Skip to main content

Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.Browse every model at this filtering level
minimax/h3-x/text-to-video

H3 X requires a prompt of 1–7,000 characters; give the action, camera direction and sound, using double quotes for dialogue. For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps.

From $0.022/s
Run model

Workbench

Input

Text to video
467/7000
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
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.
affects price
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.
Durationaffects price
For an H3 X sequence, choose 2–15 whole seconds; the default is 5 and the output runs at 24 fps. no lower than 2 · no higher than 15
Prompt expansion
The H3 X prompt enhancer starts disabled; enabling it rewrites the instructions before generation without changing the price.
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
Extend
Video translation
Video editing

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

Affordable MiniMax H3 video with spoken lines, timed shots and up to 9 references

MiniMax H3 X is a lower-cost, tuned build of MiniMax H3 that is only available on SpicyAPI, from $0.022 per second of video. It is not an official MiniMax release. Three endpoints make 2 to 15 second clips at 480p, 720p or 1080p with stereo sound: text to video, image to video from an opening frame with an optional closing frame, and reference to video from up to 9 images, 2 video clips and 3 voice clips. A line written in double quotes is spoken, and all three endpoints are uncensored. Its content limits are slightly stricter than MiniMax H3's, in exchange for a lower price per second.

Open the playground
Affordable MiniMax H3 video with spoken lines, timed shots and up to 9 references
Model
A tuned build of MiniMax H3, only on SpicyAPI since 26 October 2026; not an official MiniMax release
Endpoints
Text to video; image to video from an opening frame and an optional closing frame; reference to video from up to 9 images, 2 video clips and 3 voice clips
Clip length
2 to 15 whole seconds at 24 fps, default 5
Resolution
480p, 720p (default) or 1080p; text and reference to video in 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9
Sound
Stereo sound with every clip; a line in double quotes is spoken
Starting price
From $0.022 per second of output at 480p; 720p and 1080p cost more
Price
MiniMax H3 for lessH3 X starts at $0.022 per second of output at 480p, and the live price for each resolution is on the model page. Sound and prompt expansion are included in the per-second price. For drafts there is H3 X Turbo, a distilled version of this model at about half the price.
Dialogue
Quoted lines come back word for wordWrite what a person says in double quotes. In our speech-to-text checks the quoted line came back word for word with matching lip movement, including a run that used a reference voice clip. Rain, breathing and music follow the prompt as well.
Content
Uncensored on all three endpointsNo request was refused in our tests. Image and reference to video delivered an intimate scene between two adults. Content limits are slightly stricter than MiniMax H3's, and text to video tones down some single-person and male-subject NSFW scenes, so start from an image or references for those.

Working with MiniMax H3 X

Start from a frame and write what happens next

Start from a frame and write what happens next

Our opening frame was a bartender in an open-back black velvet dress behind a walnut bar in a dim vinyl lounge. The prompt only described what follows: she sets the glass down, leans across the bar, rests her chin on her hand and whispers "Last call.", with record crackle and soul music underneath. The 720p clip kept her face, the dress and the amber light. The top-left panel is the opening frame, the other three are moments from the clip.

Name each reference by its position

Name each reference by its position

We sent a portrait of a woman with a platinum bob, a product photo of a camel trench coat and an empty art deco fitting room as Image 1, 2 and 3, and wrote "the woman from Image 1", "the trench coat from Image 2" and "the fitting room from Image 3" into the prompt. Her face, the coat and the gold-framed mirror came back in one 720p scene. The left column shows the three references.

  • Give every spoken line its own quotes

    On a rain-soaked rooftop at night, a couple dances a slow tango and she says "Not yet." The prompt names the sound too: rain hissing on the tiles, distant traffic, a bandoneon. Picture, voice and effects come back in one pass, so there is no separate audio step. Keep each line short enough to be said comfortably within the clip length you ask for.

  • Copy a motion from a reference clip

    A reference video can act as a motion template. We sent a portrait of a man in a white tank top as Image 1 and a short dance clip as Video 1, and asked for the same spin and shoulder roll in a sunlit loft, keeping his face and top from Image 1 and only the motion from Video 1. Seconds of reference video are billed like output seconds, so trim the clip to the moves you need.

  • Choose 1080p for the frame size, not for detail

    720p is the default. 1080p returns a real 1920 x 1088 file, but at a lower bitrate than 720p, about 3.3 against about 10 Mbps, and in our crops it did not look sharper. A 5-second clip took about 50 to 110 seconds at 480p, 100 to 130 at 720p and about 171 at 1080p, so draft at 480p, where the price per second is lowest, and finish the take you like at 720p.

Who uses MiniMax H3 X

  • Three shots from one prompt

    Short-drama and series makers

    Three shots from one prompt

    Give each shot a start time and keep the same people in every line of the prompt. Our 12-second clip asked for a wide shot down a hotel corridor, a cut at 4 seconds to a key turning in a brass lock and a cut at 8 seconds to the room, and came back as those three shots with the same woman. In a 15-second test the cuts landed at about 5 and 10 seconds.

  • Vertical clips straight from text

    Social and short-form teams

    Vertical clips straight from text

    Text and reference to video take 9:16 as well as 16:9, 1:1, 4:3, 3:4 and 21:9, and a 9:16 clip at 720p comes back at 736 x 1312. Our lifeguard leaned on the railing of a white tower at golden hour, pushed his sunglasses up and grinned down at a low camera. Pictured: two moments from that 5-second clip.

  • Start from your users' own frames

    Adult-platform builders

    Start from your users' own frames

    Image to video accepted NSFW opening frames in our tests and delivered an intimate scene between two adults, with contact itself out of frame. Some male-subject scenes were toned down; for those, use reference to video. Everyone on screen must be an adult. Pictured: two women sharing a gentle kiss on a Paris balcony at dawn, from a clothed opening frame.

  • Two people from two portraits

    Character and series creators

    Two people from two portraits

    Send one portrait per person and refer to each by position, such as "the man from Image 1" and "the man from Image 2". Two portraits of two men in suits came back in a mirrored brass hotel elevator, one loosening the other's tie, with both faces kept. Reuse the same portraits in the same order for every shot of a sequence.

From first prompt to final take with MiniMax H3 X

  1. Pick the endpoint by what you have

    Only words go to text to video. A frame of the scene goes to image to video, with an optional closing frame. Separate photos of people, outfits, products or places, a motion clip or a voice go to reference to video. Frames and references cannot be mixed in one request.

  2. Write the action, the camera and the sound

    Describe what happens in order, what the camera does and what we hear, and put every spoken line in double quotes. On image to video, describe what follows the frame rather than the frame itself. Keep prompt expansion off when you want your wording used as written.

  3. Draft at 480p

    Run a few 5-second takes at 480p, where the price per second is lowest, and check faces, hands and the spoken line before you go longer.

  4. Finish at 720p, only as long as needed

    Rerun the take you like at 720p and at the length the scene needs, up to 15 seconds; every clip is billed per second of output. Choose 1080p only when you need the bigger frame.

  5. For single-person or male-subject NSFW scenes, start from an image

    Text to video tones down some single-person and male-subject NSFW scenes. Use image to video from a frame of the people, or reference to video. Everyone on screen must be an adult.

MiniMax H3 X prompts worth copying

  • A tango with one quoted line

    For text to video at 16:9. The clothing, the move, the line in quotes and the sound are each named once.

    Night, a rain-soaked rooftop above a glittering city. A couple dances a slow tango in the downpour: she wears a clinging crimson satin slip dress, he wears a soaked white shirt open at the collar. He dips her low, she pulls him back up by the collar, their faces an inch apart, and she says, "Not yet." Rain hissing on the tiles, distant traffic, a bandoneon playing a slow tango. Cinematic, shallow depth of field, wet reflections, warm city glow against blue night.

  • A frame that whispers

    For image to video from a frame of one clothed adult. Describe only what happens after the frame.

    She sets the glass down slowly, leans across the bar toward the camera, rests her chin on her hand and whispers, "Last call." A record crackles on the turntable, low soul music, ice clinking. The camera pushes in slowly. Warm amber light, shallow depth of field.

  • Three references, one fitting

    For reference to video with a portrait, a garment photo and a location photo, in that order. Name each image by its position.

    The woman from Image 1, in her black slip dress, tries on the trench coat from Image 2 in the fitting room from Image 3. She belts it, turns to the gold mirror, then lets the coat slide off one shoulder and glances back at the camera over that shoulder. Soft spotlight, rustle of wool, quiet lounge music. Cinematic, shallow depth of field.

More MiniMax H3 X questions

Use it when the scene needs a voice and the budget is tight

Pick text to video when words are enough, image to video when you already have the frame, and reference to video when people, outfits, places or a motion must match.

Quote every line, time your shots, draft at 480p and finish at 720p. For single-person or male-subject NSFW scenes, start from an image or from references.

All MiniMax models17

Video models12

Audio models3

Text models1

Image models1