Image à Vidéo
MiniMax H3
Télécharger une image
Cliquez ou faites glisser pour télécharger, ou choisissez parmi Mon Historique
Pas d'idées ? Générer une image d'abord. >
Formats d'importation pris en charge : jpg, jpeg, png, webp, heic, heif, images jusqu'à 30 Mo, avec une largeur/hauteur minimale de 256 px.
Prompt
0/5000
5s
2K
Vidéo d'échantillon
Unified multimodal video creation

MiniMax H3 AI Video Generator

Need more than another silent text-to-video clip? Media.io brings MiniMax H3 into one browser-based workflow where text, images, video, and audio can work together as creative context. Direct the character, motion, camera, style, voice, and pacing, then generate 2K video up to 15 seconds with native stereo audio.

  • 2K Output
  • Up to 15 Seconds
  • Native Stereo Audio
  • Up to 12 Mixed References
Reference Image
Reference image of a green military-style insulated cup with handle
Creative Direction

Use the uploaded images as the sole reference for the product's appearance, and create a 15-second, 16:9 landscape version of a high-end product official website landing page UI/UX demonstration video. The core product is the green military-style cup with a handle and insulated feature shown in the picture. Throughout the process, strictly maintain the cup body color, cylindrical proportion, transparent cup lid, top opening structure, right square handle, metal decorative ring, and ""media.io"" logo, without changing the product shape or creating other cup models.
The overall design adopts a dark outdoor functional style and the visual style of the official website of a high-end sports brand: a pure black background, interwoven with dark carbon fiber, metal brushed and waterproof fabric textures, and incorporating fast-flowing military green and warm golden dynamic light bands. The use of rough and powerful, slightly tilted oversized sans-serif fonts, with the text creating front-back occlusion with the product as the background graphic. The page layout is sophisticated, restrained, with a sense of speed and strength.
0 - 4 seconds: Start from the main screen of the product's official website. The insulated cup is suspended in the center of the screen in an extremely large proportion. A clear and legible giant title ""BUILT TO GO"" appears in the background. The product slowly approaches and slightly rotates. Green and gold light quickly sweeps across the cup body, highlighting the matte material, transparent lid, metal ring and wide handle. The top retains a minimalist navigation bar, and at the bottom, a few product parameters and thumbnails are arranged.
4 - 6 seconds: The mouse scrolls down the webpage, and the first screen smoothly exits upwards. The camera follows the page into the product detail area. The cup lid, handle, cup body and bottom details appear one by one in horizontal cards. The scrolling has a natural inertia and a slight speed blur, but the product information remains clear.
6 - 10 seconds: Enter the product function section. On the left side of the page, a clear title ""ALL-DAY TEMPERATURE"" is displayed, and on the right side, a complete suspended insulated cup is shown. The product rotates slowly and steadily at a quarter of a turn angle. Abstract cold air currents, fine water droplets, and metal waves appear in the background, demonstrating the insulation and cold preservation performance. When the mouse hovers over the function button, the product slightly enlarges, and the background changes from black to light gray, with the text and button colors reversing synchronously; only the UI background and interface colors are changed, without altering the insulated cup itself.
10–12 seconds: Continue scrolling to the ""LEAKPROOF LID"" and ""EASY-GRIP HANDLE"" function area. Use close-ups of the cup lid and handle, with geometric lines pointing to the structural details. The page elements slide in quickly and stop steadily in sequence.
12-15 seconds: The webpage smoothly returns to the home screen, the thermos cup is re-sized to the center of the screen, the giant title ""BUILT TO GO"" is located behind the product. A clear ""ADD TO CART"" capsule button appears at the bottom. The mouse accurately moves to the center of the button and triggers the hover effect: the button enlarges, turns black and white, and emits a greenish glow; the product remains stable. Finally, it stays on the complete product main visual and the purchase button.
Shots and animations: Smooth vertical scrolling of the webpage, product floating rotation, slow zoom in, layer parallax effect, high-speed light band sweeping, card sliding in, mouse hover zooming in, UI color inversion. All movements are compact and smooth, resembling the demonstration of a real and interactive high-end product website.
Restrictions: The color, structure, handle position, cup lid shape and brand logo of the insulated cup must not be altered; the product must not be transformed into other cup models; no figures or hands are allowed; unpacking is prohibited; the cup must not soften, bend, melt or deform; no garbled codes, incorrect labels or extra products are permitted; color inversion only affects the web background, buttons and text; the product must always remain clear, complete and stable.

MiniMax H3 Demo — Ecommerce Product Website Ad
Creative Direction

## 15-Second K-Pop Girl Group MV Prompt

Subject: A three-member K-pop girl group. Keep each member consistent with the established character design. Preserve their facial features, hairstyles, body proportions, outfits, and individual identities throughout the entire video.

SOL: Long, straight black hair with a calm yet commanding presence. She wears a black deconstructed cropped leather jacket, a fitted top with metallic fastenings, a high-waisted asymmetrical pleated miniskirt, and pointed over-the-knee boots, accessorized with silver rings and a delicate waist chain.

LUNA: Short silver-gray wolf cut with a cold, detached presence. She wears a cropped silver-gray jacket with sculptural paneling, a black high-neck bodysuit, metallic asymmetrical skort, and black platform biker boots, accessorized with transparent mesh sleeves and metallic earrings.

CORONA: Long, wavy dark auburn hair with thin braids, projecting a rebellious and sharp attitude. She wears a distressed dark-red leather biker jacket, a black strappy crop top, low-rise layered skort, and dark-red lace-up boots, accessorized with a studded belt and metallic bracelets.

The outfits must feature convincing high-fashion materials, including realistic leather texture, metallic reflections, fabric folds, and natural garment movement. Avoid cheap stage costumes, plastic-looking materials, or generic girl-group uniforms.

Setting: A white cyclorama studio transformed into an underground music magazine photoshoot set. It should not resemble a clean commercial studio. The background features overexposed white walls, heavy gray-black shadows, photocopied paper fragments, torn paper edges, halftone dots, black tape marks, misaligned scans, and localized black spray-paint textures. In selected shots, the background instantly switches between pure black, overexposed white, and high-contrast gray, using hard cuts only.

Visual Style: A dark-pop, cyber-grunge, trap-rap music video with realistic high-fashion production value, combining late-1990s to early-2000s independent music magazines, underground concert posters, photocopied paper, scanned film, and zine collage aesthetics. Use high-contrast black and white with low-saturation dark red and cold silver-gray accents. Maintain natural skin tones, deep blacks, and clearly visible details in dark clothing. Add coarse film grain, subtle film jitter, exposure flicker, halftone dots, rough printed edges, scan displacement, shutter drag, and localized color-registration errors. Avoid cheap cyberpunk neon, video-game CGI, plastic skin, and excessive beauty smoothing.

Cinematography: Create the feeling of a real, high-budget music video, simulating the natural skin tones and high dynamic range of an ARRI Alexa 35. Rapidly alternate between 24mm, 35mm, and 50mm cinema lenses. Use 24mm for close-range wide-angle pressure, low-angle group shots, and hand-to-camera impact shots; 35mm for medium close-up performances, two-person spatial compositions, and fast lateral camera moves; and 50mm for facial close-ups, profiles, and intense eye contact. Frequently use confrontational wide-angle close-ups, beat-synchronized head movements, shoulder details, low-angle medium shots, hand-driven camera pushes, and tight profile close-ups. The camera may use controlled handheld movement, sudden forward lunges, rapid lateral slides, or abrupt stops precisely on the beat. Maintain convincing camera weight, physical inertia, perspective changes, spatial parallax, and accurate focus. The performers must hit the beat through expressive choreography and rap performance. This is not a runway walk or a static fashion shoot.

Music: A 15-second dark-pop, cyber-grunge, trap-rap segment at approximately 138 BPM. Use powerful sub-bass, tight and crisp snares, distorted industrial electronic noise, and short vocal chops. The first three seconds establish tension, the middle section enters a continuous rap sequence with dense editing, and the final two seconds conclude with concentrated heavy bass.

Typography: The text is not ordinary subtitles. It is an active part of the composition and spatial design. Use original, unbranded phrases such as “CUT THE FEED,” “DEAD AIR,” “STRIKE FIRST,” “SYSTEM / VOID,” “NO SURRENDER,” “UNMUTED,” “FINAL WARNING,” and “BLACKOUT CODE.” Do not include real brands, real group logos, or existing song titles. Combine heavy black sans-serif lettering, condensed industrial fonts, typewriter text, and handmade photocopied lettering. Position the typography across the foreground, middle ground, and background. The performers may partially obscure the words, while the text may overlap their bodies, outfits, or hands, but it must never cover their eyes or essential facial expressions. Synchronize every appearance, shake, stretch, compression, fracture, flip, and hard-cut rearrangement with rap accents, snares, and bass hits. Add ink bleed, photocopy noise, misregistered color plates, and rough scanned edges to the typography.

Editing Rules: Use only hard cuts, jump cuts, white-flash cuts, black-flash cuts, and beat-matched cuts. No fades, dissolves, soft transitions, or layered slow motion. Each shot should last between 0.3 and 1.2 seconds. The pacing must be fast but visually readable. Every movement stop, camera stop, typographic change, and edit point must land precisely on the beat.

SHOT 01|0.0–0.8s

A pure black frame hard-cuts on the first bass hit to an extreme frontal close-up of SOL.

A 24mm wide-angle lens is positioned very close to her face. The camera remains level with SOL’s eyes, with focus firmly locked onto her pupils. SOL leans aggressively toward the lens and suddenly snaps her head downward. Her long black hair reacts with short, physically accurate momentum.

The image is highly contrasted, and the background is severely overexposed, but the highlights on her face must not clip. The enormous bold black word “DEAD” appears behind her, with her head obscuring its center. On the beat, the film grain suddenly intensifies and the camera produces one short vertical jolt.

SHOT 02|0.8–1.5s

Hard cut to a close-up of LUNA’s shoulder and jawline.

A 50mm lens films her from the front-left side. The framing crops part of the top of her head to emphasize her jawline and gaze. She lowers one shoulder and initially looks toward the camera in profile, then abruptly turns her face forward on the snare.

Her silver-gray hair sweeps across the edge of her eye without covering the pupil. Halftone dots spread from the background into the edges of her clothing. The condensed word “AIR” is hard-cut and compressed into the frame from the right, positioned between the background and her shoulder in the foreground.

SHOT 03|1.5–2.3s

Hard cut to a low-angle medium shot of CORONA.

The 24mm camera is positioned below waist level and tilted slightly upward. She presses her body forward and thrusts one hand toward the lens. Her palm sweeps across the foreground before immediately retracting on the bass hit.

As her hand approaches the lens, it naturally enlarges through perspective and briefly falls out of focus. When she pulls it back, focus rapidly returns to her eyes. Her long dark-auburn hair and thin braids move with realistic inertia. Dark-red and black color layers are slightly misregistered, making the image resemble a scanned vintage magazine. Her hand partially obscures the word “STRIKE.”

SHOT 04|2.3–3.0s

A tight frontal medium shot of all three members.

Use a stable 35mm central composition. SOL stands in the center, slightly closer to the camera, while LUNA and CORONA stand on either side. All three simultaneously lower their shoulders, then sharply lift their heads and stare into the camera on the beat.

The background hard-cuts from white to black and back to white while the performers remain perfectly aligned. The large word “CUT” stretches horizontally behind them, with their bodies obscuring and fragmenting its center.

SHOT 05|3.0–3.7s

A 50mm profile close-up of SOL.

She turns and suddenly stops, rapidly rotating her face from profile toward the camera while sliding a finger along her jawline. Focus begins on her fingertip before quickly shifting to her eyes.

At the exact moment her movement stops, the camera makes a fast 20-centimeter push along the lens axis and immediately brakes. The word “SYSTEM” appears in the left foreground, slightly overlapping her shoulder without covering her face.

SHOT 06|3.7–4.4s

A low-angle full-body shot of LUNA.

A 24mm lens is placed close to the floor and tilted slightly upward. She performs a short, realistic, and forceful dance phrase: her knees turn inward, her upper body pulls backward, her arms slice apart from across her chest, and then she immediately leans toward the camera.

Her silver-gray jacket, asymmetrical skort, and transparent mesh sleeves respond with clear fabric inertia. On the snare, the image switches to a high-contrast black-and-white halftone treatment for three frames before immediately returning to low-saturation color.

SHOT 07|4.4–5.1s

A tight 35mm facial close-up of CORONA.

The camera rapidly slides from her right side into a frontal position. She turns her head on the rhythm, lifts her chin slightly, and then deliberately moves closer to the lens. She does not smile; her expression is confrontational, sharp, and grounded in an authentic rap performance.

The dark-red word “VOID” suddenly cuts from behind her into the foreground, overlapping parts of her hair and jacket while leaving her eyes, nose bridge, and mouth fully visible.

SHOT 08|5.1–6.0s

A rapid montage of alternating head movements from all three members.

SOL delivers a forceful frontal nod, LUNA performs a fast profile head turn, and CORONA leans forward and snaps her head downward. Retain only three to five frames from each action and connect them with three consecutive hard cuts.

Keep their eyes positioned in approximately the same area of the frame to create precise match cuts. The background jump-cuts between white, gray, and black while film scan lines briefly roll from top to bottom.

SHOT 09|6.0–6.8s

A close two-shot of SOL and CORONA.

Use a 35mm lens. SOL stands in front and raps directly into the camera. CORONA suddenly emerges from behind her and rapidly approaches along the camera axis.

Maintain clear spatial separation between them. Focus briefly shifts from SOL to CORONA before settling between both performers. The word “NO” appears in the background, while “SURRENDER” sits along the lower foreground, overlapping the edges of their outfits without covering their faces.

SHOT 10|6.8–7.6s

A solo medium close-up of LUNA.

She rapidly crosses both hands beside her face and slices them outward, precisely hitting consecutive snares. Using a 35mm lens, the camera simultaneously slides left by approximately half a body width and abruptly stops on the final accent.

During the lateral move, the background displays limited, physically realistic motion blur while her face remains sharp. Torn photocopied paper, black tape, and rough spray-painted graphics appear as if physically attached to the studio wall.

SHOT 11|7.6–8.5s

A low-angle, full-body wide shot of all three members.

Use a slightly upward-tilted 24mm lens. Each member performs a different but complementary action:

SOL steps forward and drops her shoulders;

LUNA folds her body sideways and lowers her center of gravity;

CORONA turns rapidly, then looks back and moves toward the camera.

Their movements should not be perfectly synchronized, but all three must stop simultaneously on the same bass hit. The camera slowly pulls backward before locking instantly on the accent.

SHOT 12|8.5–9.3s

An extreme close-up montage of hands and outfit details.

Use 50mm macro-style close-ups to rapidly reveal metallic rings, waist chains, jacket zippers, fingers tapping against a thigh, and a palm thrust toward the lens. Each detail appears for only two to four frames.

The text fragments “CU / T / OUT” behave like physical paper cutouts, hard-cutting, rotating, and recombining in different areas of the frame. Metallic reflections must look realistic, with crisp detail but no excessive digital sharpening.

SHOT 13|9.3–10.2s

CORONA raps in the foreground, filmed extremely close with a 24mm lens.

She points one finger directly toward the camera, creating dramatic foreground perspective enlargement, then rotates her wrist and presses her hand downward. Keep focus on her eyes while allowing slight depth-of-field changes as her hand approaches the lens.

LUNA and SOL remain in the shallow-focus background, performing subtle head and shoulder movements. Create a three-layer composition: “THE FEED” in the background, the two members in the middle ground, and CORONA with her hand gesture in the foreground.

SHOT 14|10.2–11.1s

LUNA rapidly rotates into frame from the left.

A 35mm lens performs a short pan to the right, accurately following her turn. Her garment hem swings naturally with the movement. She then brings her face close to the right edge of the frame and looks directly into the lens.

On the musical accent, add a brief horizontal scan displacement. Introduce a restrained red-cyan color-registration offset around her silhouette while keeping her skin tone and facial features natural and stable.

SHOT 15|11.1–12.0s

A centered frontal medium shot of SOL.

A 50mm lens locks focus onto SOL’s eyes. She lowers her head and pauses for half a beat, then violently snaps it upward. LUNA and CORONA simultaneously press forward from behind her as the camera rapidly pulls backward by one step, using realistic parallax to create spatial impact.

The enormous word “UNMUTED” suddenly reconstructs itself from the background into the space between the performers and the foreground. Their bodies divide it into multiple fragments, but none of their eyes may be obscured.

SHOT 16|12.0–13.0s

Three consecutive hard cuts:

SOL is filmed with a 24mm lens as she rapidly pushes one finger toward the camera;

LUNA is filmed with a 35mm lens as she forcefully drives her shoulder toward the lens;

CORONA is filmed with a 50mm lens as she turns away and suddenly looks back directly into the camera.

Each shot corresponds to one snare hit. The focal lengths become progressively tighter, continuously increasing the visual pressure. Film grain, print noise, and rough photocopied edges intensify with each cut.

SHOT 17|13.0–14.0s

A low-angle, full-body hero shot of all three members.

Use a 24mm lens with a strong central-axis composition. SOL stands in the center and steps forward. LUNA turns sideways and lowers her center of gravity. CORONA leans forward. Together, they form a stable, sharp, asymmetrical triangular composition.

They maintain natural breathing, garment inertia, and subtle body movement rather than freezing in a static pose. The large background text “FINAL WARNING” is partially obscured by their bodies, leaving only fragmented letterforms visible.

SHOT 18|14.0–15.0s

The final bass hit.

Using a 35mm lens, the camera suddenly pushes forward along the central axis, moving from a three-person medium shot into a tight close-up of their faces. All three simultaneously deliver one powerful head hit and stare directly into the lens.

The image instantly compresses into a high-contrast black-and-white photocopy effect. The original phrase “BLACKOUT CODE” briefly appears between the background and the foreground plane of their shoulders, shaking twice.

On the final frame, cut directly to pure black. The music and image stop simultaneously, with no fade and no lingering audio tail.

Negative Requirements: No generic girl-group lineup, prolonged runway walking, static fashion poses, weak or slow movements, extended synchronized group choreography, excessively mechanical dancing, cartoonish expressions, cyberpunk neon cities, cheap RGB light pollution, real brands, existing group logos, existing song titles, ordinary subtitles, fades, dissolves, soft transitions, dreamy soft-focus filters, video-game CGI aesthetics, plastic-looking skin, excessive beauty smoothing, facial identity drift, random hairstyle or outfit changes, inaccurate focus, floating camera movement without physical inertia, overly smooth AI camera motion, malformed fingers, or additional people.

MiniMax H3 Demo — K-Pop Music Video
Reference Image
Reference image of a silhouetted western cowboy beside a telegraph pole
Creative Direction

Using the attached reference image as the exact opening frame and character reference,

RHYTHM — HIGHEST PRIORITY:

Edit the entire sequence to an energetic 150 BPM jazz-western rhythm with rapid snare hits, double-bass pulses, sharp brass stabs, hand claps, spur sounds, whip cracks, and short revolver-cylinder clicks. Every cut, pose, camera move, graphic bar, and typography hit must land precisely on a beat or half-beat.

Create at least 20–24 clearly different shots within 15 seconds. Most shots last only 0.2–0.7 seconds. No shot may remain visually unchanged for longer than 0.8 seconds. Maintain constant visual progression and escalating energy.

Use only very short 2–4 frame micro-freezes, approximately 0.08–0.15 seconds, immediately followed by explosive movement. Never hold a freeze-frame for a full beat or several seconds. The rhythm must feel like: burst—hit-stop—burst—hard cut—burst.

VISUAL STYLE:

Create a high-energy 1960s Spaghetti Western title sequence combining mid-century graphic design, Saul Bass-inspired composition, Bauhaus geometry, pulp-western posters, comic-book panels, and rough screen-print texture.

Render the cowboy and environmental details as bold, high-contrast black silhouettes. Use only saturated red, mustard yellow, deep navy, black, and small cream highlights. Every cut should introduce a new color arrangement, crop, graphic shape, or camera angle.

Use dramatic diagonal compositions, oversized sun circles, sharp horizontal desert lines, rapidly moving telegraph wires, rotating revolver chambers, boot and spur close-ups, coat silhouettes, hat-brim close-ups, split screens, geometric shutters, expanding circles, diagonal bars, and bold typography.

TIMELINE AND SHOT DENSITY:

0.0–1.5s — RAPID OPENING:

Begin from the exact reference image for only 0.2 seconds.

Immediately perform four rapid cuts:

wide cowboy silhouette against red sky → snap zoom into hat brim → telegraph wires whip across frame → extreme close-up of shadowed eyes.

Insert a 3-frame black flash between selected shots.

The cowboy pushes sharply away from the pole on the first major brass hit.

1.5–4.0s — DETAIL IMPACT MONTAGE:

Create six fast close-ups, each lasting approximately 0.3–0.5 seconds:

boot strikes the ground → spur spins → coat tail snaps → fingers flex beside the holster → belt buckle flashes → revolver cylinder rotates.

Alternate red, mustard yellow, and deep navy backgrounds on every cut.

Use snap zooms, whip pans, crash zoom-outs, diagonal wipes, and 2–4 frame pose freezes.

4.0–7.0s — ACCELERATING WALK:

Show the cowboy walking rapidly across a flat mustard-yellow desert field.

Cut between full-body side view, low-angle boots, overhead graphic silhouette, extreme close-up of the swinging hand, and comic-book split panels.

His walking rhythm must land exactly on the percussion.

The telegraph pole and desert lines streak past in opposite directions to create intense parallax and speed.

Do not show a continuous slow walk—break the movement into fast rhythmic fragments.

7.0–10.0s — GEOMETRIC ACTION BURST:

Divide the screen into two, three, then four comic-book panels.

In different panels, show the hat turning, boots stopping, the coat opening, the hand approaching the revolver, and the cowboy sharply looking toward camera.

Panels slam open and shut on consecutive beats.

Use alternating full-frame red, yellow, navy, and black flashes.

Add rapid horizontal bars, expanding sun circles, rotating star shapes, and angular shadow wipes.

Every composition changes within 0.5 seconds.

10.0–13.0s — CLIMAX BUILD:

Increase the editing speed further with 8–10 cuts:

eyes → hand → holster → spur → revolver grip → coat → telegraph wire → oversized sun → full cowboy silhouette.

Use progressively tighter close-ups and stronger snap zooms.

The cowboy reaches the revolver, turns sharply, and hits a powerful gunfighter pose.

Use a 3-frame dead stop, then immediately cut to the next angle.

No slow motion and no long suspenseful pause.

13.0–15.0s — FINAL IMPACT:

Use three rapid false endings:

cowboy silhouette against yellow → hard cut to navy extreme close-up → hard cut to red wide shot.

On the final musical hit, reveal the cowboy beside the telegraph pole against a solid red sky and oversized mustard-yellow sun.

The exact title “OUTLAW” slams into the empty left side as large cream geometric letters.

Hold the final composition for only the last 0.6 seconds, then end with an immediate hard stop.

MiniMax H3 Demo — Comic Western Title Sequence

More Ways to Direct the Video You Intended

MiniMax H3 is a general-purpose omni-modal generation model built to understand text, images, video, and audio as one creative context.
Describe how those inputs relate, and H3 can follow the instruction to generate or edit a controllable video with native stereo sound.

One unified context

Direct complex multimodal tasks in natural language

Combine up to 9 images, 3 videos, and 3 audio clips within 12 mixed files, then explain their roles in one instruction—for example, reference a camera move, a character image, and a voice or song together.

2K + stereo

2K video with native stereo sound

Create sharper product, campaign, and social visuals in 2K, with H3 using the original multimodal context again to retain more small text and fine detail than a conventional upscale alone.

Up to 15 sec

Room for a complete video moment

Generate clips up to 15 seconds to establish a scene, develop motion, and land on a clear ending.

Commercial-grade instruction following

Translate detailed creative briefs into more controllable brand, product, advertising, editing, and multi-shot video results, with improved handling of text and brand information.

V2V motion transfer and editing

Reference an existing video to guide body movement, camera behavior, performance, or editing rhythm while applying the creative structure to a new subject or scene.

MiniMax H3 Use Cases Built Around Outcomes

Choose a mode based on what you need to hold constant and what you want the model to invent.

Product ads with directed movement

Generate 2K product reveals, branded scenes, dynamic posters, or ecommerce ads with coordinated visuals, native sound, clearer instruction following, and improved presentation of text and brand information.

Consistent character videos

Use multiple character and wardrobe images to communicate identity across a short narrative beat or branded scene.

Motion and camera studies

Provide a short video reference to guide movement, framing, camera behavior, or pacing while describing your own subject.

Film titles and multi-shot storytelling

Direct connected shots, atmosphere, camera progression, dialogue, sound effects, and music as one audiovisual sequence instead of treating every modality as a separate task.

Music and performance concepts

Combine a performer or character reference with music, voice, or timing instructions to create native stereo audiovisual performance concepts in one generation.

Game, UI, and dynamic design

Create game cinematics, animated interfaces, dynamic posters, product-design concepts, and UI/UX motion studies from a detailed multimodal brief.

MiniMax H3 Specifications

The MiniMax H3 capabilities that matter when you create videos on Media.io.

Official model name MiniMax-H3
Output resolution 2K in the official API documentation
Native audiovisual output Video with native stereo audio, including unified modeling of voice, sound effects, and music
Duration Integer duration from 4 to 15 seconds
Creation modes Text-to-video; first-frame or first-and-last-frame image-to-video; reference-to-video
Image references Up to 9
Video references Up to 3; each 2–15 seconds; combined video-reference duration no more than 15 seconds
Audio references Up to 3; each 2–15 seconds; combined audio-reference duration no more than 15 seconds
Mixed reference maximum Up to 12 image, video, and audio reference files combined
Aspect ratios Prompt- and input-dependent; available choices may vary in the Media.io interface
2K detail rendering Uses in-context regeneration with the original multimodal context to improve small text and fine visual detail at 2K
Commercial strengths Complex instruction following, text and brand-information presentation, V2V motion transfer, multimodal editing, and native multi-shot audiovisual generation
Important restrictions Choose the creation mode that matches your inputs. First/last-frame controls and mixed-reference controls are separate workflows in the current documented interface.

See the official video generation guide, API reference, and model release notes. Media.io may expose a subset of documented controls.

MiniMax H3 vs. Seedance 2.0

Both models support multimodal references and native audio. The practical choice comes down to commercial precision versus cinematic complexity.

Comparison MiniMax H3 Seedance 2.0 Edge
Input and reference control Text, image, video, and audio context; up to 9 images, 3 videos, 3 audio clips, and 12 mixed files Text, image, video, and audio references with comparable 9-image, 3-video, and 3-audio limits Tie
Resolution and duration Native 2K output and video up to 15 seconds Up to 15 seconds; the official launch page does not state a public maximum resolution H3
Native audio Native stereo audio with voice, music, and sound effects jointly modeled Dual-channel audio with detailed voice, music, ambient sound, Foley, and parallel audio direction Tie
Commercial presentation Strong focus on instruction following, product and brand information, visible text, advertising, ecommerce, UI/UX, and dynamic design Strong commercial and cinematic output, while official evaluation still identifies text accuracy as an area to improve H3
Motion and camera Strong V2V motion transfer for movement, performance, camera behavior, and editing rhythm Major strength in complex movement, choreography, multi-character interaction, camera planning, and physically plausible action Seedance
Editing and continuation General prompt-based multimodal editing and reference-led regeneration More explicitly documented targeted clip, subject, action, storyline editing, and video continuation Seedance
Start and end control Explicit first-frame and first-and-last-frame workflows Not highlighted as a central official workflow H3
Best fit 2K product ads, ecommerce, branded video, dynamic posters, UI/UX concepts, and controlled commercial content Action-heavy scenes, dance and sports, cinematic storytelling, multi-character interaction, rich sound design, and video continuation Depends

Quick verdict: Choose MiniMax H3 when sharp 2K detail, product and brand presentation, visible text, first/last-frame control, and efficient commercial production matter most. Choose Seedance 2.0 when complex movement, multi-character interaction, cinematic direction, immersive sound design, or video continuation takes priority.

How to Create a Video with MiniMax H3

01

Select H3 and a Creation Mode

Open MiniMax H3 in Media.io, then choose text-to-video, first/last-frame image-to-video, or reference-to-video.

02

Add Frames or References

Upload a first frame and optional last frame, or switch to reference mode for images, videos, and eligible audio. Frame and reference modes cannot be combined.

03

Generate, Review, and Download

Describe the complete task in natural language: which reference controls the character, which video guides motion or camera, how the audio should shape voice or music, and what brand, text, or final composition must appear.

Built for Creators Who Need More Control

These comments describe the broader Media.io workflow experience and the control creators expect from advanced models such as MiniMax H3; they are not presented as verified reviews of the newly launched model.

poster
Daniel Kim
E-commerce Marketer
star

The Media.io workflow makes it easier to organize an idea, references, and creative direction in one place instead of starting from a blank canvas.

poster
Sophia Chen
Social Media Manager
star

For daily content, clear model and input choices can help our team test video directions without adding a steep editing learning curve.

poster
Marcus Lee
DTC Brand Owner
star

What I value in Media.io is accessibility: prompts and reference inputs make advanced video concepts easier to communicate.

poster
Emily Carter
UGC Creator
star

For product content, I want to reference the subject, describe the motion and camera, and quickly review whether the concept fits a social campaign.

poster
Kriti Agarwal
Content and Film Programmer
star

Media.io makes experimentation feel lighter. A guided workflow helps me focus on character, movement, framing, and the intended outcome.

poster
Ryan Patel
Performance Marketer
star

For campaign testing, reference-led generation can explore different motion and camera treatments while keeping the creative brief in view.

poster
Daniel Kim
E-commerce Marketer
star

The Media.io workflow makes it easier to organize an idea, references, and creative direction in one place instead of starting from a blank canvas.

poster
Sophia Chen
Social Media Manager
star

For daily content, clear model and input choices can help our team test video directions without adding a steep editing learning curve.

poster
Marcus Lee
DTC Brand Owner
star

What I value in Media.io is accessibility: prompts and reference inputs make advanced video concepts easier to communicate.

poster
Emily Carter
UGC Creator
star

For product content, I want to reference the subject, describe the motion and camera, and quickly review whether the concept fits a social campaign.

poster
Kriti Agarwal
Content and Film Programmer
star

Media.io makes experimentation feel lighter. A guided workflow helps me focus on character, movement, framing, and the intended outcome.

poster
Ryan Patel
Performance Marketer
star

For campaign testing, reference-led generation can explore different motion and camera treatments while keeping the creative brief in view.

poster
Daniel Kim
E-commerce Marketer
star

The Media.io workflow makes it easier to organize an idea, references, and creative direction in one place instead of starting from a blank canvas.

poster
Sophia Chen
Social Media Manager
star

For daily content, clear model and input choices can help our team test video directions without adding a steep editing learning curve.

MiniMax H3 FAQs

1. What is MiniMax H3?

MiniMax H3 is a general-purpose omni-modal generation model launched July 31, 2026. It unifies text, image, video, and audio context so creators can describe complex generation or editing tasks in natural language, then produce up to 15-second 2K video with native stereo audio.

2. What references does MiniMax H3 support?

MiniMax H3 supports image, video, and audio references within a shared prompt context. You can describe which asset defines the character or product, which clip guides movement or camera work, and which audio guides voice, song, or rhythm.

3. How many reference files can I use?

The maximum is 9 images, 3 videos, and 3 audio clips, with no more than 12 mixed files. Each video or audio reference must be 2–15 seconds, and each type has a combined 15-second limit.

4. Does MiniMax H3 support 2K video up to 15 seconds?

Yes. Official API documentation lists 2K output and integer duration from 4 to 15 seconds. Media.io controls depend on current implementation.

5. Does MiniMax H3 generate audio with the video?

Yes. MiniMax describes H3 as generating native stereo audiovisual output, with voice, sound effects, and music modeled together rather than added as a separate silent-video post-production step. Reference audio can also help define the intended voice, song, timing, or rhythm.

6. Can I combine first/last frames with reference media?

No. Frame image-to-video and reference-to-video are separate modes. A first frame can be used alone, but a last frame requires a first frame.

7. What is MiniMax H3 best used for?

It is particularly suited to advertising and ecommerce video, brand and product presentation, film titles, dynamic posters, game cinematics, UI/UX motion concepts, consistent character scenes, V2V motion transfer, and other commercial workflows that combine visual references, movement, text, and sound.

8. How can I access MiniMax H3 on Media.io, and is it free?

Use the MiniMax H3 page on Media.io and sign in to check availability, controls, and credit requirements. Starter or promotional credits may be offered, but terms can change. See the official guide for model details.

Media.io Online AI Tools Quality Rating:
vote 4.7 (162,387 Votes)