Your vacancy has been open six weeks. The ad is accurate, the salary is stated, and it has been reposted three times. Applications: eleven, four of them irrelevant. Somewhere in that gap, good people opened the listing and closed the tab. Nobody can see them, because a rejection that never became an application leaves no trace in the ATS.
Recruitment video is usually sold as brand decoration. It is not. It is a filter and a proof document, and it is measured in time-to-hire and ninety-day retention, not in views. This is the operational version of what we do at KAMOD, an AI video studio in Tashkent working with companies hiring across Uzbekistan, the CIS and further afield — what the video has to do, which roles justify one, what belongs in the first fifteen seconds, and how to keep it usable when an employee leaves.
A recruitment video has three jobs and they fight each other
The first job is to filter. Say the hard part out loud — shift work, cold calls, an eight o'clock finish, a site two hours outside the city, a probation period, the actual salary range. Rejecting the wrong applicant in fifteen seconds is a success, not a failure. Every irrelevant application costs a screening call.
The second job is proof. Show the real workplace: the actual corridor, the actual machines, the actual noise. A rendered office that does not exist reads as a promise the company cannot keep, and candidates who visit for an interview notice within four minutes.
The third job is de-risking. Changing jobs is a risk decision, not a desire decision. Candidates are asking who they will sit next to, what week one looks like, whether the manager is a person or a rumour, and whether the last person in this role lasted four months.
These three pull in different directions. A film that maximises applications filters badly and buries HR in screening. A film that filters hard produces fewer applications and a shorter, cleaner hiring process. Before writing a word, decide which number you are optimising: volume of applications, quality of applications, or offer acceptance. You cannot optimise all three in one piece, and pretending otherwise is how recruitment videos end up as ninety seconds of smiling people and no information.
Which roles justify a video and which do not
Four categories pay back, for different reasons.
High-volume, high-turnover roles — call centres, retail floor, delivery, hospitality, production lines. A sixty to ninety second day-in-the-role that filters honestly. Cost per hire is low and volume is high, so a small improvement in retention is worth more than the production cost.
Hard-to-fill specialists — engineers, pharmacists, developers, machine operators on specific equipment. Here the video is not for the public feed. It is a two to three minute deep piece the hiring manager sends directly to eight named candidates, plus a thirty second cut for the job board. The economic logic is simple: one unfilled specialist role for three months costs more than any video.
Graduate and internship programmes. Energy and progression, vertical format, shot with people one or two years ahead of the applicant, not with executives.
Senior leadership. Usually no public video. A sixty second introduction from the hiring manager, sent privately, outperforms a glossy employer brand film, because senior candidates want specifics about mandate and reporting line, not culture slogans.
Skip video entirely when you hire fewer than three people a year, when there is no retention problem to solve, or when the workplace genuinely is poor. A well-made recruitment video accelerates the exit of people who would have discovered the truth in week two. It does not fix a broken team, and if a client asks us to paper over a real problem we say so before quoting.
The first fifteen seconds decide the outcome
Candidates watch the way they scroll. Hook, proof, action. We build the opening like this: seconds zero to three put the role in plain words with location and salary band on screen — warehouse team lead, Sergeli, from six million sum a month. Seconds four to twelve give the single hardest truth of the job, said by someone who does it. Seconds twelve to fifteen name the people they would work with.
Do not open with a logo animation. Do not open with drone footage of the city skyline. Both cost about four seconds and lose a third of viewers before the offer is even mentioned.
Write the script with someone who has done the job, or at minimum with someone who has interviewed four people who have. Vocabulary is the whole game: the words an employee uses for the shift, the tool, the client, the paperwork are what makes a viewer believe the rest. Marketing language — dynamic team, fast-paced environment, growth opportunities — is read as noise and skipped. Specific language is remembered.
One structural rule we apply: every claim in the script must be visible in the frame, or spoken by a person whose name and role appear on screen. If the video says the team is small, show the room. If it says training is real, show the training.
Employees on camera, and the trust problem
The strongest asset in a recruitment video is a real employee who is slightly awkward on camera. A polished actor reads as an advertisement. A nervous warehouse supervisor who fumbles a sentence still reads as true, provided the sentence is his own.
Practically: record ten to fifteen minutes of conversation with each of three or four employees, ask all of them the same five questions, then cut the answers. You need roughly forty seconds of usable speech from each person. Do not hand out scripts. Send the questions a week ahead, in the language they think in, and let them prepare in their own words.
This does not require a crew of six or a full shooting day. Two people, one camera, one light and a lavalier microphone get it done in about ninety minutes per site, during a working shift, without emptying the floor. That matters in a warehouse or a clinic where stopping operations for a day is not an option.
AI enters the production in a specific and limited way: it builds the scenes the client cannot shoot. The city at six in the morning, the machinery close-up, the impossible wide of the site, the transition sequences, the second location that does not exist yet. We composite those around the human footage so the testimony carries the credibility and the generated material carries the scale. This is also where the budget saving is real — the expensive part of a recruitment film was never the camera, it was closing the site for a day and moving a crew.
The trade-off must be stated plainly: synthetic people talking about company culture do not work for recruitment. Viewers detect a generated face in a talking head quickly, and once they do, everything else in the film becomes suspect. Use AI for environments, b-roll, stylised sequences and translated voice. Use real people for anything that functions as testimony.
In Uzbekistan a hiring campaign usually needs Uzbek and Russian, and sometimes English for expatriate or regional roles. The expensive mistake is treating this as three separate productions. It is one production with three delivery passes.
We record testimony in the language the employee is most fluent in, which is almost always the language they do the job in, and subtitle it. Synthetic voice is fine over b-roll and graphic sections; for testimony, subtitles usually beat dubbing, because a mismatched mouth on a talking head undermines exactly the credibility the shot exists to provide. Cut length changes per language — Russian lines run longer than Uzbek ones, so the layout is designed around the longest variant, and the logo is placed where a wrapping headline cannot reach it.
Placement, usage rights and what to measure
Placement usually matters more than production quality. In order of impact for most of our clients: the careers page hero, the job board listing itself, Telegram hiring channels, vertical cuts for Instagram and TikTok, then the offer and onboarding sequence.
That last one is where most companies forget video and where the measurable effect is highest. A ninety second message from the hiring manager watched in the week before the start date reduces no-shows and early exits more reliably than anything on the public feed, and it is cheap because the material already exists.
Measure in the ATS, not in the video platform. Applications per source, quality of applications as judged by your screening pass rate, time-to-hire, and ninety-day retention for the cohort hired through the video. Views tell you the ad ran. Application rate per thousand impressions tells you whether it worked. We do not promise multipliers, because hiring funnels are noisy and a change in application quality usually shows in two to four weeks while a retention effect takes two quarters to read honestly.
Get written consent that names the usage — organic, paid social, job boards, internal — with a defined term, twenty-four months is common, and a defined territory. Consent for a careers page is not consent for a paid ad in another country.
If someone is identifiable and later leaves the company, you generally cannot put their face in new campaigns, and in some markets you should not keep it in existing ones either. The practical fix is cheap if planned: shoot a second version of every scene that features an employee, with b-roll only. Keep the employee-free cut as the evergreen, and keep a registry of who appears in which shot and when their consent expires. Rebuilding a cut is hours of editing. Reshooting it is a new production.
One hard rule: never generate an AI likeness of a real employee. It is a contractual problem, a reputational problem, and in several jurisdictions a legal one, and no production saving justifies it.
Timeline and what drives the price
A single-role video cut from footage you already have: three to five working days. A recruitment piece shot with a small crew at one site for one role: about a day of recording plus seven to ten days to delivery, including two review rounds. A multi-role employer brand set of five to eight short videos in three languages: roughly three weeks. Price follows shoot days, number of languages, number of cuts and how much of the environment is generated rather than filmed — which is why we quote after a one-page treatment instead of before it.
Review rounds are where these projects fail, not production. Two rounds with a consolidated comment sheet and forty-eight hours per round keeps a date. Ten rounds by email with five stakeholders does not.
If you are hiring and the ad is not converting, send us the vacancy text, the location and the salary band. We will come back with a one-page treatment: the hook, three scenes, what we record and what we generate, and a delivery date. KAMOD is an AI video studio in Tashkent working with teams across Uzbekistan, the CIS and worldwide, and we will tell you honestly if a video is not the right fix for the problem you actually have.