A launch team expects its Chinese production partner to deliver an initial product ad on Friday. The partner's AI tool allows unlimited generation, so the brief seems to have room for plenty of exploration. Close to delivery, however, requests may still be waiting, simultaneous work may be limited, and returned footage still needs review.
This is a hypothetical engagement, not an ONCE client story. It separates two purchases: permission to keep trying and the capacity to deliver acceptable work on time.
Identify what is actually unlimited
Runway's current Generating in Unlimited Mode guide, checked October 30, 2026, says supported models and tools can generate without using credits, with potentially slower starts during busy periods and limits on simultaneous generations. Credits Mode uses credits, starts faster and supports more simultaneous work.
The guide distinguishes Unlimited Mode from the legacy Unlimited subscription tier and notes that availability can vary with plan and promotional period. A historical plan name cannot establish today's entitlement for a particular model, resolution or session. This article considers the capacity implications, rather than recommending a subscription or guaranteeing its features.
Runway's current pricing page lists monthly credits and parallel generations separately. That is evidence of one provider's service design, not proof that the whole industry uses an identical capacity model.
Keep four counts separate
Attempts available over a cycle, tasks running at once, results returned before a cutoff, and approved usable shots are different quantities. A production promise cannot substitute one for another.
Ten attempts at one movement do not necessarily produce ten deliverable shots. Ten completed outputs do not establish that their product structure, motion and duration suit the edit. A successful first attempt also cannot establish the yield of the next assignment.
ONCE proposes treating software access as an input and approved footage as an output. A submitted request is not a completed shot; a completed generation is not yet an accepted creative asset.
Two lanes: exploration and committed delivery
Let exploration wait, with a decision point
Exploration covers unapproved looks, movement ideas and compositions, provided their waiting time does not jeopardize an existing commitment. Record what is being explored, who chooses the direction and when exploration stops. More attempts do not replace a clear product idea.
For example, a home-appliance campaign could compare two background moods using authorized inputs. If the team keeps revising prompts without agreeing on what the product should communicate, upgrading the account alone will not resolve the creative decision.
Resource deadline work, including necessary revisions
For an approved shot needed by the editor, confirm the actual model and mode, output requirements, existing workload, credit balance and exception owner. A faster paid path is not a provider guarantee of a particular finish time, nor a promise that its output will be acceptable on the first attempt.
Reserve resources for inspection and correction. Spending all available time and credits on first-round exploration leaves the production dependent on an untested assumption that nothing needs fixing. Changing models or specifications also requires a fresh quality and cost check; it is not automatically an equivalent substitution.
Rehearse a burst of work rather than extrapolating a monthly allowance
Choose a plausible group of tasks that might arrive together. Use authorized, nonconfidential test inputs and the actual model, resolution and duration. Base the workload on your needs, not an invented industry standard. No rehearsal was performed for this article.
Record submission, actual start, return, selection, correction and approval times, along with other work already in progress. An idle-system test of one short clip cannot establish performance when several brands request revisions at once.
| Record | Question it answers |
|---|---|
| Outputs returned by the cutoff | Which requests actually finished in time? |
| Shots approved by the cutoff | Which outputs could enter the agreed production? |
| Waiting, failed and correction-needed tasks | What risk and work remain outside the success count? |
| Credits consumed and human effort | Which resources were required to meet this cutoff? |
One rehearsal documents one set of conditions. It does not establish year-round delivery reliability. Add evidence when demand, tools or specifications materially change. With insufficient records, narrow the promise rather than filling unknown waits with an optimistic average.
Maintain a cost ledger without counting included resources twice
A practical starting point is actual period spending = subscriptions + additional resource purchases + labor and post-production + necessary external resources. Include only costs actually incurred and not already included elsewhere.
Do not add an imagined retail charge for credits already included in the subscription. An internal project allocation of the subscription can be useful, but its allocation rule belongs in a separate cost view; it is not another payment.
If a top-up purchase retains unused balance, distinguish cash paid this period from consumption by this project. Compare the same delivery scope, quality standard and deadline, then examine how exploration and committed work use resources. Dividing a monthly fee by a theoretical maximum number of attempts does not produce the cost of an accepted advertisement.
Buy priority and a workable arrangement
A brand can ask its partner which requests may wait, which have an agreed cutoff, who prioritizes simultaneous revisions, and who approves additional expenditure. If results are unsuitable or a tool is unavailable, clarify what can be delivered first and when the schedule will be reassessed.
For the production company, separate exploratory requests from work already promised to clients. Record software constraints, review capacity and output quality independently. An ample monthly allowance does not establish capacity for more urgent work on a particular afternoon.
Brands connected to Chinese supply chains can use this distinction when commissioning ongoing product imagery, rather than interpreting a software subscription as an unlimited production retainer. ONCE's video marketing articles cover adjacent planning questions.
For ONCE's long-term content services, the productizable value would be clear priorities, capacity supported by relevant records and a credible exception process. This is a service-design proposal, not an existing ONCE package or a claim of improved margins.
Sources and scope
Checked October 30, 2026: Runway's Generating in Unlimited Mode and Pricing (neither shows a publication date). Legacy-plan migration material was also reviewed but is not used to infer current mode entitlements.
The two-lane method, rehearsal and cost ledger are ONCE analysis. No generation-speed test, subscription change or client-account access took place. Examples are hypothetical, and the original AI cover illustrates production rather than a client engagement.
