Pay for the GPU time your run uses.
We rent GPUs for the queue and charge each run for the time it held one, at that card’s hourly price, plus 30%. Nothing is a plan or a pack of credits.
A run on a GPU
+30% on our cost
The rented card’s hourly price × the time your run held it, plus 30%. Cards cost us up to $1.30 an hour: a two-minute run on a $0.90 card is $0.03, or $0.039 with the 30%.
A step with a paid API
+30% on our cost
Some steps call a paid model API — image or video generation on kie.ai, for example. You pay the API’s price for that call, plus 30%.
Free
the GPU starting up and waiting between runs
a run that fails
server steps — resizing, cutting, transcoding
your workflows as forms, pipelines and the API
results kept in Media
One balance for the workspace, in dollars
One credit is one dollar. The balance belongs to the workspace and is shared by Slate, Fleet and Studio.
Top it up in USDT on Tron, BNB Chain, Polygon or Ethereum, from any wallet or exchange. The network fee is the sender’s.
Every charge is a line in Billing that says what it was for, and for every run, what it cost us and the 30%.
The balance does not expire, and what you haven’t spent is refunded on request.
If it reaches zero, you have three days to top up before the work that costs money pauses.
Never Manual is in limited access. We open it to teams one by one and walk you through the beta in Telegram first.