Ask anyone who runs a small foundation or a CSR programme what their hardest communication problem is, and you will hear a version of the same thing. The work is real, the outcomes are measurable, and almost nobody outside the organisation ever sees it.
The reason is not indifference. It is that showing impact visually is expensive. A field visit with a camera crew costs more than most grassroots programmes spend on their entire annual communications budget, and the footage is often unusable anyway — poor light, no consent for the faces that matter, a monsoon that did not cooperate.
So annual reports stay text-heavy, donor decks stay stock-photo-heavy, and the genuinely moving parts of the work stay invisible.
What actually shifted
Over the last eighteen months the cost of producing short, watchable video has collapsed, and the reason is not cheaper cameras. It is that a usable clip can now be generated rather than filmed, for cents rather than a day rate.
That is a loaded statement in this sector, and it deserves immediate qualification: this does not mean fabricating beneficiaries. It means something much narrower and much less troubling.
Every impact film contains two kinds of shot. There is the material that must be real — the person speaking, the classroom, the borewell, the hands doing the work. And there is the connective material that exists only so the edit makes sense: an establishing view of the district, a map animating, a sequence explaining how a water filtration process works, a visual metaphor for compounding change over ten years.
The second category has always consumed a disproportionate share of production budgets while carrying none of the truth-telling burden. That is the category this technology has taken over.
The line that matters
Sector guidance on this is still forming, but the practitioners thinking clearly about it have converged on a simple rule: generative imagery may illustrate, never testify.
An animated explanation of how your sanitation intervention works is illustration. A generated image of a smiling child presented as a programme participant is testimony, and it is fabrication regardless of how carefully it is captioned. The distinction is not subtle, and the reputational cost of getting it wrong in this sector is severe — far higher than in commercial marketing, because trust is the entire asset.
Practical safeguards that organisations are adopting:
- Label generated sequences in the credits, the same way stock footage has always been credited
- Never generate human faces presented as identifiable individuals
- Keep beneficiary voice and testimony entirely in real footage, always with documented consent
- Put the policy in writing before the first project, not after the first complaint
The economics, plainly
The pricing structure of these tools is unusual and worth understanding before budgeting.
Unlike software you license per seat, generative video is billed per second of output. Every second costs the provider real computation, so a clip you discard costs the same as one you use. And you will discard most of them — the working ratio is roughly one keeper in ten.
This means the meaningful budget number is not a monthly subscription. It is the cost of a hundred attempts. For a communications officer costing out an impact film, the calculation is: how many seconds of generated material does the edit need, multiplied by ten for discards, multiplied by the per-second rate.
Those rates vary considerably between models producing broadly comparable output, so it is worth comparing directly rather than defaulting to whichever tool appeared in a newsletter — the current rates for the Kling API and competing video models give a reasonable sense of the range. For most small organisations the total lands somewhere between a few hundred and a couple of thousand rupees per finished minute of connective footage, against tens of thousands for a crew day.
Where it still fails
Knowing the limits saves money, since every failed generation is billed.
Close-up human faces fall apart — which, given the rule above, is arguably convenient. Hands performing specific actions come out wrong. Text inside the frame is unreliable for anything longer than a word or two. Generating the same location or person consistently across two separate clips remains unreliable. And five to ten seconds is the honest maximum per generation, so the workflow is many short clips cut together, not a continuous scene.
None of these limits matter much for establishing shots, process animations, and abstract sequences. They matter enormously for anything you would want to call documentary.
The governance question nobody is asking yet
There is a second-order issue that this sector, of all sectors, should probably get ahead of.
These models run on data centres, and data centres consume electricity and water. An organisation publishing a sustainability report illustrated with generated imagery is making a small, unacknowledged environmental expenditure to communicate about environmental stewardship. The amounts are genuinely modest at the volumes a small nonprofit would use — orders of magnitude below a single flight to a field site — but the principle of disclosure applies, and the organisations that will look best in five years are the ones that started reporting it early.
There is also a labour question. The videographers and editors who serviced the development sector were often small independent operators, and the shots being automated were frequently the affordable end of their work. A sector that talks about just transitions should be able to think about this one honestly rather than simply pocketing the savings.
Where this leaves a communications budget
The realistic position is unglamorous. This technology does not let a small organisation make a documentary. It lets them stop paying for the boring half of one, and redirect that money toward the half that has to be real — more days in the field, better audio, proper consent processes, an editor who understands the work.
That is a smaller claim than the technology’s promoters make. It is also, for an organisation trying to show what it actually does on a budget that was never adequate, a genuinely useful one.
