MiniMax H3 Fast AI Video Generator

Describe the short video you want, choose 5–15 seconds, and create it with MiniMax H3 Fast at 480P. For a video made from words alone, choose a horizontal or vertical shape such as 16:9 or 9:16. You can also provide a supported picture or video clip as a guide.

Your AI

Use with your AI

For tasks with several steps, continue in the AI you already use.

CodexClaude CodeCursorWindsurfCline

View instructions

Use Toolmu Tool minimax-video-h3-fast-minimax, ID b0a83c9f-8042-44ed-a8f9-962f69680952. Reuse the current Toolmu connection, or follow https://toolmu.com/SKILL.md if needed. Read this exact Tool’s current contract, availability and credit ceiling; check identity, balance and the user’s established budget in the same environment. Collect the video description, integer duration of 5–15 seconds and desired mode. Set model to MiniMax-H3-Fast and resolution to 480P. Include exactly one nonempty text item. Text-only requests require an explicit non-adaptive ratio. Keep first/last-frame images separate from reference-image/video/audio inputs; a last frame requires a first frame, and audio cannot be the sole reference. Validate the selected media’s formats, sizes and lengths. Use only the supported settings. Execute the user’s requested generation with one persisted idempotency key and a maxCredits bound within the authorized budget. Recover the same run when submission is unconfirmed; do not buy a replacement generation. Report the actual returned video result and confirmed charge; preserve missing fields and do not claim permanent storage, exact reference matching or reviewed audiovisual quality.

Task:
Create a [5-second] video showing [a paper boat drifting across a calm pond]. Have the camera [follow the boat slowly]. Use a [wide horizontal view].

Paste into your connected AI's conversation.

Connection guide

Up to 1.848 credits / call

The final charge is based on confirmed usage. Unused reserved credits return to your available balance.

MiniMax

MiniMax H3 Fast AI Video Generator

You receive

Create a 5–15-second, 480P MiniMax H3 Fast video from a written scene description. You can also provide a supported picture or video clip as a guide. The finished video comes with a link; save the video before that link expires.

Start with your task

Create a [5-second] video showing [a paper boat drifting across a calm pond]. Have the camera [follow the boat slowly]. Use a [wide horizontal view].

What to know

The opening and ending picture option cannot be combined with separate reference pictures, clips or sound. A sound clip cannot be used by itself as a guide. Links to finished videos expire, so save the video soon.

Inputs & outputs

Inspect the current inputs, availability, and maximum charge before starting.

Input schema

Input schema
Field / Path	Type	Requirement	Meaning
Video request`body`	`object`	Required	Required request object containing model, content, resolution and duration. Ratio and media requirements depend on the generation mode. Required fields `model, content, resolution, duration`
Video model`body.model`	`string`	Required	Required string selecting MiniMax-H3-Fast for video generation, fixed to this model. Fixed value `MiniMax-H3-Fast`
Aspect ratio`body.ratio`	`string`	Optional	Frame width-to-height ratio: adaptive, 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16. Text-only requests require an explicit value other than adaptive. Image-to-video follows the image and resolves to adaptive; reference-to-video defaults to adaptive but accepts a concrete supported ratio. Options `adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16`
Prompt and media items`body.content`	`array`	Required	Required array with exactly one nonempty text item. Add first/last-frame image items OR reference-image/video/audio items; these two media families cannot be mixed. Minimum items `1`
[]`body.content[]`	`object`	—
Video description`body.content[].text`oneOf · 1	`string`	—	Required on the single text item: a nonempty string of 1–7,000 characters describing the desired video. Text remains required when reference media is supplied. Minimum length `1` Maximum length `7000`
Content type`body.content[].type`oneOf · 1	`string`	—	Required tag identifying an item as text, image_url, video_url or audio_url. Each tag has its own required text or media URL fields. Fixed value `text`
Media role`body.content[].role`oneOf · 2	`string`	—	Images use first_frame, last_frame or reference_image; a single image without a role defaults to first_frame, and last_frame requires first_frame. Reference video and audio require reference_video and reference_audio respectively. Reference-image maximum is nine, video and audio maxima are three each; audio cannot be the sole reference. Options `first_frame, last_frame, reference_image`
Content type`body.content[].type`oneOf · 2	`string`	—	Required tag identifying an item as text, image_url, video_url or audio_url. Each tag has its own required text or media URL fields. Fixed value `image_url`
Image source`body.content[].image_url`oneOf · 2	`object`	—	Required URL object for an image_url item. A first/last frame supplies an image boundary; reference_image supplies visual reference material. Required fields `url`
Image URL or data URI`body.content[].image_url.url`oneOf · 2	`string`	—	Nonempty public HTTPS URL or data URI; mm_file:// references are unsupported. Images support JPG/JPEG/PNG/WEBP/HEIC/HEIF, up to 30 MB, 256–5,760 pixels per side and aspect ratio 0.4–2.5. Total request body must not exceed 64 MB. Minimum length `1`
Media role`body.content[].role`oneOf · 3	`string`	—	Images use first_frame, last_frame or reference_image; a single image without a role defaults to first_frame, and last_frame requires first_frame. Reference video and audio require reference_video and reference_audio respectively. Reference-image maximum is nine, video and audio maxima are three each; audio cannot be the sole reference. Fixed value `reference_video`
Content type`body.content[].type`oneOf · 3	`string`	—	Required tag identifying an item as text, image_url, video_url or audio_url. Each tag has its own required text or media URL fields. Fixed value `video_url`
Video reference source`body.content[].video_url`oneOf · 3	`object`	—	Required URL object for a video_url item, used with role reference_video. Video references belong to the reference-media mode, not the first/last-frame mode. Required fields `url`
Reference video URL`body.content[].video_url.url`oneOf · 3	`string`	—	Nonempty public HTTPS URL or data URI, with URLs preferred for video. Supports MP4/MOV with H.264 or H.265, up to 50 MB per clip. Each clip is 2–15 seconds, with at most 15 seconds total video reference; mm_file:// is unsupported and the whole request is limited to 64 MB. Minimum length `1`
Media role`body.content[].role`oneOf · 4	`string`	—	Images use first_frame, last_frame or reference_image; a single image without a role defaults to first_frame, and last_frame requires first_frame. Reference video and audio require reference_video and reference_audio respectively. Reference-image maximum is nine, video and audio maxima are three each; audio cannot be the sole reference. Fixed value `reference_audio`
Content type`body.content[].type`oneOf · 4	`string`	—	Required tag identifying an item as text, image_url, video_url or audio_url. Each tag has its own required text or media URL fields. Fixed value `audio_url`
Audio reference source`body.content[].audio_url`oneOf · 4	`object`	—	Required URL object for an audio_url item, used with role reference_audio. Audio supplies reference material within video generation and must accompany image or video reference. Required fields `url`
Reference audio URL`body.content[].audio_url.url`oneOf · 4	`string`	—	Nonempty public HTTPS URL or data URI. Supports WAV/MP3, up to 15 MB per clip; each clip is 2–15 seconds with at most 15 seconds total audio reference. mm_file:// is unsupported and the whole request is limited to 64 MB. Minimum length `1`
Requested duration`body.duration`	`integer`	Required	Required integer video length in seconds, from 5 through 15. The current credit ceiling is separate from the final charge for a chosen duration. Minimum `5` Maximum `15`
Output resolution`body.resolution`	`string`	Required	Required setting for the generated video’s resolution, limited to 480P. Options `480P`

View raw JSON

{
  "type": "object",
  "required": [
    "body"
  ],
  "properties": {
    "body": {
      "type": "object",
      "$schema": "https://json-schema.org/draft/2020-12/schema",
      "required": [
        "model",
        "content",
        "resolution",
        "duration"
      ],
      "~standard": {
        "vendor": "zod",
        "version": 1,
        "jsonSchema": {}
      },
      "properties": {
        "model": {
          "type": "string",
          "const": "MiniMax-H3-Fast",
          "description": "Fixed to MiniMax-H3-Fast. Required."
        },
        "ratio": {
          "enum": [
            "adaptive",
            "21:9",
            "16:9",
            "4:3",
            "1:1",
            "3:4",
            "9:16"
          ],
          "type": "string",
          "description": "Aspect ratio. REQUIRED and must not be adaptive for text-to-video. Image-to-video always resolves to adaptive (the input image decides). Reference-to-video defaults to adaptive but accepts a concrete ratio."
        },
        "content": {
          "type": "array",
          "items": {
            "oneOf": [
              {
                "type": "object",
                "required": [
                  "type",
                  "text"
                ],
                "properties": {
                  "text": {
                    "type": "string",
                    "maxLength": 7000,
                    "minLength": 1,
                    "description": "Prompt describing the video (max 7000 characters). REQUIRED in every request, in every scenario."
                  },
                  "type": {
                    "type": "string",
                    "const": "text"
                  }
                },
                "additionalProperties": false
              },
              {
                "type": "object",
                "required": [
                  "type",
                  "image_url"
                ],
                "properties": {
                  "role": {
                    "enum": [
                      "first_frame",
                      "last_frame",
                      "reference_image"
                    ],
                    "type": "string",
                    "description": "first_frame (default when a single image carries no role), last_frame (must be paired with first_frame), or reference_image (reference-to-video, max 9). Formats: JPG, JPEG, PNG, WEBP, HEIC, HEIF; 256-5760px per side; aspect ratio 0.4-2.5."
                  },
                  "type": {
                    "type": "string",
                    "const": "image_url"
                  },
                  "image_url": {
                    "type": "object",
                    "required": [
                      "url"
                    ],
                    "properties": {
                      "url": {
                        "type": "string",
                        "minLength": 1,
                        "description": "Public HTTPS URL, or a data: URI (data:image/jpeg;base64,...). MiniMax mm_file:// references are not supported. Limits: image <=30MB, reference video <=50MB, reference audio <=15MB, whole request body <=64MB — prefer URLs over data: URIs for video."
                      }
                    },
                    "additionalProperties": false
                  }
                },
                "additionalProperties": false
              },
              {
                "type": "object",
                "required": [
                  "type",
                  "video_url",
                  "role"
                ],
                "properties": {
                  "role": {
                    "type": "string",
                    "const": "reference_video",
                    "description": "Reference video (reference-to-video only, max 3). MP4/MOV, H.264 or H.265, 2-15s per clip and <=15s total."
                  },
                  "type": {
                    "type": "string",
                    "const": "video_url"
                  },
                  "video_url": {
                    "type": "object",
                    "required": [
                      "url"
                    ],
                    "properties": {
                      "url": {
                        "type": "string",
                        "minLength": 1,
                        "description": "Public HTTPS URL, or a data: URI (data:image/jpeg;base64,...). MiniMax mm_file:// references are not supported. Limits: image <=30MB, reference video <=50MB, reference audio <=15MB, whole request body <=64MB — prefer URLs over data: URIs for video."
                      }
                    },
                    "additionalProperties": false
                  }
                },
                "additionalProperties": false
              },
              {
                "type": "object",
                "required": [
                  "type",
                  "audio_url",
                  "role"
                ],
                "properties": {
                  "role": {
                    "type": "string",
                    "const": "reference_audio",
                    "description": "Reference audio (reference-to-video only, max 3, cannot be the only reference). WAV/MP3, 2-15s per clip and <=15s total."
                  },
                  "type": {
                    "type": "string",
                    "const": "audio_url"
                  },
                  "audio_url": {
                    "type": "object",
                    "required": [
                      "url"
                    ],
                    "properties": {
                      "url": {
                        "type": "string",
                        "minLength": 1,
                        "description": "Public HTTPS URL, or a data: URI (data:image/jpeg;base64,...). MiniMax mm_file:// references are not supported. Limits: image <=30MB, reference video <=50MB, reference audio <=15MB, whole request body <=64MB — prefer URLs over data: URIs for video."
                      }
                    },
                    "additionalProperties": false
                  }
                },
                "additionalProperties": false
              }
            ]
          },
          "minItems": 1,
          "description": "Multimodal input. Every request needs exactly one non-empty text item. Add first_frame/last_frame images for image-to-video, OR reference_image/reference_video/reference_audio items for reference-to-video — the two families cannot be mixed."
        },
        "duration": {
          "type": "integer",
          "maximum": 15,
          "minimum": 5,
          "description": "Output length in seconds, 5-15."
        },
        "resolution": {
          "enum": [
            "480P"
          ],
          "type": "string",
          "description": "Output resolution. Drives the per-second rate: 480P $0.046/s. The USD amounts in this input description are upstream supplier values; Toolmu charges the current catalog credits."
        }
      },
      "additionalProperties": false
    }
  },
  "additionalProperties": false
}

Result schema

Output fields have not been documented yet.

Additional notes

One text item describes the scene and motion in every mode. First and last frames define image-to-video boundaries; reference images, video and audio instead guide a reference-media request. Keep these families separate, and use the current Tool’s fixed model and resolution.

Asset size limits apply to each image, video or audio file, while 64 MB applies to the whole request body. Public HTTPS URLs avoid inserting large media strings into the request. An audio reference needs an image or video reference alongside it.

The returned result.task carries the supplier task ID and output settings. Its content.url retrieves the video, and supplier status succeeded means generation completed. Save the video promptly because the URL is time-limited. When returned, created_at and updated_at are Unix seconds for task creation and latest status update; the supplier task ID is separate from the Toolmu run ID.

Usage distinguishes input reference-video seconds, output-video seconds and their total, submitted-image count and billed extra-image count. Keep these figures separate from the confirmed Toolmu charge. Returned style labels are metadata, not additional request settings.

MiniMax video input reference

MiniMax video task and download definitions

Questions about this tool.

Inputs, results and details to know before running it.

Get connected

Yes. Describe the scene, choose a length from 5 to 15 seconds, and choose a horizontal or vertical shape such as 16:9 or 9:16.

MiniMax H3 Fast AI Video Generator

MiniMax H3 Fast AI Video Generator

You receive

Start with your task

What to know

Inputs & outputs

Input schema

Result schema

Describe what you want to see

Choose the video length and shape

Open and save your video

Questions about this tool.

You receive

Start with your task

What to know

Inputs & outputs

Input schema

Result schema

Describe what you want to see

Choose the video length and shape

Open and save your video

Questions about this tool.

Can I make a video by describing it in words?

Can I use a picture or video clip as a guide?

Will the video link last forever?

How much will one video cost?