Introduction
MiniMax H3 X API combines a tuned H3 build with frame control and multimodal scene assembly; it starts at $0.022/s on SpicyAPI and is not an official MiniMax release.
An H3 X reference scene accepts 9 images, 2 videos totalling 15 seconds and 3 audio clips at most; name their roles with Image 1, Video 1 and Audio 1. For an H3 X product scene, assign the portrait to the performer, a second image to clothing and another to the set, then describe how they interact.
For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps. H3 X offers 480p, default 720p and 1080p; our SpicyAPI crops showed no visible detail gain at 1080p despite the larger 1920 x 1088 file. For H3 X text and reference scenes choose 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9, with 16:9 selected initially; an opening image instead determines the shape. H3 X creates AAC stereo audio with its H.264 MP4 picture; write dialogue inside double quotes and describe the surrounding sound separately.
H3 X is classified uncensored for NSFW video on all three endpoints; SpicyAPI tests recorded no refusals, but text-only and opening-frame results sometimes softened requests. For H3 X, obtain permission for any person whose likeness or voice you use and keep every depicted person an adult.
In H3 X reference mode, image order determines Image 1 through Image 9; separate identity, wardrobe and setting instructions, with only the first 5 images included in the base charge. H3 X can borrow motion from up to 2 reference videos with 15 seconds combined; their duration contributes to the bill as output duration does. Use up to 3 audio references to guide an H3 X voice; the reference supplies its sound, while the prompt still supplies the quoted words, without an audio-reference surcharge. H3 X billing follows output seconds and resolution; reference video adds billable seconds, the first 5 images are included and each later image adds a charge, while reference audio is free.
SpicyAPI reference-led H3 X tests delivered male-subject requests in 5 of 5 trials; the Very strong test category was delivered in 6 of 8, so fidelity still varied.
| Field | Value |
|---|---|
| Input | Text |
| Output | Video |
| Publisher | SpicyAPI / MiniMax H3 X |
| Model | MiniMax H3 X |
| Released | 26 October 2026 |
| Endpoints | MiniMax H3 X — Text to video / Image to video / Reference to video |
| Duration | For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps. |
| Resolution | 480p, 720p (default), 1080p |
| Aspect ratio | For H3 X text and reference scenes choose 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9, with 16:9 selected initially; an opening image instead determines the shape. |
| Prompt length | 1 to 7,000 characters |
| References | An H3 X reference scene accepts 9 images, 2 videos totalling 15 seconds and 3 audio clips at most; name their roles with Image 1, Video 1 and Audio 1. |
| Content policy | H3 X is classified uncensored for NSFW video on all three endpoints; SpicyAPI tests recorded no refusals, but text-only and opening-frame results sometimes softened requests. |
| Execution | H3 X billing follows output seconds and resolution; reference video adds billable seconds, the first 5 images are included and each later image adds a charge, while reference audio is free. |
Key features & highlights
A quoted line in the rain
A couple dances a slow tango on a rain-soaked rooftop; he dips her low, then in a close-up she pulls him up by his shirt collar and says "Not yet." The prompt also lists rain, traffic and a bandoneon; in the clip the line is heard over slow, melancholic music. Quoted lines were spoken word for word in our speech-to-text checks.
Three shots in one 12-second clip
One prompt lists three shots with start times: a wide corridor walk, a hand turning a key in a door lock at 4 seconds, a medium shot in the room at 8 seconds. In the clip the cuts landed at about 3.9 and 8.3 seconds, with the same woman throughout.
A frame that whispers
A woman with dark wavy hair in a black velvet dress behind a walnut bar goes in as the opening frame. The prompt only says what follows: she sets her glass down, leans toward the camera and says "Last call." The clip has quiet music under the line.
Three references, one scene
A portrait, a product photo of a trench coat and an empty fitting room go in as Image 1, 2 and 3. The prompt names them by position and asks for one scene with all three; in the clip the face, outfit, coat and room all match the references.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | H3 X requires a prompt of 1–7,000 characters; give the action, camera direction and sound, using double quotes for dialogue. at least 1 characters · at most 7000 characters |
reference_image_urls | No | In H3 X reference mode, image order determines Image 1 through Image 9; separate identity, wardrobe and setting instructions, with only the first 5 images included in the base c… at least 1 items · at most 9 items |
reference_video_urls | No | H3 X can borrow motion from up to 2 reference videos with 15 seconds combined; their duration contributes to the bill as output duration does. at least 1 items · at most 2 items |
reference_audio_urls | No | Use up to 3 audio references to guide an H3 X voice; the reference supplies its sound, while the prompt still supplies the quoted words, without an audio-reference surcharge. at least 1 items · at most 3 items |
aspect_ratio | No | For H3 X text and reference scenes choose 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9, with 16:9 selected initially; an opening image instead determines the shape. |
resolution | No | H3 X offers 480p, default 720p and 1080p; our SpicyAPI crops showed no visible detail gain at 1080p despite the larger 1920 x 1088 file. |
duration_seconds | No | For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps. |
enable_prompt_expansion | No | The H3 X prompt enhancer starts disabled; enabling it rewrites the instructions before generation without changing the price. Default: |
Pricing
H3 X billing follows output seconds and resolution; reference video adds billable seconds, the first 5 images are included and each later image adds a charge, while reference audio is free.
| Specification | Price |
|---|---|
| $0.022 / second |
| $0.032 / second |
| $0.045 / second |
Use cases
Dialogue scenes
An H3 X speaking shot can pair a character’s turn with a short quoted line and the ambient sound of the set.
Multi-shot sequences
Use H3 X for a planned sequence: allocate time to each shot and repeat the subject description when the camera angle changes.
Fashion, product and location films
For an H3 X product scene, assign the portrait to the performer, a second image to clothing and another to the set, then describe how they interact.
Motion copying
H3 X can borrow motion from up to 2 reference videos with 15 seconds combined; their duration contributes to the bill as output duration does.
Apps without a content filter
For H3 X, obtain permission for any person whose likeness or voice you use and keep every depicted person an adult.
Prompt guide & pro tips
Put spoken lines in double quotes
An H3 X speaking shot can pair a character’s turn with a short quoted line and the ambient sound of the set.
Time your shots
Use H3 X for a planned sequence: allocate time to each shot and repeat the subject description when the camera angle changes.
Name references by type and position
An H3 X reference scene accepts 9 images, 2 videos totalling 15 seconds and 3 audio clips at most; name their roles with Image 1, Video 1 and Audio 1.
Observed limits
For H3 X, obtain permission for any person whose likeness or voice you use and keep every depicted person an adult.
Scene direction
In an H3 X image-led scene, spend the prompt on what changes after the supplied frame: a gesture, a camera move or a spoken reply.
MiniMax H3 X
MiniMax H3 X API combines a tuned H3 build with frame control and multimodal scene assembly; it starts at $0.022/s on SpicyAPI and is not an official MiniMax release.
Scene direction
An H3 X reference scene accepts 9 images, 2 videos totalling 15 seconds and 3 audio clips at most; name their roles with Image 1, Video 1 and Audio 1. For an H3 X product scene, assign the portrait to the performer, a second image to clothing and another to the set, then describe how they interact. An H3 X speaking shot can pair a character’s turn with a short quoted line and the ambient sound of the set.
Output and sound
For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps. H3 X offers 480p, default 720p and 1080p; our SpicyAPI crops showed no visible detail gain at 1080p despite the larger 1920 x 1088 file. H3 X creates AAC stereo audio with its H.264 MP4 picture; write dialogue inside double quotes and describe the surrounding sound separately.
Billing
H3 X billing follows output seconds and resolution; reference video adds billable seconds, the first 5 images are included and each later image adds a charge, while reference audio is free.
Observed limits
H3 X is classified uncensored for NSFW video on all three endpoints; SpicyAPI tests recorded no refusals, but text-only and opening-frame results sometimes softened requests. For H3 X, obtain permission for any person whose likeness or voice you use and keep every depicted person an adult.
| only an idea | MiniMax H3 X Text to video | | a frame to animate, with an optional closing frame | MiniMax H3 X Image to video | | people, outfits, places, motion clips or voices | MiniMax H3 X Reference to video |
SpicyAPI reference-led H3 X tests delivered male-subject requests in 5 of 5 trials; the Very strong test category was delivered in 6 of 8, so fidelity still varied.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "minimax/h3-x/reference-to-video",
"input": {
"aspect_ratio": "16:9",
"duration_seconds": 5,
"prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
"resolution": "720p",
"reference_image_urls": [
"https://example.com/reference.png"
]
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.
