Before an overseas team books flights to a shoot in China, there is a useful production question to ask: what would change if the same work were completed by a local crew with remote approval?
The answer is more specific than “remote is greener.” Some journeys and hotel stays may disappear. Other activities may be added, while the studio, product filming and much of the crew’s work remain. Introducing AI changes the comparison again. It does not remove the need to understand the complete route to an acceptable film.
For brands commissioning content around a Chinese supply chain, ONCE proposes comparing production routes against one shared delivery brief, then examining the activities removed, added and still uncertain. This supports a procurement decision without pretending that incomplete estimates prove an environmental outcome.
What the current research supports
SLR’s June 30, 2026 announcement describes DIMPACT’s commission of Carbon Trust research into AI video emissions. The Carbon Trust’s public research overview highlights variation across models, settings and workflows, alongside limited and inconsistent environmental disclosure. A generic emissions figure for one generated clip cannot resolve a full commercial-production comparison.
AdGreen’s current methodology provides a practical distinction: Early Insights can support creative development and bidding using estimates; Final Footprints use actual activity information at wrap. A route can therefore be assessed before it is commissioned, while the estimate remains distinct from a result achieved in production.
These materials were checked on October 23, 2026. This is an analysis using current resources, rather than a claim of a new launch that day. ONCE’s recommendation is to make the comparison while travel, locations and production methods can still be changed.
Establish the same finished assignment
Start with what both proposals must deliver: the agreed film and versions, necessary product actions, accurate appearance and accessories, local-language needs and acceptance requirements. Keep those requirements fixed while evaluating the route.
A film showing a real product in continuous operation and an AI concept sequence may both be valuable. They do not automatically satisfy the same brief. If one route omits the difficult demonstration, its lower estimated footprint would not establish a more efficient way of completing the original assignment.
Where a product claim needs observable support, retain suitable filming or another verifiable approach. Where imaginative expression is appropriate, consider generated or virtual scenes. Carbon estimates should inform a viable creative plan; they cannot authorize inaccurate product representation.
Build a production-route change sheet
This is ONCE’s proposed comparison method, not an accounting standard or a reported client test. Compare two routes at a time. Keep shared activities in a common baseline, then describe the changes.
| Change when an overseas team attends remotely | What the comparison should record |
|---|---|
| Removed activities | The specific people, flight segments, hotel stays and transfers genuinely cancelled. |
| Added activities | Remote equipment, transmission and workspaces, rehearsals, and additional local personnel or hours. |
| Unchanged activities | The studio, core filming, existing local crew and other work still needed for the brief. |
| Uncertain activities | Contingency travel, extra sample shipments, extended studio time, pickups and digital services lacking sufficient information. |
Consider a hypothetical visit by two brand representatives staying four nights each. Removing the visit removes eight person-nights from that accommodation plan. It does not remove “eight units of carbon,” or establish a percentage reduction for the film. The relevant accommodation factors and the other route changes still need to be assessed under the chosen method.
Likewise, local staffing does not automatically remove equipment freight, and an initial sample shipment may not represent all required logistics. The comparison concerns particular arrangements, not a fixed environmental ranking of “travelling,” “remote” and “AI.”
Keep the method consistent
Use the same method and version, coverage and factor-selection rules for both routes. Preserve the geography, period and assumptions behind the estimate. If one proposal covers only shoot days and the other includes preparation and post-production, reconcile that scope before ranking them.
Lower fees, smaller files and fewer shoot days are not sufficient proxies for emissions. Show commercial costs alongside estimated activity impacts, with their different units and purposes intact.
Count the AI workflow, and state what the estimate excludes
AdGreen’s current Activity Breakdown describes AI Usage in terms of the CO2e associated with inference electricity for generated responses. That component excludes initial model training, storage of generated assets on the model provider’s cloud, and CO2e associated with data-centre cooling water. It is a defined estimate, not a complete environmental lifecycle assessment of AI.
Do not rank a lifecycle estimate for one route against an inference-only figure for another as if their coverage matched. An excluded or unknown activity also does not become zero simply because the table lacks a number.
ONCE recommends logging the work attempted, including discarded outputs: model, media type, settings, duration and response count, together with upscaling, regeneration or other processing. A finished edit containing one generated shot may have required many attempts. Check which processing is already included by the selected tool before adding it separately, and retain unknowns where reliable information is unavailable.
For work on owned or shared equipment, agree how energy will be recorded and allocated under the client’s chosen method. Measurements from one setup should remain specific to that setup. They cannot establish the impact of every model, platform or production. No calculator session, GPU energy measurement or upload of client production records was performed for this article.
Ask whether an additional round would reverse the decision
Remote attendance can require more preparation. A generated sequence can require further attempts. Neither observation establishes that one route always needs more rework.
Beside the base comparison, include a plausible additional-work scenario: a defined pickup session, for example, or another specified batch of generation and compositing. Derive its scale from the production plan, a pilot or confirmed requirements. Where there is no basis, record an unresolved scenario rather than inventing its probability.
A useful relationship is: net route change = estimated emissions added − estimated emissions removed. Shared activities may cancel within a consistent comparison of differences; they still belong in the complete project total.
If unknowns or the additional-work scenario could reverse the direction of that change, record that the lower-emissions route is not yet established. A precise-looking savings percentage would conceal the uncertainty.
The production decision can still proceed on quality, product accuracy, timing and cost. The carbon ranking simply remains unresolved until the missing information is sufficient.
Turn the comparison into a useful commissioning deliverable
Request the common brief, route changes, assumptions and gaps, additional-work scenario and named decision owner. Separate responsibility for supplying activity information, calculating estimates and checking the result. A brand with an appointed measurement team may need organized production inputs rather than a second end-to-end service.
For a remotely approved China shoot, settle sample approval, prelighting or pilot work, monitoring windows and who can authorize changes. For partial AI production, specify which scenes may be generated, which product facts must remain accurate and when unsuccessful attempts should trigger another method. These arrangements help the chosen route remain executable; they are not themselves evidence of a reduction.
Discuss those practical inputs with a China-based video production partner before locking travel. The useful question is which people need to be physically present for this particular assignment, and what credible arrangement replaces their presence.
At wrap, review actual activities against the plan. A difference between the initial estimate and final total is not automatically an achieved reduction, especially if the deliverables changed as well.
Over a longer relationship, retain comparable activity records with their limits. They can eventually support better choices about local collaboration, necessary attendance and alternative production techniques. ONCE proposes this as a capability to build; it is not a claim that ONCE already holds an emissions database, certification or verified client reduction results.
Sources and limits
- SLR, DIMPACT commissions new research to manage the carbon impact of AI video generation, June 30, 2026.
- Carbon Trust, The carbon impact of AI video generation, public research overview checked October 23, 2026. No specific emissions values from the full report are used here.
- AdGreen, Carbon Calculator Methodology and Activity breakdown, checked October 23, 2026. These dynamic pages do not state separate publication dates.
The route-change sheet, additional-work check and accommodation example are ONCE’s recommendations or explicit hypotheticals. This article supports production procurement, not a “zero-carbon” or “greener product” advertising claim. The original AI-generated cover is conceptual, not a client shoot or environmental certification.
