MiniMax H3 reference input limits: 30 MB, 50 MB, 64 MB request
MiniMax H3 caps images at 30 MB, reference video at 50 MB and the request at 64 MB, and says to use public URLs. A pre-flight list, and how Sume takes URLs.
MiniMax's H3 create-task docs set per-file limits: a single image up to 30 MB, a reference video up to 50 MB, and a total request body up to 64 MB, with public URLs recommended over Base64 for large files. Check each file against those numbers before you send it. Sume reference entries are URLs, so the same habit applies: host the file at a reachable URL and send the link.
Limits below are as written on the MiniMax page, read 2026-10-01. They describe MiniMax's own API; read your Sume model's supported_input_references for what Sume accepts.
What are the MiniMax H3 file limits?
| Input | Limits as written |
|---|---|
| Request body | Total at most 64 MB; use public URLs for large files, avoid Base64 |
| Image | JPG, JPEG, PNG, WEBP, HEIC, HEIF; up to 30 MB; 256 to 5760 px; aspect ratio 0.4 to 2.5 |
| Image count | First frame at most 1, last frame at most 1, reference images at most 9 |
| Reference video | MP4 or MOV; up to 50 MB; at most 3; each 2 to 15 s, total at most 15 s; 256 to 5760 px; 23.976 to 60 fps |
| Reference audio | WAV or MP3; up to 15 MB; at most 3; each 2 to 15 s, total at most 15 s |
How do Sume requests carry reference media?
As URLs. In the video endpoint a reference entry looks like { "type": "image_url", "image_url": { "url": "https://example.com/style-ref.png" } } inside input_references. That means the 64 MB request-body limit is mostly about what you embed, and a URL keeps the body small. Only models whose supported_input_references lists a type accept it; minimax-h3-max reports image, video, and audio.
Output envelope is separate: minimax-h3 accepts 5 to 15 seconds at native 480p or 768p on Sume. For the reference side of that model see MiniMax H3 reference-to-video limits.
What should I check before sending?
Run a short pre-flight: file size, pixel range, aspect ratio between 0.4 and 2.5, clip length of at least 2 seconds, and a total across clips of at most 15 seconds. Re-encode or crop anything outside the range first. Then read the model's catalog entry for the reference types and counts that Sume applies.
Sources
Related posts
- MiniMax H3 reference limits: 9 images, 3 videos, 3 audio, 12 total
- Video to video motion transfer with AI: MiniMax H3 reference video
- MiniMax H3 voice reference: match a voice with an audio clip
- MiniMax H3 Max reference images: 2 free, then per image
- MiniMax H3 API: H3 and H3 Max video at native 768p with stereo audio
More in Developers
- Modal 150-second web timeout and 303 redirect: Sume polling
Modal web endpoints return a 303 redirect after 150 seconds. Sume returns 202 with status_url and result_url, so a client polls and follows no redirects.
- Capacity fallback vs Sume queued jobs at full concurrency
Runway's Model Router can fall back to another model at its concurrency limit. Sume instead accepts the job as queued on the same model until a slot opens.
- Nano Banana batch API: Gemini's 24 h batch vs Sume async jobs
Gemini's Batch API trades up to 24 hours of turnaround for higher rate limits. Sume has no batch tier for images: send async or webhook jobs per request.
- Nano Banana Pro 21:9: Gemini's ratio list vs Sume's catalog
Gemini's image docs list 21:9 among ten ratios. Sume's Nano Banana Pro and Nano Banana 2 catalogs include 21:9 too; GPT Image 2.5's list does not.
Written by Sume