Answer in brief
A memorised testimonial gives itself away through rhythm, borrowed vocabulary and a fixed stare. Here is who to film, which questions produce a usable answer, how consent and release differ by jurisdiction, and what to hand the editor.
The short answer: specifics persuade, phrasing does not
A testimonial works when the viewer recognises their own situation in the person on screen. That takes one checkable detail: what the situation was before, what the person did, and what changed afterwards. Praise about quality and professionalism never supplies that detail.
The usual request from an owner sounds like "say something nice about us". The customer honestly tries to comply and delivers marketing copy. The viewer hears marketing copy and concludes it was written in your office. The damage to trust outweighs whatever the clip was meant to achieve.
The practical conclusion is simple: you do not write your customer's words. You ask questions they already have answers to, and you keep the passages where they answer in their own vocabulary. Everything after that is an editor's job, not a scriptwriter's.
The rest of this article covers who to put in front of the camera, which questions produce a usable answer, how the consent and release side works when you film a real customer, and exactly what to hand an editor to get back a finished vertical cut.
Why a memorised line is audible even through a phone microphone
A memorised sentence changes the rhythm of speech. The delivery flattens out and the pauses fall where the line ends rather than where the thought ends. The ear catches this even when the viewer cannot explain what bothered them about it.
The second tell is vocabulary. Phrases turn up in the customer's mouth that were never there before: "end-to-end solution", "tailored approach", "a team of professionals". That is the language of your website, not of their warehouse or their kitchen table.
The third tell is the eyes. Someone recalling an event looks away and down, then comes back to the interviewer. Someone recalling a text fixes on one point and stops moving. In a tight vertical frame that difference is usually easy to spot.
None of the three can be fixed in the edit. They are fixed on location, by removing the script, asking a question and waiting. The pause before an answer is not a defect; it is evidence that the person is genuinely remembering rather than reciting.
A testimonial is evidence, not a second advert
An advert and a testimonial answer different questions. The advert explains what you sell. The testimonial answers whether you actually deliver it for people who resemble the viewer. Attempting both in one clip weakens each of the two jobs.
That imposes a limit on content: one person, one claim. When a customer covers deadlines, price and service inside ninety seconds, none of the three is remembered. Separate claims ask for separate clips and separate people.
It imposes a limit on promises as well. Your customer can describe what they received. They cannot promise the viewer will receive the same thing. Lines such as "it will definitely work for you" are worth cutting, because they turn testimony back into advertising.
A case film differs from a testimonial in depth, not in genre. There the person describes the problem, the constraints and the decision that was taken. That version lives on the site and in proposals, while short vertical extracts from it go into the feeds.
Who to put in front of the camera, and who to leave out
Film the customer whose story your own records support: the contract, the correspondence, the delivered result. When what is said on camera drifts from what sits in your folder, the clip turns into a liability rather than an asset.
Film the customer who resembles your buyer. A recognisable name performs worse than a recognisable situation, because the viewer decides that a large company had budgets and conditions they do not have, and stops listening halfway through.
Film the customer who speaks in connected sentences and is not frightened of a lens. This is not about eloquence: someone who answers in short concrete phrases edits far more easily than someone who talks at length about nothing in particular.
Do not film a customer you gave something to in exchange for the testimonial unless you are ready to disclose it. Rules for disclosing paid and incentivised endorsements differ between countries and between platforms, and they belong in your checks before the shoot, not after publication.
What to send the customer in advance, and what to keep back
Send the themes, not the questions word for word, and certainly not the answers. Three or four lines are enough: what you will talk about, how long it takes, where it will be published. That lets the person recall detail without arriving with a memorised text.
Send the logistics: the address or the link, the time, the duration, whether to bring anything, whether there will be makeup. Anxiety about procedure eats the first twenty minutes of the conversation and is clearly audible in the voice on the recording.
Send the consent paperwork ahead of time instead of sliding it across the table on the day. Someone reading a document a minute before recording either signs without looking or starts to worry, and both outcomes are visible in the footage.
Do not send your list of selling points. The moment the customer sees your marketing phrasing they begin repeating it, because they sincerely want to help you. You will get exactly what you asked for, and that is precisely what will sound false.
Questions that produce a usable answer
Ask about the past rather than for an assessment. "What were you doing before this" produces a scene; "how did you find it" produces an adjective. Adjectives do not survive the edit, because there is nothing to verify them with and nothing to illustrate them with.
Ask about the moment of decision. "What made you start looking" and "what were you unsure about" pull the viewer's own objections out of somebody else's mouth, and that works harder than your own answer to the same objections.
Avoid anything that can be answered with yes. A closed question returns one syllable, after which you have to splice a sentence together from two fragments, and the splice is audible. Open with what, how, when, or "walk me through".
Ask the person to fold the question into the answer. "We started because…" instead of "Yeah, basically" is the only way to get a self-contained passage that can later be dropped into the cut without your voice running underneath it.
What to do when an answer does not land: A memorised text is given away by rhythm, borrowed…
Do not feed them the words. The second you say the phrasing you were hoping for, the customer repeats it in your intonation and you are back at a memorised text, except that this text now belongs to somebody else as well.
Re-ask through a different door. The same meaning by another route: "where did that show up", "give me an example", "what was different on the Monday morning". An example is nearly always more concrete than the generalisation the person offered first.
Hold the silence. Two or three seconds is usually enough for the person to add a second sentence, and that second sentence is often the one you came for. Do not fill the gap with nods and "mm-hm", because they land in the audio track.
Shoot the same question twice with a break in between. The second take runs calmer, because the person has heard their own voice by then. Say in advance that there will be several takes, or the repeat reads as "I answered wrongly".
Room, sound and frame: the minimum that decides it
Start with sound rather than the camera. A roughly shot but intelligible answer still works; a beautifully shot and unintelligible one does not. A lavalier or on-camera microphone close to the speaker buys more than any change of lens.
Choose a room with soft surfaces: curtains, a sofa, shelves of books. Bare walls and glass produce a hollow reflection that editing does not remove but only masks, and the masking takes the voice down along with the noise.
Shoot against a calm background and leave air above and below the head. A vertical frame crops the sides and a horizontal one crops the top; if you need both, shoot horizontal and keep the framing loose enough to survive the vertical crop.
Switch off everything that hums: air conditioning, extractor fans, notifications on the phone. A steady hum passes unnoticed in the room because the ear adapts to it, and it becomes obvious in a viewer's headphones.
Consent and release: what gets signed before you record
A verbal "sure, go ahead" is not enough, because a year later nobody remembers the terms and the person may have changed jobs and changed their mind. What you want is a short written document signed before the camera is switched on.
It normally records who was filmed, what was filmed, where it may be published, for how long, whether it may be edited and passed to platforms, and whether anything was paid. The more specifically the platforms are named, the less room is left for argument.
Requirements for such a document differ between countries, and sometimes between regions inside one country. Give the text to a lawyer in the jurisdiction where you publish and where the person lives: this is not the place for a template found online.
Settle the withdrawal of consent separately. Agree in advance what happens if the person asks for the clip to come down: how quickly you act, and what becomes of copies already handed to partners or reposted by them.
Employees, children and sensitive stories
In many legal systems a recording of an identifiable person counts as their personal data, and the rights over it stay with them after publication. The extent of those rights varies, so work on the basis that you must be able to remove or restrict the material on request.
Employees are harder than customers: a relationship of dependence sits between you, and consent given to a manager is easy to challenge later. State plainly that a refusal changes nothing, and do not post the shoot in a work channel as an assignment.
Children appear on camera only with the consent of a parent or guardian, and the rules there are stricter. The applicable age, the form the consent must take and the permitted platforms are set by local law, so check before the shoot rather than after it.
Medical and financial stories, and any story touching on vulnerability, need extra care: under a running camera people say more than they intended to. Show them a draft before publication even when you are not formally obliged to.
Editing: shorten it without rewriting the meaning
The editor's core job is removing pauses, repetitions and filler without reordering the claims. Fragments moved around each other are capable of producing a meaning the person never expressed, and at that point the recording stops being testimony.
Captions are not only for watching without sound. WCAG 2.2 asks for captions on prerecorded video with audio (success criterion 1.2.2, Level A), and audio that starts on its own and runs beyond three seconds needs a mechanism to stop it (1.4.2, Level A).
Keep burned-in captions and a separate track apart. Burned-in text is visible everywhere and cannot be switched off; a separate WebVTT file can be turned off by the viewer and read by machines. Feeds usually want the first, your own site the second.
For the web, deliver MP4 carrying H.264 video and AAC audio: that combination plays in browsers without additional conditions, and the MDN reference on media containers explains which codecs belong inside which container.
What the edit costs, and what to hand the editor
You film, we assemble. Filming is not part of the video-production packages: Single Reel at $70 is assembled from your footage in 1-2 working days and includes 1 round of revisions, captions, a licensed track and exports for every vertical placement.
Reels Pack at $90 covers up to 5 edits in 1-2 working days from footage delivery with 2 rounds on the edit; filming is not included, and you supply the footage. Reels Pack Express at $150 runs 1 working day with a 24-hour turnaround from footage and 1 round of revisions, excluding shooting and licensing beyond the supplied track.
Campaign Edit at $180 is a hero cut plus short derivatives in 3-5 days, with 2 rounds on the cut and 1 on the grade; licensed music is not included and the licence is purchased by you. Monthly Content Engine at $290/mo runs a monthly cycle with volume agreed at the start of each cycle and 30 days notice to stop, excluding filming days and paid stock.
Hand over the source files rather than your own rough assembly: complete takes, the separate lavalier audio, a list of timecodes for the moments you liked, and the exact spelling of names and job titles. Package terms are set out on the service page.
Practical checklist
- Pick the customer whose story your contracts and correspondence can support.
- Send the themes, the logistics and the consent paperwork in advance, but not your selling points.
- Ask about the past and ask the person to fold the question into the answer.
- Put the microphone close to the speaker and switch off everything in the room that hums.
- Sign the release before recording and have the wording checked by a lawyer in the relevant jurisdiction.
- Hand the editor complete takes, the separate audio, and the exact spelling of names and job titles.
Questions and answers
Does a nervous customer need a script?
A script trades nerves for falseness. Give themes and the order of questions instead of a text, open with a simple factual question, and shoot two takes: the second almost always runs calmer than the first.
What does the edit cost if I shoot the footage myself?
Single Reel costs $70: assembly in 1-2 working days, 1 round of revisions, captions, a licensed track and exports for every vertical placement. Shooting, scripting and actors are not included in the package.
How long should a testimonial clip be?
Aim at one claim rather than at a number of seconds. A vertical cut usually fits into 30-60 seconds, while a full case film lives on the site and can run several times longer.
What if the customer will not give figures?
Do not insist and do not supply figures on their behalf. Ask them to describe the change in words: what got simpler, what stopped taking time. An unverifiable number is riskier than no number at all.
Is verbal permission to film enough?
No. You want a document signed before recording that lists the platforms, the term and the right to edit. The requirements for its form differ between countries, so have the text checked by a local lawyer.