MiniMax H3 Max lip sync image aspect ratio: 0.4 to 2.5

Sume's MiniMax H3 Max Lip Sync describes image_url as a public HTTPS still for audio-driven image-to-video with an aspect ratio of 0.4 to 2.5. If that ratio is width divided by height, which the schema does not spell out, a vertical 9:16 still (0.5625), a square, a 16:9 frame (1.78) and a 21:9 frame (2.33) all fit.

Which common ratios fit?

The table applies the schema's 0.4 to 2.5 range to familiar shapes, assuming width over height. Treat the last two rows as a caution, since the schema gives only the range and not the direction.

What should I do with a still outside the range?

Crop or pad it before you submit. A tall banner or a wide strip is not a face portrait anyway, and cropping to the head and shoulders usually puts you well inside the range.

You can also skip the URL: avatar_id or avatar_handle resolves to the avatar identity still on Sume's side, and the three fields are mutually exclusive.

What else decides the output frame?

The resolution field accepts 480p, 768p, or 1080p, default 768p; 2K is not offered. Output length follows the audio, which must be 5 to 14.8 seconds.

See MiniMax H3 Max lip sync API for the whole request, and Sume's models overview for how this route relates to VEED Fabric 1.0.