MiniMax H3 reference formats: HEIC, MOV, WAV, MP3 limits

MiniMax H3 accepts HEIC/HEIF images, MP4/MOV video and WAV/MP3 audio as references. A pre-flight checklist before you submit minimax-h3 on Sume.

5 min readSume
All posts

MiniMax's H3 page lists the reference media it accepts: images as JPG, JPEG, PNG, WEBP, HEIC or HEIF, video as MP4 or MOV, and audio as WAV or MP3, each with a size and dimension limit. If you send references to minimax-h3 or minimax-h3-max through Sume, check your files against these before you submit, because Sume's own docs only say references must be public HTTPS URLs in supported formats.

The format and limit figures are from MiniMax's Create Video Generation Task, read 2026-10-02. Sume's side is from Video generation and Errors and credits.

What are MiniMax's reference limits per type?

Images: at most 1 first frame, 1 last frame and 9 reference images, up to 30 MB each, 256 to 5,760 pixels per side, aspect ratio (width over height) between 0.4 and 2.5. Videos are reference only: up to 3 clips, 2 to 15 seconds each with 15 seconds in total, up to 50 MB each, 256 to 5,760 pixels, 23.976 to 60 fps. Audio is reference only: up to 3 clips, 2 to 15 seconds each with 15 seconds total, up to 15 MB each. The whole request body is capped at 64 MB.

MiniMax H3 reference media (read 2026-10-02)
TypeFormatsCountPer-file limits
ImageJPG, JPEG, PNG, WEBP, HEIC, HEIF1 first, 1 last, 9 reference30 MB, 256-5,760 px, ratio 0.4-2.5
VideoMP4, MOV350 MB, 2-15 s each, 15 s total, 23.976-60 fps
AudioWAV, MP3315 MB, 2-15 s each, 15 s total

Why does this matter when I call through Sume?

Sume fetches your public HTTPS URL and hands the media on, so an unsupported format fails after you have waited in the queue. Sume's errors table lists input_media_unreachable and image_not_fetchable for media it cannot fetch or mirror, and tells you to check that input is a public HTTPS URL. A file the vendor will not accept is a different failure, and the Sume docs do not list the exact error string for it.

Also remember that Sume documents minimax-h3 as 5 to 15 seconds, while the MiniMax page allows 4 for H3 itself. The Sume catalog is the authority for what Sume accepts; the MiniMax limits tell you what the model can read.

How do I check my files before I submit?

Run a local pre-flight on size and extension. This one covers the checks that need no media library; add dimension and fps checks with your own probe tool.

import os

LIMITS = {
    "image": ({".jpg", ".jpeg", ".png", ".webp", ".heic", ".heif"}, 30),
    "video": ({".mp4", ".mov"}, 50),
    "audio": ({".wav", ".mp3"}, 15),
}

def check(kind, path):
    exts, mb = LIMITS[kind]
    ext = os.path.splitext(path)[1].lower()
    size = os.path.getsize(path) / 1e6
    if ext not in exts:
        return f"{path}: {ext} not accepted for {kind}"
    if size > mb:
        return f"{path}: {size:.1f} MB over {mb} MB"
    return None

for kind, p in [("image", "ref.heic"), ("audio", "voice.mp3")]:
    print(check(kind, p) if os.path.exists(p) else f"{p}: missing")

What about HEIC from an iPhone?

MiniMax lists HEIC and HEIF as accepted image formats, which is useful for phone photos. I did not find HEIC called out in Sume's pages, so the safe route is to convert to PNG or JPEG before you upload and host the converted file on a public HTTPS URL. See the 30 MB, 50 MB and 64 MB limits for how those figures interact with Sume's URL-based inputs.

Sources

Related posts

More in Media tools

All Media tools posts

Written by Sume