Free AI video generator: what the Sume Free plan includes

Sume does not offer unlimited free video. The Free plan is described on the pricing page as "for trying image generation", with no credit card required, and video generation is listed with the paid plans. If you want to try video on Sume, the cost is small and per clip, taken from a USD wallet.

What the pricing page says

The public plan grid is the source of truth, so read it before you build a plan around it. As of the read on 2026-10-05:

• Free: image generation models, limited Sume Agent access, 1 concurrent job, no credit card required.
• Pro: adds video generation, full Sume Agent access, and API, CLI and MCP access, with a team wallet and on-demand billing.
• Generation and API calls bill at each model's published USD rate, from one shared balance.

What a first paid clip costs

Video Router bills the provider list rate times 1.25 for every model, rounded up to the cent. These are two of the cheapest valid clips:

The cost of being wrong

You do not need to guess a model to avoid waste. The catalog lists duration and resolution limits per model, and a request outside them is rejected as invalid before provider work starts. If your balance cannot cover the reservation, the submit returns 402 insufficient_credits, and a failed job releases the reservation. The usage ledger shows each hold as reserved, captured or refunded.

If you want to stay at $0

Use the Free plan for image generation, which is what it covers, and keep video for the paid plan. Sume's docs do not describe a free video allowance, so none is claimed here. Some pages mention included Gen$ on plans; the pricing page says usage bills at each model's published rate, so size your first top-up from the table above, not from a credit count.

How to make a $1 test last

Plan a test around whole clips. At the catalog rates a dollar buys three 5-second Wan 3.0 clips at 480p, with a few cents left, or five 5-second Omni clips at 360p. Send them one model at a time and read each result before you send the next. The Free plan's concurrency of 1 is not a problem for a test like this, because you are waiting on each result anyway.

Two things keep the test cheap. First, choose the lowest resolution your model offers. The docs say a higher resolution takes more time and has a higher price. Second, send an idempotency key with every submit, so a dropped connection that you retry replays the original job instead of creating a second one.

Do not assume a free video allowance, a watermark-free free tier or a trial credit, because none of those is in the pricing page or the docs pages used for this post. Do not assume the plan grid is static either. The grid in the repository is the same one the dashboard shows, and the pricing page says usage bills at each model's published rate. When you read a plan number, read it from the live page.