Transparency notice: MiniMax H3 is not connected to ImageMake yet. This page is an independent workflow preview and guide; generation currently runs through Grok Imagine Video 1.5.
Explore the official MiniMax H3 audio-video workflow, study prompt patterns, and test your own video concept in one workspace. ImageMake currently runs this page with Grok Imagine Video 1.5 while direct MiniMax H3 support is being evaluated.
设置图片或视频，并在当前页面直接查看生成结果。
Open a clip to inspect its prompt and send that prompt to the current Grok-powered generator.
16:9 · 10s · 720p
Grok Imagine Video 1.5
The generator above is fully usable, but it does not call MiniMax H3 today. Every prompt submitted on this page is routed to Grok Imagine Video 1.5 using ImageMake's current video settings.
Prompt gallery
Hover to preview motion. Open any card for the complete prompt, video settings, and a one-click handoff to the generator above.
These ImageMake library clips illustrate useful prompt patterns and page interactions. They are not represented as MiniMax H3 outputs; recreating a prompt uses the current Grok engine.
Model overview
MiniMax H3 is a general-purpose multimodal generation system designed to understand text, images, video, and audio together, then generate video with synchronized stereo sound. Its official materials describe first-frame, last-frame, first-and-last-frame, and full-reference workflows, plus a regeneration path for higher-resolution 2K output.
4–15s
Official output duration range
24 FPS
Official output frame rate
32 kHz
Native stereo audio
Up to 2K
With H3 Regenerate-2K
Synchronized audio-video
Official MiniMax H3 materials describe joint audio-video generation instead of treating sound as a separate finishing step. A strong prompt can coordinate the visible action, camera movement, dialogue, physical sound, ambience, and non-diegetic music on one timeline.
Flexible reference control
H3 Base FL2VA covers text-to-video plus optional first and last image anchors. H3 Base Ref2VA expands the workflow to reference images, video clips, and audio. That makes the model relevant to both fast ideation and tightly directed production briefs.
Context-aware 2K workflow
The official H3 workflow describes a 768p base result and a separate H3 Regenerate-2K pass that combines the result with the original multimodal context. The goal is to recover small details from the source instructions instead of guessing them after generation.
Workflow
The page separates H3 research from the engine that actually generates your video, so you can experiment without losing track of the execution model.
Write a text brief or upload one image for the current Grok workflow. Describe the subject, environment, framing, and starting state before adding motion.
Use a clear event sequence, name the camera move, and specify what must remain consistent. Borrow the structure of a gallery prompt when you need a faster start.
Submit to the current Grok engine, then review the result in your history panel. Refine one variable at a time, such as motion intensity, duration, or framing.
Use cases
Turn a product brief into a controlled hero shot, launch clip, or short campaign variation.
Use a supplied frame as a stable visual anchor, then define believable motion and an ending beat.
Block camera movement, character action, scene rhythm, and audio intent before production.
Write dialogue, ambience, physical effects, and music as coordinated layers on the same timeline.
Combine a character, location, motion source, and audio direction into a more explicit creative brief.
Plan concise 9:16 scenes with one readable action, strong subject continuity, and a clear final beat.
Model comparison
This is a product-truth comparison, not a quality ranking. It shows which capabilities belong to official MiniMax H3 materials and which settings are available in the working generator on this page.
| Model comparison | MiniMax H3 official workflow | ImageMake current Grok engine |
|---|---|---|
| Availability on this page | Not connected to ImageMake yet | Connected and available for generation |
| Inputs | Text, first and last frames, plus multimodal image, video, and audio references in the documented variants | Text prompt with optional single image input |
| Duration and output | Official materials list 4–15 seconds at 24 FPS, with a separate Regenerate-2K workflow | 6, 10, or 12 seconds at 480p or 720p |
| Audio | Official workflow describes synchronized 32 kHz stereo audio | No native audio generation in the current ImageMake route |
ImageMake pricing
The plans below cover ImageMake credits and the currently connected generation engines. They do not include MiniMax H3 API access or guarantee future H3 availability.
非常适合初次使用 ImageMake 的用户
图片创作者最爱
给使用 ImageMake 制作图片的资深用户
Questions
No. ImageMake has not connected MiniMax H3 yet. The usable generator on this page currently routes requests to Grok Imagine Video 1.5.
MiniMax H3 is a multimodal audio-video generation model. Its official materials describe understanding text, images, video, and audio together and generating video with synchronized stereo sound.
Yes, according to the official model materials. They list native 32 kHz stereo audio and describe prompts that coordinate dialogue, physical sounds, ambience, and music.
The official MiniMax H3 model card lists an output duration range of 4 to 15 seconds and a frame rate of 24 FPS.
Yes. The documented FL2VA workflow supports optional first and last image anchors, including first-frame, last-frame, and first-and-last-frame generation.
It is the official higher-resolution stage described for MiniMax H3. It sends the base result and original context back through H3 to regenerate a 2K output with context-aware detail.
No. They are ImageMake library previews used to demonstrate prompt patterns and page interaction. Every card states that it is not being presented as a MiniMax H3 output.
The prompt and compatible settings are filled into the page generator. Nothing is generated automatically; you review the form and choose when to submit it to the current Grok engine.
Start creating
Study the official workflow ideas, choose a prompt pattern, and test your direction with the current Grok-powered ImageMake generator.
Try the current video generator