MiniMax H3 text-to-video rejects adaptive: send an explicit ratio

On Sume the H3 text-to-video builder accepts 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16 and rejects adaptive. Only reference-to-video allows it. Preflight in Python.

4 min readSume
All posts

Wan 3.0 and Seedance accept auto or adaptive. If you reuse that payload on MiniMax H3 for text-to-video, the Sume builder returns the message "aspect_ratio must be 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16." I read this in the repo on 2026-10-07 and have not recorded the HTTP status code, so catch any 4xx rather than one code.

Where adaptive works

Aspect ratio handling, read 2026-10-07
ModeadaptiveNotes
H3 text-to-videorejectedsend one of six ratios
H3 image-to-videonot sentratio follows the image
H3 reference-to-videoallowedup to 9 images, 3 videos, 3 audios
Wan 3.0auto and adaptive acceptedboth sent as adaptive

A small preflight that fixes the payload before you pay for a job:

ALLOWED = {"21:9", "16:9", "4:3", "1:1", "3:4", "9:16"}
def h3_text_ratio(value, fallback="16:9"):
    if value in ALLOWED:
        return value
    return fallback
for v in ("adaptive", "auto", "9:16", None):
    print(v, "->", h3_text_ratio(v))

Sources

Related posts

More in Developers

All Developers posts

Written by Sume