Wan 3.0 vs MiniMax H3: clip length, ratios and reference caps
Wan 3.0 runs 2 to 30 s at up to 1080p; MiniMax H3 runs 5 to 15 s at 480p or 768p. Where they differ on Sume: ratios, reference caps and audio inputs.

The two differ mainly in clip length and frame shape. On Sume, wan-3.0 makes clips of 2 to 30 seconds at 480p, 720p or 1080p, while minimax-h3 makes 5 to 15 seconds at native 480p or 768p and adds a 21:9 ratio. Both accept image, video and audio references and can generate audio. This page compares limits only, not output quality.
Sume limits are from Video generation; Wan facts from fal's Wan 3 page and H3 facts from MiniMax's H3 post, all read 2026-09-29.
How do the limits compare?
| Item | `wan-3.0` | `minimax-h3` |
|---|---|---|
| Clip length | 2 to 30 s | 5 to 15 s |
| Resolutions | 480p, 720p, 1080p | 480p and 768p native (768p, not 720p) |
| Aspect ratios | 16:9, 4:3, 1:1, 3:4, 9:16 | Those five plus 21:9 |
| Image references | Up to 10 | Up to 9 |
| Video references | Up to 5 | Up to 3 |
| Audio references | Up to 5 | Up to 3; not as the only reference |
Which is longer in one request?
Wan by a wide margin: 30 seconds against H3's 15. fal's page gives Wan 3 the same 2 to 30 second range, and MiniMax's post says H3 makes up to 15 seconds at 2K, with 768p also available. On Sume the 2K and 4K H3 outputs are upscales, priced if you request them, so the native sizes above are what you generate.
Do both make audio?
Yes. fal says Wan 3 generates audio in the same pass, and MiniMax says H3 produces native stereo sound. On Sume, generate_audio defaults to each model's audio capability, and GET /v1/videos/models reports it per model.
Which ratios matter for vertical?
Both list 9:16, so a Reels or Shorts clip works on either. The one shape only H3 accepts is 21:9. Send the ratio explicitly:
curl -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: wan-h3-001" \
-d '{
"model": "minimax-h3",
"prompt": "A skater crosses an empty plaza at dusk",
"duration": 8,
"resolution": "768p",
"aspect_ratio": "21:9"
}'What is the practical split?
- Need more than 15 seconds in one clip, or under 5: use
wan-3.0. - Need a 21:9 frame: use
minimax-h3. - Need 1080p natively: use
wan-3.0; the H3 variant that lists 1080p isminimax-h3-max, a latent refinement from 768p. - Unsure: read
supported_durationsandsupported_resolutionsfromGET /v1/videos/modelsand test one short clip on each.
Sources
Related posts
More in Models
- Wan 3.0 vs Seedance 2.0: which model to call when
Wan 3.0 and Seedance 2.0 are both on Sume's video API. Wan runs 2 to 30 s and bills per second; Seedance 2.0 runs 4 to 15 s with a 21:9 frame. Pick by the job.
- Wan 3.0 web page to video: link input and how to do it on Sume
Wan 3.0 can read a public web link to make a video on Alibaba's API. Sume's wan-3.0 has no link field, so fetch the page yourself and send text and images.
- An OpenRouter-compatible video API: sume/auto or a pinned model
Sume's POST /v1/videos follows OpenRouter's video generation API field for field. Let sume/auto pick the model, or pin a catalog id like seedance-2.5.
- Image generation API with reference images: POST /v1/images
Send a prompt plus public HTTPS reference images to Sume's POST /v1/images. Pin a catalog model or send sume/auto; the catalog lists each model's limits.
Written by Sume