.
