$50 in a Sume wallet: cutouts, upscales, transcripts, minutes

At the rates on Sume's API pricing page (read 2026-10-02), a $50 balance covers about 2,222 background removals, 250 image upscales, 5,000 minutes of speech-to-text, 500 minutes of timeline output or 400 music tracks. The wallet is in US dollars, so the arithmetic is a division; there is no credit conversion to do.

The table

Each row divides $50 by the listed price and rounds down. These are the fixed-rate tools; model-priced generation (video and image models) varies by model and is not in this table.

Read the unit column

Units differ and matter. Video upscale is billed per input second, with a 5-second default reserve and a 30-second maximum per job, so 5,555 seconds is across many jobs, not one. Speech to text bills per audio minute and reserves one minute when you omit a duration. Image upscale reserves an estimate of about 16 megapixels of generative output.

Those reserves are why the usage ledger shows reserved before captured: the figure you see first is the hold, and the settled amount follows. The usage docs list the states.

What the table leaves out

Avatar and model-priced video are not rows here because they depend on tier or model. As one data point, the Standard avatar tier without a product image is listed at $0.184 per second, so $50 is about 271 seconds of avatar video, a little over 4 minutes. For a video model, read the per-second rate from GET /v1/videos/models and apply the 1.25 reserve rule from the video docs.

Check before you budget

The rates come from the generated pricing page, which is built from the same constants the API reserves with, but prices change. Look at the page on the day you plan the spend. Balance and top-ups live in the dashboard under Billing & subscription, and GET /v1/balance reads the wallet from code.