Marey 5 s 1080p at 24 fps vs Sume catalog clips
Moonvalley's licensed-data Marey makes clips up to five seconds at 1080p and 24 fps. Marey is not on Sume; here is how its limits compare with the catalog.

Marey, from Moonvalley, generates clips of up to five seconds at 1080p and 24 fps, and is described as trained on licensed content. It is not on Sume. If your question is about training-data provenance, Sume's docs do not make such claims for the catalog, so check each model vendor's terms.
What the article reports about Marey
The PetaPixel piece (2025) lists these features and a credit price. Prices and plans may have changed since, so check Moonvalley's site.
| Item | Reported |
|---|---|
| Clip length | Up to 5 seconds |
| Output | 1080p at 24 fps |
| Controls | Camera control, motion guidance, pose transfer, inpainting |
| Credits | From $14.99 for 100 credits |
| Training data | Licensed content |
What Sume documents for clips
Sume's catalog is wider than one model, and limits differ per model. The docs say seedance-2.5 accepts 4-30 seconds at 480p, 720p and 1080p, wan-3.0 accepts 2-30 seconds, minimax-h3 accepts 5-15 seconds, and every other catalog model tops out at 15 seconds. Clips of five seconds fit in the range of most of them.
GET /v1/videos/models lists the id, resolutions, aspect ratios, durations and pricing SKUs per model. Sume does not publish a frame rate setting in the pages reviewed for this post, so check the model before assuming 24 fps.
Provenance questions
Sume's job record gives you a receipt, not a rights statement. A completed job shows its id, model, status and usage cost, and the same job is visible at GET /v1/jobs/{id}/status and /result. That is useful for audit trails: which model made which file, and what it cost.
It does not say what a model was trained on. For clean-provenance requirements, ask the model's vendor for documentation, and write the answer next to the job ids in your own records. A model that is not in the Sume catalog cannot be requested through Sume.
- Marey is a separate product with its own credits.
- Sume receipts show model and cost, not training data.
- Read model limits from the catalog before planning a shot list.
Sources
Related posts
More in Comparisons
- OpenAI Batch has no video endpoint listed: where a video batch goes
The OpenAI Batch guide lists text, embeddings, moderation and image endpoints but no video endpoint. Here is what that leaves for a 100-video holiday job.
- OpenAI batch output order differs from input: how Sume bulk keys rows
OpenAI batch output may not follow input order, so you join on custom_id. A Sume bulk queue lists items by index in submitted order. Here is the join for each.
- Pika's auto model pick vs Sume Auto and pinned ids
Pika's Sept 17, 2026 relaunch lists 25-plus apps, auto or manual model pick and cheap Seedance. How that compares with Sume Auto and a pinned seedance-2.5.
- Pika API Club: 100+ models at reduced pricing, questions to ask first
Pika's API Club (Aug 5) offers 100+ models at reduced pricing, and the Sep 17 relaunch adds audio models. Seven questions to ask an aggregator.
Written by Sume