Hailuo AIGC

A powerful Dify plugin providing comprehensive AI-powered image and video generation capabilities using MiniMax Hailuo models. Supports text-to-image, image-to-image with subject reference, text-to-video, image-to-video, first-last frame video, subject reference video, MiniMax-H3 multimodal reference, prompt enhance, 768P-to-2K regeneration, and video result query with optional download.

Version Information

Current Version: v0.0.2
Release Date: 2026-08-9
Compatibility: Dify Plugin Framework
Python Version: 3.12

Version History

v0.0.2 (2026-08-9): Added MiniMax-H3 video generation (V2 API) for text-to-video, image-to-video, and first-last frame video; added multimodal reference, H3-Context-IR prompt enhance, and 768P-to-2K regeneration; video query now supports both Hailuo V1 and MiniMax-H3 V2 tasks
v0.0.1 (2026-02-22): Initial release with image and video generation capabilities

Quick Start

Install the plugin in your Dify environment
Configure your MiniMax API credentials (API Key)
Start generating images and videos with AI

Key Features

Multiple Generation Modes: Text-to-image, image-to-image, text-to-video, image-to-video, first-last frame video, subject reference video, MiniMax-H3 multimodal reference
Latest AI Models: Supports MiniMax-H3, MiniMax-Hailuo-2.3, MiniMax-Hailuo-2.3-Fast, MiniMax-Hailuo-02, T2V-01-Director, I2V-01-Director, I2V-01-live, I2V-01, S2V-01, image-01, image-01-live
MiniMax-H3 2K Output: Generate 768P or 2K videos, duration 4-15 seconds
Prompt Enhance: Use H3-Context-IR to expand a video prompt before generation
Video Regeneration: Upscale a MiniMax-H3 768P video to 2K
Flexible Image Ratios: Multiple aspect ratios from 1:1 to 21:9
Video Generation: Task-based asynchronous generation with status query and download
First-Last Frame Video: Create videos from first and last frame images
Subject Reference Video: Create videos using subject reference images
Watermark Control: Optional AIGC watermark for content authenticity

Core Features

Image Generation

Text to Image (text_2_image)

Generate images from text descriptions using MiniMax image models.

Supported Models: image-01, image-01-live
Features:
- Multiple aspect ratios (1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9)
- Optional watermark
- Optional prompt optimizer
- Style options for image-01-live

Image to Image (image_2_image)

Generate images from text and a subject reference image.

Supported Models: image-01, image-01-live
Features:
- Subject reference guidance (character)
- Multiple aspect ratios
- Optional watermark
- Reference image (jpeg/png/webp; max 10MB)

Video Generation

Text to Video (text_2_video)

Create a video generation task from text.

Supported Models: MiniMax-H3, MiniMax-Hailuo-2.3, MiniMax-Hailuo-02, T2V-01-Director, T2V-01
Features:
- Task-based async generation
- MiniMax-H3: 768P / 2K, duration 4-15s, required aspect ratio
- Older models: 720P / 768P / 1080P
- Optional prompt optimizer and fast pretreatment (older Hailuo models)

Image to Video (image_2_video)

Create a video generation task from a single image.

Supported Models: MiniMax-H3, MiniMax-Hailuo-2.3, MiniMax-Hailuo-2.3-Fast, MiniMax-Hailuo-02, I2V-01-Director, I2V-01-live, I2V-01
Features:
- Single image input as first or last frame (MiniMax-H3)
- MiniMax-H3 prompt is required; output 768P / 2K
- Older models: 512P / 720P / 768P / 1080P

First-Last Frame Video (images_2_video)

Create a video generation task from first and last frame images.

Supported Models: MiniMax-H3, MiniMax-Hailuo-02
Features:
- First and last frame inputs
- MiniMax-H3: 768P / 2K, duration 4-15s
- MiniMax-Hailuo-02: 768P / 1080P

Subject Reference Video (subject_reference_2_video)

Create a video generation task using a subject reference image.

Supported Models: S2V-01
Features:
- Subject reference guidance (character)
- Single reference image input

MiniMax-H3 Reference to Video (reference_2_video)

Create a MiniMax-H3 video from text plus reference images, videos, or audio.

Supported Models: MiniMax-H3
Features:
- Up to 9 images, 3 videos, and 3 audios (12 files total)
- Output 768P / 2K, duration 4-15s
- Adaptive or fixed aspect ratio

MiniMax-H3 Prompt Enhance (h3_context_ir)

Expand a video prompt with MiniMax H3-Context-IR. This does not generate a video.

Query the task later and read

MiniMax-H3 Video Regeneration (video_regeneration)

Regenerate a MiniMax-H3 768P video into 2K.

Use the original task ID, or resubmit the 768P video plus the original prompt

Video Result Query (video_query)

Query video generation task status and optionally download the video.

Features:
- Supports both Hailuo V1 tasks and MiniMax-H3 V2 tasks
- MiniMax-H3 success returns a direct download URL
- H3-Context-IR success returns the enhanced prompt
- Optional file download when a video is ready

Requirements

Python 3.12
Dify Platform access
MiniMax API credentials (API Key)
Required Python packages (installed via requirements.txt):
- dify_plugin>=0.2.0
- requests>=2.31.0,<3.0.0
- pillow>=10.0.0,<11.0.0

Installation & Configuration

Install the required dependencies:
Configure your MiniMax API credentials in the plugin settings:
- API Key: Your MiniMax API key
- Get API key: https://platform.minimaxi.com/user-center/basic-information/interface-key
Install the plugin in your Dify environment

Usage

Image Generation Tools

1. Text to Image

Parameters:
- : Text description of the image (required)
- : image-01 / image-01-live
- : 1:1, 16:9, 4:3, 3:2, 2:3, 3:4, 9:16, 21:9
- : url / base64
- : 1-9
- : true / false
- : true / false
- / (image-01-live only)

2. Image to Image

Parameters:
- : Text description (required)
- : Subject reference image (required)
- : image-01 / image-01-live
- : 1:1, 16:9, 4:3, 3:2, 2:3, 3:4, 9:16, 21:9
- : url / base64
- : 1-9
- : true / false
- : true / false

Video Generation Tools

3. Text to Video

Parameters:
- : Text description (required)
- : MiniMax-H3 / MiniMax-Hailuo-2.3 / MiniMax-Hailuo-02 / T2V-01-Director / T2V-01
- : 720P / 768P / 1080P / 2K (2K is MiniMax-H3 only)
- : MiniMax-H3 uses 4-15; older models use 6 or 10
- : Required for MiniMax-H3 text-to-video (16:9, 9:16, 1:1, 4:3, 3:4, 21:9)
- : true / false (older models)
- : true / false (Hailuo 2.3/02 only)
- : true / false

4. Image to Video

Parameters:
- : Input image (required)
- : Text description (required for MiniMax-H3)
- : MiniMax-H3 / MiniMax-Hailuo-2.3 / MiniMax-Hailuo-2.3-Fast / MiniMax-Hailuo-02 / I2V-01-Director / I2V-01-live / I2V-01
- : 512P / 720P / 768P / 1080P / 2K
- : MiniMax-H3 uses 4-15; older models use 6 or 10
- : MiniMax-H3 first_frame or last_frame
- : true / false
- : true / false (Hailuo 2.3/02 only)
- : true / false

5. First-Last Frame Video

Parameters:
- : First frame image (required)
- : Last frame image (required)
- : Text description (required for MiniMax-H3)
- : MiniMax-H3 / MiniMax-Hailuo-02
- : 768P / 1080P / 2K
- : MiniMax-H3 uses 4-15; MiniMax-Hailuo-02 uses 6 or 10
- : true / false
- : true / false

6. Subject Reference Video

Parameters:
- : Subject reference image (required)
- : Optional text description
- : S2V-01
- : true / false
- : true / false

7. MiniMax-H3 Reference to Video

Parameters:
- : Text description (required)
- / / : Optional files
- / / : Optional public URLs
- : 4-15 seconds
- : 768P / 2K
- : adaptive or a fixed ratio
- : true / false

8. MiniMax-H3 Prompt Enhance

Parameters:
- : Original description (required)
- Optional first/last frame or reference media
- : 4-15 seconds
- Query the task later to read the enhanced prompt

9. MiniMax-H3 Video Regeneration

Parameters:
- : Successful MiniMax-H3 768P task ID
- or plus the original final prompt
- Do not provide both modes at once

10. Video Result Query

Parameters:
- : Video generation task ID (required)
- : Download video file when available (default: true)

Notes

Video generation is asynchronous; use Video Result Query to check status and retrieve results
MiniMax-H3 uses a new V2 API. Select MiniMax-H3 in the existing video tools, or use the dedicated H3 reference / prompt enhance / regeneration tools
MiniMax-H3 text-to-video requires an aspect ratio and cannot use adaptive
MiniMax-H3 prompt is required for all generation modes
First/last frame video on older models uses MiniMax-Hailuo-02 only
Subject reference video currently supports a single subject image (S2V-01). MiniMax-H3 multimodal reference uses the H3 Reference to Video tool
MiniMax-H3 prefers public media URLs; uploaded files are sent as Base64 when no URL is available
Reference size limits: image <= 30MB, video <= 50MB, audio <= 15MB

Developer Information

Author:
Email: [email protected]
License: Apache License 2.0
Source Code:
Support: Through Dify platform and GitHub Issues

License Notice

This project is licensed under Apache License 2.0. See LICENSE file for full license text.

Ready to create stunning images and videos with AI?

Development Notes

MiniMax-H3 uses the new API with a multimodal payload. It is not compatible with the older Hailuo API. This plugin routes to V2 and keeps older models on V1. Video query tries V2 first, then falls back to V1. Prefer public media URLs because the V2 request body limit is 64MB.