# MiniMax Image Edit Before to After

A ComfyUI workflow for transforming a single input image from a “before” state into an “after” state using MiniMax H3 FL2V.

## Features

- Before and after states controlled by text inputs

- Fixed four-shot prompt structure

- Adjustable aspect ratio and target resolution

- 5-second output at 5 fps / 25 frames

- Automatic last-frame extraction and PNG preview

## Inputs

- Image: Select the source image before running the workflow.

- Before text: Describes the Shot 1 state.

- After text: Describes the Shot 4 state.

- Aspect ratio: Select the target aspect ratio, such as 1:1, 16:9, or 9:16.

- Megapixels: Sets the target resolution scale.

- Multiple: Controls resolution rounding for the H3-compatible dimensions.

## Default Settings

- Aspect ratio: 1:1 (Square)

- Resolution: 512 × 512

- Duration: 5 seconds

- Frame rate: 5 fps

- Frames: 25

- Sampling steps: 8

- LoRA: MiniMax H3 Turbo 4-step

## Prompt Structure

```text

Spot the Difference

[Shot 1] {before_text}

[Shot 2] A momentary blackout

[Shot 3] The lights come back on

[Shot 4] {after_text}

```

Only Shot 1 and Shot 4 are user-editable. Shot 2 and Shot 3 remain fixed so the workflow consistently describes a before-to-after transition.

## Output

The workflow creates a 5-second MP4 at 5 fps with 25 frames. It also extracts the final frame and saves it as a PNG with a preview.

## Required Models

- minimax_h3_fl2va_pruned_int8_convrot.safetensors

- qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors

- minimax_h3_video_vae_fp16.safetensors

- minimax_h3_turbo_4step.safetensors

Download the required models separately and follow their respective licenses. Model weights are intentionally not included in this package.