How to make a mini-documentary from interviews
Upload your interviews and footage to Vizard Agent and say what the piece is about. Vizard Agent transcribes the interviews, finds the passages that carry the story, builds them into a spine and cuts the b-roll over them. Finding ninety usable seconds inside two hours of interview is the work, and it is the work being handed over.
What is the short version?
Three moves, and the second is the one that decides whether you get a documentary or a long testimonial. A mini-doc is an argument with a person in it. Vizard Agent will find the strongest passages either way, but what it builds them into depends entirely on whether you named the argument.
- Go to Vizard Agent and upload the interviews and the b-roll.
- Say what the piece is about in one sentence — the point, not the topic.
- Watch it beside the source and confirm nobody's meaning changed.
The rest of this page is what goes into each of those, and where the approach stops working.
What do you need before you start?
Interviews with clean audio, footage to cover them, and one honest sentence about what the piece argues. That last item is not a formality — it is the difference between a documentary and a collection of nice quotes, and it is the one thing Vizard Agent cannot infer from the material.
- The interviews, uncut. Vizard Agent works from transcripts, so send the whole recording rather than your selects.
- Clean audio. Every cut lands on a word boundary from the transcript. A distant mic degrades the edit itself, not just the sound.
- B-roll to cover the cuts. Hands working, the place, the object. Without it the piece is a person in a chair for two minutes.
- The argument, in one sentence. "She kept the bakery open through the flood" is a story. "About the bakery" is a topic.
- A length. Two to three minutes is the usual ask, and long enough for one argument and no more.
What do you type into Vizard Agent?
Vizard Agent ships this as a template. The brackets are straightforward, but the sentence it produces is better when you replace the subject with the point — Vizard Agent structures the piece around what you say it is about, so "about the bakery" and "about keeping it open" cut differently from identical footage.
Prompt
Three variants worth keeping:
When there are two interviewees
When the argument is specific
When you want it narrated
What are the steps inside Vizard Agent?
Three, and the last one is a factual obligation rather than a polish pass. Vizard Agent transcribes the interviews, selects the passages that carry the argument, orders them, cuts the b-roll over them, captions the result and mixes the music under — as one request, so what you review is a finished piece.
- Upload the interviews and footage to the Vizard Agent chat with the prompt above.
- Pick the tier that matches the job. For anything published about a real person the product recommends Max; the free Flash tier is built for quick, simple edits.
- Watch it beside the source. Not for pacing — for meaning. A shortened answer can say something the person did not say, and only reading it against the original catches that.
What does the result look like?
An assembled piece: interview as the spine, b-roll cut over the answers, captions, music mixed under the speech, in the aspect ratio you named. Vizard Agent keeps every source file and the transcripts in the project, so a shorter social cut or a subtitled version costs a sentence rather than another upload of hours of footage.
It is not instant, and this category was not separately measured. The closest measured work — finding moments inside long recordings — runs a median of about 38 minutes end to end, with the middle half between 18 and 61 minutes. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.
When does this not work well?
A mini-documentary is a set of claims about real people, published under your name. That makes the failures here different in kind: a weak trailer is ineffective, a weak documentary can misrepresent someone who trusted you with two hours of their time.
- Meaning shifted by compression. An answer cut down can become a sentence the person would not stand behind. This is the risk that matters, and the reason for the review.
- No argument. Without one, Vizard Agent produces a competent sequence of pleasant quotes. That is a testimonial, and there is a separate guide for it.
- No b-roll. Nothing to cut away to means visible jump cuts or a static talking head for the full length.
- Poor interview audio. Every cut comes from the transcript, so echo and a distant mic damage the edit and not just the sound.
- Answers that need the question. "That was the hardest part" means nothing alone. Ask for the question on screen, or pick answers that stand up.
- Claims made on camera. If your subject states a figure or accuses someone, that is now published. Vizard Agent does not fact-check it and neither does the edit.
- Consent for the actual use. Agreeing to be interviewed is not agreeing to a paid ad. Check the scope, not just the signature.
The first one deserves repeating: nothing in this process will tell you that a legitimate-looking cut changed what somebody meant.
How do you fix a result that came back wrong?
Reply in the same Vizard Agent conversation instead of starting over. The interviews, the transcripts and the b-roll are all still in the project, so you never re-upload and you are not charged again for the transcription Vizard Agent has already done.
| What you see | What to say next |
|---|---|
| A cut changed her meaning | "Keep the 'at first I said no' clause — it is misleading without it" |
| It has no through-line | "Restructure it around why she refused to sell, drop the rest" |
| Talking head for too long | "Cut the workshop footage over her second and third answers" |
| An answer makes no sense alone | "Put my question on screen before that answer" |
| Music fights the speech | "Drop the music under the interview, keep it only on the b-roll" |
How does Vizard Agent compare to editing it yourself?
By hand this starts with a transcript and a paper edit: read two hours, mark the passages, arrange them into an argument, then cut and cover. Documentary editors do exactly this, and the first pass alone runs longer than the recording — which is why most interview footage never becomes anything.
| Vizard Agent | Editing the documentary by hand | |
|---|---|---|
| Transcribing | Part of the same request | A separate tool and a clean-up pass |
| Finding the passages | Automatic, steerable | Reading the whole transcript |
| Building the structure | Automatic from your sentence | The paper edit, by hand |
| Covering with b-roll | Part of the same request | Shot by shot |
| Time to a first cut | Tens of minutes | Days |
| Judgement about meaning | Yours, and it must be exercised | Yours |
Neither replaces the other. Vizard Agent wins on getting the piece made at all and on the mechanical hours; a documentary editor still wins when the argument has to be discovered in the footage rather than stated up front.
Common questions
How long should a mini-documentary be?
Two to three minutes carries one argument. Past five, a piece needs a second thread to stay alive, and asking Vizard Agent for a longer cut without more material usually just loosens the pacing rather than adding anything.
Can it use two or more interviewees?
Yes — upload them all and say so. Vizard Agent will cut between people where they describe the same events, which is the structure that makes a short documentary feel like more than one person talking.
Will it write narration?
Ask for it and Vizard Agent will. Narration is useful for carrying a timeline between answers, but keep it thin: the reason the format works is that a real person is speaking, and narration competes with that.
Can I get a short version for social?
Ask in the same conversation. The footage and transcripts are already in the project, so a sixty-second cut costs a sentence — though it is genuinely a different edit, and worth telling Vizard Agent what the short version should argue.
Do I need a release from the people in it?
Yes, and for the specific use. Nothing about the editing changes who is making the claims or where they can run, and consent for a community piece is not consent for advertising.