Midjourney video length: reported 5-21 s, and Sume's ranges
Releasebot's feed reports Midjourney video clips of 5 to 21 seconds. Sume lists no Midjourney model; here are the durations its video models accept.

Midjourney video clips are reported at 5 to 21 seconds. That figure comes from a search result for Midjourney's docs, summarized in Releasebot's Midjourney feed (read 2026-10-02); I could not open the primary page, so treat it as reported.
Sume does not list a Midjourney model in the docs I read, so you cannot get Midjourney clips through Sume. What Sume offers instead is a catalog of video models, each with its own duration range, which you read from GET /v1/videos/models.
What durations do Sume's video models accept?
The ranges differ by model and by route, so read supported_durations before you pin a length. These are the figures stated in the Video docs.
| Model or route | Duration |
|---|---|
| Video Router Auto default | 8 s; clips of 3-10 s at 16:9 or 9:16 |
| Gemini Omni Flash 1.1 | 3-10 s |
| MiniMax H3 / H3 Max | 5-15 s |
| Seedance 2.5 | 4-30 s |
| Wan 3.0 | 2-30 s |
| Other catalog models | Up to 15 s |
How do I ask for a length?
Send duration as whole seconds on POST /v1/videos; the value is validated against the model. A model that does not list your value rejects the request, so check the descriptor first.
What should I do if I need a long clip?
Pick a model whose range covers it, or generate shorter clips and join them. For a first frame you already have, frame_images with frame_type: "first_frame" makes image-to-video, on models that list it in supported_frame_images. The Video generation docs have the request fields.
Sources
Related posts
More in Comparisons
- Ming-Image open weights or a hosted image API: what you operate
Ming-Image-0.1-Design is MIT-licensed and runs on one 80 GiB GPU. This post lists what self-hosting means in practice and what a hosted API call replaces.
- MiniMax Speech 2.8 pitch and emotions vs Sume TTS controls
MiniMax T2A lists speech-2.8 models, nine emotions, pitch -12 to 12 and 10,000 characters. Sume TTS has speed, volume and emotion text, no pitch.
- MiniMax <#1.5#> pause markup vs Sume TTS: no pause marker
MiniMax inserts pauses with <#x#> markers from 0.01 to 99.99 seconds. Sume TTS documents no pause marker; here is how to split a script and what to use instead.
- MiniMax Video Agent template API: submit, poll, download
MiniMax Video Agent builds a video from a template_id plus your media and text. The endpoints, statuses, and what Sume offers for template-driven video.
Written by Sume