MiniMax H3

Model IDminimax/h3/text-to-video

MiniMax
Text to video
Video

Generate a video from a text prompt.

Call this model API quickstart

Overview

Model developer: MiniMax
API channel: fal.ai
Resolution: 480P · 768P · 2K · 4K
Duration: 5–15 s
Aspect ratios: 6 aspect ratios
Supported protocols: fal-task
openai-response
openai-video

Call this model

Create an API key, add account credit, and confirm this model is available to your key before sending a request.

LLM index

Copy the complete model reference, including every published protocol, for your AI assistant.

API protocol

POST/v1/tasks/fal

Request parameters

Request parameters
Parameter	Required	Type	Description
`model`	Required	`string`	Use this exact public model ID. Constraints const = "minimax/h3/text-to-video"
`target_audio_url`	Optional	`string`	Accessible HTTP(S) audio URL for reference audio. Constraints minLength = 1 pattern = "\\S"
`resolution`	Optional	`string`	Output resolution. Values are case-sensitive and specific to this model. Default `"2K"` Constraints enum = ["480P","768P","2K","4K"]
`prompt_expansion_mode`	Optional	`string`	Provider prompt expansion strategy. Default `"balanced"`
`aspect_ratio`	Optional	`string`	Requested output width-to-height ratio. Default `"16:9"` Constraints enum = ["21:9","16:9","4:3","1:1","3:4","9:16"]
`sync_mode`	Optional	`boolean`	Only false is supported. This gateway submits asynchronous tasks. Default `false` Constraints enum = [false]
`enable_safety_checker`	Optional	`boolean`	Whether the provider safety checker is enabled. Provider policy may reject attempts to disable it. Default `true`
`duration`	Optional	`integer`	Requested video duration in seconds. Default `5` Constraints minimum = 5 maximum = 15
`prompt`	Required	`string`	Describe the subject, scene, and desired generation. Constraints maxLength = 50000 minLength = 1
`seed`	Optional	`integer`	Integer seed for controlling generation randomness.
`seconds`	Optional	`integer`	Video duration in seconds. This is an alias for duration. Default `5` Constraints minimum = 5 maximum = 15
`size`	Optional	`string`	Alias for image_size or resolution. Prefer the exact native value in this table; pixel-size aliases may not match case-sensitive resolution values.
`parameters`	Optional	`object`	Alternative container for the native parameters in this table. Do not repeat a setting with a different value.
`metadata`	Optional	`object`	Alternative container: place native parameters inside metadata.parameters.

Request notes

Authenticate every request with Authorization: Bearer and your Scenepond API key. Keep the key on your server.
Parameter values are case-sensitive. Missing or null settings use documented defaults when available; valid zero and false values are retained.
The parameter table uses native names at the JSON top level. The same settings can be placed in parameters or metadata.parameters; unknown nested settings and conflicting duplicate values are rejected.
Task submission returns a queued task, not the generated media. Poll status, then list artifacts after SUCCESS.
Do not automatically resubmit a timed-out POST. Preserve the returned task ID and inspect task status before creating another billable task.
Defaults and limits follow this gateway and the linked provider schema. Settings may change after a provider or gateway update.

Minimum request

1curl --fail-with-body -X POST "https://scenepond-api-production.up.railway.app/v1/tasks/fal" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY" \
3  -H "Content-Type: application/json" \
4  --data-binary @- <<'JSON'
5{
6  "model": "minimax/h3/text-to-video",
7  "prompt": "A quiet lakeside at sunrise"
8}
9JSON

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.

Illustrative responses

Response examples show the structure only. IDs, URLs, and values are placeholders.

Submission response

1{
2  "id": "task_EXAMPLE",
3  "task_id": "task_EXAMPLE",
4  "status": "queued",
5  "model": "minimax/h3/text-to-video",
6  "created_at": 1700000000
7}

Completed task status

1{
2  "task_id": "task_EXAMPLE",
3  "platform": "fal",
4  "status": "SUCCESS",
5  "progress": "100%",
6  "fail_reason": "",
7  "created_at": 1700000000,
8  "finished_at": 1700000060
9}

Artifact listing

1{
2  "task_id": "task_EXAMPLE",
3  "artifacts": [
4    {
5      "key": "video",
6      "type": "video",
7      "content_url": "https://scenepond-api-production.up.railway.app/v1/tasks/task_EXAMPLE/artifacts/video/content"
8    }
9  ]
10}

Follow-up requests

2. Check task status

Set SCENEPOND_TASK_ID to task_id. Poll every few seconds until SUCCESS or FAILURE. Status includes progress and fail_reason, but no media.

1curl --fail-with-body -X GET "https://scenepond-api-production.up.railway.app/v1/tasks/$SCENEPOND_TASK_ID" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY"

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.

3. Get the results

After SUCCESS, list artifacts. Use each returned content_url to download the media with authorization; treat signed URLs as private.

1curl --fail-with-body -X GET "https://scenepond-api-production.up.railway.app/v1/tasks/$SCENEPOND_TASK_ID/artifacts" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY"

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.

Use it in your application

Use the task API in your application to submit generation requests and retrieve results by task ID.

API documentation

Model access and request limits depend on your account, API key, and service configuration.

Pricing

Video generation unit priceUSD 0.05 – 0.16 / secondPrices vary by specifications · 4 tiers

Prices by specification

Output resolution: 480P

Video generation unit price

USD 0.05 / second

Output resolution: 768P

Video generation unit price

USD 0.06 / second

Output resolution: 2K

Video generation unit price

USD 0.13 / second

Output resolution: 4K

Video generation unit price

USD 0.16 / second

Applicable conditions	Video generation unit price
Output resolution: 480P	USD 0.05 / second
Output resolution: 768P	USD 0.06 / second
Output resolution: 2K	USD 0.13 / second
Output resolution: 4K	USD 0.16 / second

Get an API key Add funds

Prices depend on model settings. Usage records show recorded deductions; asynchronous task charges may be adjusted after completion.

Call this model

Create an API key, add account credit, and confirm this model is available to your key before sending a request.

LLM index

Copy the complete model reference, including every published protocol, for your AI assistant.

API protocol

POST/v1/tasks/fal

Request parameters

Request parameters
Parameter	Required	Type	Description
`model`	Required	`string`	Use this exact public model ID. Constraints const = "minimax/h3/text-to-video"
`target_audio_url`	Optional	`string`	Accessible HTTP(S) audio URL for reference audio. Constraints minLength = 1 pattern = "\\S"
`resolution`	Optional	`string`	Output resolution. Values are case-sensitive and specific to this model. Default `"2K"` Constraints enum = ["480P","768P","2K","4K"]
`prompt_expansion_mode`	Optional	`string`	Provider prompt expansion strategy. Default `"balanced"`
`aspect_ratio`	Optional	`string`	Requested output width-to-height ratio. Default `"16:9"` Constraints enum = ["21:9","16:9","4:3","1:1","3:4","9:16"]
`sync_mode`	Optional	`boolean`	Only false is supported. This gateway submits asynchronous tasks. Default `false` Constraints enum = [false]
`enable_safety_checker`	Optional	`boolean`	Whether the provider safety checker is enabled. Provider policy may reject attempts to disable it. Default `true`
`duration`	Optional	`integer`	Requested video duration in seconds. Default `5` Constraints minimum = 5 maximum = 15
`prompt`	Required	`string`	Describe the subject, scene, and desired generation. Constraints maxLength = 50000 minLength = 1
`seed`	Optional	`integer`	Integer seed for controlling generation randomness.
`seconds`	Optional	`integer`	Video duration in seconds. This is an alias for duration. Default `5` Constraints minimum = 5 maximum = 15
`size`	Optional	`string`	Alias for image_size or resolution. Prefer the exact native value in this table; pixel-size aliases may not match case-sensitive resolution values.
`parameters`	Optional	`object`	Alternative container for the native parameters in this table. Do not repeat a setting with a different value.
`metadata`	Optional	`object`	Alternative container: place native parameters inside metadata.parameters.

Request notes

Authenticate every request with Authorization: Bearer and your Scenepond API key. Keep the key on your server.
Parameter values are case-sensitive. Missing or null settings use documented defaults when available; valid zero and false values are retained.
The parameter table uses native names at the JSON top level. The same settings can be placed in parameters or metadata.parameters; unknown nested settings and conflicting duplicate values are rejected.
Task submission returns a queued task, not the generated media. Poll status, then list artifacts after SUCCESS.
Do not automatically resubmit a timed-out POST. Preserve the returned task ID and inspect task status before creating another billable task.
Defaults and limits follow this gateway and the linked provider schema. Settings may change after a provider or gateway update.

Minimum request

1curl --fail-with-body -X POST "https://scenepond-api-production.up.railway.app/v1/tasks/fal" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY" \
3  -H "Content-Type: application/json" \
4  --data-binary @- <<'JSON'
5{
6  "model": "minimax/h3/text-to-video",
7  "prompt": "A quiet lakeside at sunrise"
8}
9JSON

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.

Illustrative responses

Response examples show the structure only. IDs, URLs, and values are placeholders.

Submission response

1{
2  "id": "task_EXAMPLE",
3  "task_id": "task_EXAMPLE",
4  "status": "queued",
5  "model": "minimax/h3/text-to-video",
6  "created_at": 1700000000
7}

Completed task status

1{
2  "task_id": "task_EXAMPLE",
3  "platform": "fal",
4  "status": "SUCCESS",
5  "progress": "100%",
6  "fail_reason": "",
7  "created_at": 1700000000,
8  "finished_at": 1700000060
9}

Artifact listing

1{
2  "task_id": "task_EXAMPLE",
3  "artifacts": [
4    {
5      "key": "video",
6      "type": "video",
7      "content_url": "https://scenepond-api-production.up.railway.app/v1/tasks/task_EXAMPLE/artifacts/video/content"
8    }
9  ]
10}

Follow-up requests

2. Check task status

Set SCENEPOND_TASK_ID to task_id. Poll every few seconds until SUCCESS or FAILURE. Status includes progress and fail_reason, but no media.

1curl --fail-with-body -X GET "https://scenepond-api-production.up.railway.app/v1/tasks/$SCENEPOND_TASK_ID" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY"

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.

3. Get the results

After SUCCESS, list artifacts. Use each returned content_url to download the media with authorization; treat signed URLs as private.

1curl --fail-with-body -X GET "https://scenepond-api-production.up.railway.app/v1/tasks/$SCENEPOND_TASK_ID/artifacts" \
2  -H "Authorization: Bearer $SCENEPOND_API_KEY"

Run this example in a shell with cURL installed. Set SCENEPOND_API_KEY before running.