Your product launch ends. The presenter signs off in California; the demonstration team in China leaves the set. Who checks that the production resources have actually stopped—and that the recording you still need has finished safely?
That question belongs in the production brief, not just in the cloud account. For an overseas brand working with a Chinese supplier or demonstration team, usage-based live production can make intermittent infrastructure worth evaluating. But a one-hour programme is not a one-hour production job, and “pay as you go” does not define every billable event.
ONCE’s recommendation is to procure a documented operating window and closeout process alongside the creative work. Compare delivery-equivalent alternatives before assuming that moving the control room changes the total cost in your favour.
A launch announcement changes the questions, not the answer
On 15 October 2026, Cerberus Tech introduced Livelink Studio, describing on-demand cloud production that can be shut down and relaunched for subsequent events under a pay-as-you-go model. Its announcement also describes linked A and B instances running simultaneously for resilience.
Those are supplier descriptions of a product and deployment approach. They do not establish a price for your event, independently demonstrate its reliability, or prove an industry-wide cost reduction. They do give a buyer a useful question: which resources run together, and what commercial terms apply to them?
A different provider offers a concrete warning about assumptions. Tencent Cloud’s cloud director billing documentation, in Chinese, updated 25 February 2026, says closing the browser does not automatically stop its output. Its postpaid output charge depends on output specification, picture type and duration; relay and other services may carry separate charges. These are Tencent-specific rules, not a description of Cerberus billing.
The practical lesson is to ask each supplier what starts and stops each meter. A technical shutdown also need not undo an existing prepaid period or minimum commercial commitment.
ONCE’s run-state ledger: give every resource an exit
The following is a proposed planning method, not a report of an ONCE deployment or customer experiment. Make a row for each chargeable resource in the actual proposal. Record its billing trigger, required operating stages, stop condition, and named person responsible for confirmation.
| Resource to clarify | Question before the show | Closeout to agree |
|---|---|---|
| Production engine and outputs | What starts the meter, and can the rehearsal configuration be restored after shutdown? | Operator records the stopped state and time |
| Redundant engine or backup path | When must it run alongside the primary, and is that already included? | Stop only after confirming no remaining dependency |
| Recording, DVR and relays | Which tasks are required during rehearsal, and what does each meter measure? | Confirm necessary files are finalised and readable before stopping related tasks |
| Retained media and configuration | What remains, for how long, under whose access, and at what charge? | Agree transfer, authorised deletion or continued retention separately |
Add billing increments, rounding, minimums, package inclusions and overage handling to the relevant rows. Do not invent charge lines for services you do not use, or count an included resource twice. Nor should a producer remove an untested backup arrangement simply to shorten a billable window.
Consider a fictional launch with a Chinese factory demonstration and an overseas host. The local camera check happens well before the host’s rehearsal; the public event is the next day. Instead of leaving everything running through the gaps, ask whether approved scenes, source mappings and media references can be retained while unnecessary runtime is stopped. Then include the work of restarting, checking picture and sound, and reverting if recovery fails. A saved configuration is not the same deliverable as a continuously running system.
Assign closeout across the teams. The show lead releases the programme output; the recording lead confirms file completion and usability; the account operator stops the agreed tasks and records their status. One person may hold several roles, but the checks remain separate. Specify who may authorise extra runtime and who tells the client about unexpected usage. The event is not operationally closed just because its audience has left.
Compare like-for-like production, not isolated rates
A useful planning equation is:
Total job cost = configuration and content preparation + crew, site and connection work + contract-calculated metered charges + agreed retention and support.
This is an omission check, not a universal tariff. Calculate each charge using its own terms: runtime, transfer, bandwidth and storage are not interchangeable units. Prepaid packages and commitments require their own treatment. Keep bundled items from being added again. If internal coordination cannot credibly be priced, record the hours rather than inventing a standard productivity rate.
Hold the programme specification constant when comparing existing local equipment, rental and cloud options: sources, languages, destinations, rehearsals, recording deliverables, resilience and response requirements. The remote control room does not automatically supply a prepared sample, good location audio, an available presenter or a dependable site connection.
Grabyo’s discussion of cloud production pricing separates runtime from wider production costs and describes several commercial approaches. We checked it on 19 October 2026; the page did not display an exact publication date. It is a provider perspective, not independent proof of savings on your programme.
Intermittent, repeatable events with checked connection conditions are sensible candidates for a bounded comparison. Continuous demand, usable existing equipment and an experienced local team may support keeping the current arrangement. Neither conclusion can be reached from a cloud hourly rate alone.
For the separate question of what to retain throughout the year, see ONCE’s discussion of annual platform access versus per-event production, in Chinese. Here, the narrower purchase is the operating and closeout plan for one job.
Keep reusable work, not accidental runtime
For a production company, the longer-term opportunity may lie in authorised, maintained scene configurations, demonstration structures, terminology, media references and recovery procedures. These are proposed capabilities to investigate—not claims that ONCE owns a cloud platform or has established savings, margins or client outcomes from one.
Before the next launch, ask the team to mark the resources and owners against six stages: preparation, rehearsal, idle gap, programme, file delivery and shutdown. If a quote supplies only an airtime rate, request the missing operating plan before approving a migration. The useful promise of on-demand production is not merely that it can start when needed. It must also be able to stop cleanly, while leaving the work required for the next show intact.
