Use MiniMax H3 on Ezier to turn a prompt, start image, or up to nine reference images into a 2-15-second 1080p or 2K video with generated audio.
Plan the shots, assign each reference image a job, and describe sound at the moment it should happen. MiniMax H3 can then treat the picture and soundtrack as one brief instead of separate production steps.
Create a 15-second cinematic 3D fantasy sequence centered on the same young warrior with black hair, silver-gray eyes, a sculpted silver headpiece, and a dark indigo robe embroidered with fine metallic patterns. Begin with an extreme macro close-up as reflected light moves across his eyes and he looks toward the sky. Pull out into a vast storm-cloud realm where a radiant golden celestial formation opens overhead: concentric rings, flowing runes, and fragments of light rotating around a bright center. Let energy travel through the clouds without obscuring their depth. Return to an intimate three-quarter close-up of the warrior as snow-like particles cross the frame and his expression hardens with resolve. Preserve his face, hair, headpiece, robe, eye color, and lighting direction across every shot. Add deep wind, distant thunder, layered choral texture, and a controlled rising impact timed to the portal. No subtitles, logos, modern objects, costume changes, extra characters, or unstable facial details.
Confirm that MiniMax H3 is active before choosing the duration, resolution, or generated-audio setting.
Start from text, add one start image, or attach up to nine supported reference images. Give every reference a specific job instead of treating them as a loose mood board.
Describe the shot order, camera movement, continuity rules, dialogue, effects, ambience, and final frame. Tie every important sound to a visible action or location.
Choose a 2-15-second duration, 1080p or 2K resolution, aspect ratio, seed when useful, and generated audio preference. Check the displayed credit estimate, then generate.
H3 responds best when the brief says what must stay exact, what may change between shots, and which visible action should produce each sound.
Choose 2K when viewers need to inspect materials, mechanisms, surface finish, product edges, or small production details. Still prompt for controlled light and a readable focal point; resolution cannot rescue an unfocused brief.
Use separate images for identity, front and rear geometry, materials, wardrobe, location, lighting, and camera angle. Name those roles in the prompt so the model knows what each input should preserve.
Describe where dialogue, effects, ambience, and music sit in the scene. A door movement, passing object, changing room, or off-screen source gives H3 a clearer basis for timing and stereo placement.
These controls reflect the model configuration Ezier can deliver now. Use the workbench, rather than specifications from another platform, as the source of truth for each generation.
Review hands, faces, small mechanisms, product seams, patterns, and material edges at full size rather than judging only the overall composition.
Confirm that identity came from the identity reference, materials from the detail reference, and framing from the camera reference. Remove inputs that compete with one another.
A cut should reveal new information, change scale, or move the action forward. If two shots do the same job, simplify the sequence before spending another generation.
Review room tone, effects, left-right placement, transitions, and the relationship between sound and motion. A convincing soundtrack needs more than a readable voice.
The practical value is not one isolated feature. It is the ability to brief visual fidelity, reference consistency, shot progression, and generated sound together.
MiniMax designed H3 to understand creative context across several input types and generate picture and sound together. Ezier currently exposes text, start-image, multi-image reference, and generated-audio controls.
The current workbench supports 2-15-second clips at 1080p or 2K, with Auto plus 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 framing.
Up to nine image references give enough room to separate subject identity, product geometry, materials, location, lighting, and camera language instead of forcing one image to explain everything.
Turn on Generate audio when dialogue, effects, or ambience contribute to the result. Describe their timing and source so they can be evaluated with the visual sequence.
MiniMax H3 in Ezier
Start with one 15-second sequence worth reviewing
Define the shot order, add only the references that protect the idea, choose 1080p or 2K, then review the result for detail, continuity, and sound timing.
Answers about duration, 2K output, reference images, generated audio, prompts, and the controls currently available in Ezier.
Keep testing
Compare a nearby model, finish the clip with tools, or rebuild the source for another render.
AI models
Compare models
Choose a better fit before you generate.
AI video tools
Turn images into motion
Animate a still or create a clip.
AI image tools
Create or edit images
Make, clean up, or upscale stills.
LTX 2.5
4K audio-video workflow
Use for 1080p or 4K video with Fast, Pro, and generated-audio controls.
Wan 3.0
Longer beta video generation
Use for text or start-image video up to 30 seconds.
Hailuo AI
Physical motion
Use for fabric, liquid, and physical motion.
Seedance 2.5
Longer reference-led video
Use for longer scenes with richer reference direction.
Veo 3.1
Cinematic control
Use for cinematic shots and camera moves.
Kling 3.0
Action and motion
Use for action, motion, and story beats.
AI models
Compare models
Choose a better fit before you generate.
AI video tools
Turn images into motion
Animate a still or create a clip.
AI image tools
Create or edit images
Make, clean up, or upscale stills.
LTX 2.5
4K audio-video workflow
Use for 1080p or 4K video with Fast, Pro, and generated-audio controls.
