MiniMax H3 Max

What is MiniMax H3 Max?

MiniMax H3 Max is a hosted AI video generation endpoint built on the MiniMax H3 model family. According to the site, MiniMax created the H3 model family, fal produced the post-trained H3 Max endpoint, and h3-max.com provides an independent interface to that hosted endpoint. The tuned weights are not public, so the model cannot be self-hosted.

The tool accepts a text prompt, a starting image, or a set of reference media and returns a 5–15 second clip with synchronized audio. Output is available at 480p or 768p. The site positions it for short, iterative production rather than maximum resolution, and notes that it does not replace the base MiniMax H3 editing or 2K workflows.

How MiniMax H3 Max works

Users interact with the generator through a web interface. They choose an input mode — Text to Video, Image to Video, or Reference to Video — and supply a prompt. For Image to Video, a first-frame image is required and an end-frame image is optional. A prompt expansion mode can be set to Disabled, Balanced, or Quality to control how the prompt is prepared before generation.

Video version (resolution) and duration are selected before submission. Duration options are 5, 10, or 15 seconds. The generator calculates the required credits before the request is submitted. Picture and stereo audio are generated together in the same pass for the same short take.

Main features

• Input modes: Text to Video, Image to Video, and Reference to Video.
• Synchronized audio: Sound is generated in the same pass as the picture.
• Duration: 5 to 15 seconds per request.
• Resolution: 480p or 768p.
• Prompt expansion modes: Disabled, Balanced, and Quality.
• Reference inputs: Reference images and reference video can be supplied in Reference-to-Video mode.
• Example library: Six example videos with prompts, covering cinematic, pixel art, dialogue, stop motion, slow motion, and product macro styles, each with a Recreate option.
• Turbo variant: H3 Max Turbo is referenced for Image-to-Video rates.

Pricing

The site presents two separate pricing structures: supplied generation rates and site credit packs.

Supplied generation rates (reproduced from a user-supplied rate sheet reviewed September 7, 2026, not a checkout quote on the site):

• Text-to-Video: $0.025 per second at 480p; $0.04 per second at 768p.
• Image-to-Video: same per-second output rates as Text-to-Video.
• Reference-to-Video: generated output $0.08 per second at both 480p and 768p; input images $0.02 per image at 1024×1024 and $0.08 per image at 2048×2048; input video $0.052 per second at 480p and $0.144 per second at 768p.
• H3 Max Turbo Image-to-Video: $0.025 per second at 480p; $0.04 per second at 768p.

Credit packs on this site:

• Free: $0, 10 credits included on sign-up.
• Starter: one-time $9.9 for 99 credits.
• Basic: one-time $29.9 for 370 credits.
• Professional: one-time $99.9 for 1,665 credits.

Credits are used at 1 credit per output second at 480p and 2 credits per output second at 768p. Reference images add 1 credit each; reference-video duration adds 1 credit per second at 480p or 2 credits per second at 768p. The site combines free and purchased balances when checking available credits.

Common use cases

The site lists six kinds of short video people make with H3 Max:

• Product videos: one product, one motion — a pour, unboxing, rotation, or close-up. Five seconds covers a single action; 16:9 for a product page, 9:16 for a feed.
• Social clips and hooks: a single visual hook meant to land in the first second, such as a reveal, transformation, or impossible camera move. Vertical, 5 to 10 seconds, with ambient sound attached.
• Dialogue and talking-head scenes: short performance clips with expression, speech timing, and ambience.
• Cinematic shots: long-form shot briefs coordinating subject, camera path, physical action, lighting, constraints, and environmental sound.
• Pixel art and rule-driven animation: prompts that make obstacles, character actions, parallax layers, visual style, and game audio explicit.
• Stop motion and slow motion: material cues, stepped motion, and separated subject/motion-blur treatment.

Limitations

• Output stops at 768p; the tool is not designed for maximum resolution.
• One request stops at 15 seconds.
• The H3 Max tuned weights are not published for self-hosting.
• The endpoint is hosted; the site is an independent third-party web application and is not the official website of MiniMax or fal.
• The supplied generation rates are not a checkout quote for a finished clip on this site; the credit packs are separate and their prices are not used to calculate or imply a dollar price per generated video.
• Whole-clip estimates are rounded down; actual use depends on duration, resolution, mode, and any reference inputs selected.
• The site advises checking resolution, duration, FPS, and aspect-ratio limits before estimating a run.

Summary

MiniMax H3 Max is a hosted AI video generation endpoint that produces short clips with synchronized audio from text, images, or reference media. It is accessed through an independent interface at h3-max.com, which offers credit packs and documents per-second generation rates. Output is limited to 480p or 768p and 5–15 seconds per request, and the tuned weights are not available for self-hosting.