True Crime & History YouTube: The Complete AI Production Guide

Most true crime advice tells you to find a gripping case. The 4-act structure, the footage filter that keeps fake stock out, and the cuts that build suspense matter more. Here is the production guide.

Most advice about starting a true crime channel is 1 sentence long: find a gripping case. That single sentence is why so many of these channels stall at 12 videos.

That advice is wrong in a specific, expensive way. Gripping cases are the most saturated material on the platform, and a 12-minute video about a famous murder is competing with a hundred versions of itself.

What separates the channels that grow is not the case. It's the structure, the footage discipline and the cutting.

All three are learnable in an afternoon. None of them are what the guides talk about.

What the guides get wrong

The standard advice tells you to pick a case, write a script, find some footage and add dramatic music.

Follow it and you produce the thing YouTube's advertiser-friendly guidelines were written to catch. Four of the fourteen categories are violence, shocking content, controversial issues and sensitive events, and a true crime video that leads with the crime hits all four before the ninety-second mark.

That's why published RPM figures for this niche scatter so badly. It isn't the niche. It's how most people shoot it.

The fix is structural. Lead with the question, not the body.

The 4-act structure that holds 10 to 15 minutes

Every investigation that works has the same skeleton, whether it's a 1970s cold case or the collapse of a bank.

Act one: the discovery. Something is found or noticed. Two minutes. End on the question the rest of the video answers.

Act two: the investigation. What was tried and what it turned up. Four to six minutes, and this is where you clear the eight-minute threshold that unlocks mid-roll ads.

Act three: the turn. The trial, the arrest, the official conclusion. Three minutes. This is where most channels stop.

Act four: the unresolved part. What still doesn't fit. Two minutes, and it is the single most valuable act in the video.

Act four is why people subscribe. A tidy ending gives a viewer permission to leave, and an honest loose end gives them a reason to watch the next one.

Here's the prompt shape:

A 12-minute documentary investigation into [case], in four acts: the discovery, the investigation, the official conclusion, and the part that still doesn't fit. Measured and neutral throughout. Anchor each act to a dated record. End on the open question, not a summary.

History works identically. Swap "case" for "event" and "official conclusion" for "what the textbooks say", and the structure holds without modification.

The footage problem nobody warns you about

Here's the thing that quietly ruins documentary channels.

Generic stock is fiction. A stock clip of an actor in a trench coat walking through rain is not evidence of anything, and viewers who watch this genre recognise the tell instantly.

Once they clock one fake shot, every subsequent claim in your video is downgraded in their heads. You spent the credits and lost the trust in the same second.

So filter footage to non-fiction and archive sources rather than accepting whatever the search returns first. In VidNext, footage sourcing was 97 credits of the 442-credit project we timed, which is more than any other single step, and it is the step most worth supervising.

Two rules that survive contact with reality.

If the footage doesn't depict the actual thing, it should depict the actual place. A real photograph of the right street beats a beautiful stock shot of the wrong one.

And if neither exists, use a document. Which brings us to the part most people skip entirely.

Documents beat footage in this genre

Investigations are made of paper. Court filings, police reports, newspaper columns, evidence logs, maps.

That's your visual language, and it costs almost nothing to produce.

The Motion panel in the editor carries the pieces this genre runs on. Five of them do most of the work.

Pin Board is the detective board with red string between the pins. Document Zoom pushes the camera into an archival document, which is how you show a filing without asking anyone to read it. Entity Intro gives a subject a dossier-style introduction.

Source Card floats a citation next to the claim it supports. Timeline Moment puts an event on a dated card with a year counter.

Then there are eight map templates, and they do more work than people expect. Map Route animates a journey, Map Zoom reveals from world to street, Location Pin drops a marker.

A route, a distance, a timeline drawn across geography answers questions the narration would otherwise spend forty seconds on.

Infographics billed 9 credits on the project we measured. Nine. That's the cheapest line on the entire invoice, and in this genre it's the one carrying your credibility.

Cutting for suspense

Documentary pacing is a cutting decision, not a music decision.

Hard cuts build tension. Dissolves release it. Most beginner true crime videos are 80% dissolves, which is why they feel soft even when the material is genuinely dark.

The same applies to the writing. If you have a documentary channel whose narration rhythm you admire, you can reference its pacing directly rather than trying to describe it in a prompt.

Three rules.

Cut hard on new information. A revelation gets a hard cut into it, never a fade.

Hold longer than feels comfortable before an answer. Two extra seconds on a still image before the narration continues does more for suspense than any music sting.

Never cut on the narrator's breath. Cut a beat after the sentence lands, so the viewer has time to absorb it.

The style pack that fits is Documentary, and the visual style is true crime. Set both on video one and leave them, because the twentieth video made under time pressure needs to look like the first.

Where to set the footage filter

The setting that does the most for this genre sits in Advanced settings on the creation form, not in the editor.

Narrow Footage Sources to Non-Fiction, which covers archive, news, interviews and documentaries. Or choose No Stock to exclude corporate footage entirely.

That one choice does more for credibility than any amount of polishing afterwards, because it changes what the pipeline is allowed to find in the first place.

That one is set before generation, so it costs you nothing to change.

The templates work at both ends. Under Design & Background on the same form you choose which ones the pipeline may use: Media, Overlays, Charts, Maps and Other, each a separate tab. Then in the editor's Motion panel you drag specific ones onto the timeline after the video exists.

Two more switches on that form matter more for this genre than anything in the editor.

Writing Style has a True Crime / Mystery preset. Its own description is the brief you would have written yourself: deliberate pacing with slow reveals, tension built through atmosphere. Pick it and the narration stops sounding like an explainer.

Professional pacing is a toggle, not a habit you have to remember. Switched on, it hard-cuts most boundaries and reserves transitions for chapter breaks, mood shifts and reveals. That is the cutting discipline from the section above, enforced by the pipeline instead of by you.

So the order is: source filter, writing style, professional pacing, generate. Then add the paper in the editor.

What the production actually costs

Here's the whole bill for the 10-minute project we timed, which is the closest thing we have to a documentary benchmark:

442 credits total. The Newbie pack is $15 for 500 credits, so a full investigation fits inside the starter pack with credits to spare.

Credits do not expire and there's no subscription running underneath, which matters in this genre specifically. Research-heavy videos come in bursts, and a month spent reading court records is a month you shouldn't be paying a platform for.

Built by hand, the same 12-minute investigation runs 16 to 30 hours across five or six tools. That range is an estimate creators report, not something we timed. What we did time is the VidNext pipeline: 65 minutes end to end, of which 22 were human attention, and 43 of those minutes ran unattended in the cloud.

The clock on all of this

The Partner Programme threshold doubles on 1 February 2027, and clearing it beforehand takes roughly ninety-six videos, which is four and a half a week from September and eleven a week from December.

Research-heavy genres feel that curve hardest. Every month you wait costs you about a video and a half per week on the pace you will need later, and in true crime each of those videos needs real reading behind it.

Why act four is also your legal safety

Return to the unresolved act, because it does double duty.

A video that asserts guilt beyond what the record supports is a defamation problem and a policy problem at once. The Partner Programme asks for "the creator's original, authentic insights or perspective", and an accusation you cannot source is neither insight nor perspective. A video that lays out what was concluded and what remains contested is neither.

The honest structure and the safe structure are the same structure. That's unusual and worth taking advantage of.

Reading the script is not optional here

More than in any other niche, Script Review earns its seven minutes in true crime and history.

You are reading for two things. Does every factual claim trace to a record you could show someone? And does the video say anything that the other hundred versions of this case didn't?

Verification scored 81 out of 100 on our project at 18 credits, and it catches factual drift. It cannot catch an implied accusation, and it will not tell you that your framing is unfair.

You have to read it. In this genre a wrong claim is not a correction. It is a person.

What we don't know

We haven't run a true crime channel to fifty videos, and we're not going to pretend our production benchmark is an audience result.

What we measured is one 600-second project: the credits, the clock and the step breakdown. Whether four acts beats three on retention for your audience is something your own analytics will tell you in a month, and nobody publishing a table has that answer for your case list either.

The structure above is what the format's own logic recommends. Test it against your retention graph and trust the graph.

Start with one case you already know

Pick a case you've read about, not one you have to research from zero. You're testing the workflow, not the subject.

Write the four acts as four sentences before you touch the prompt box. If act four is empty, the case isn't ready.

Make the first investigation at vidnext.ai and see what your act four looks like.

---

The advertiser-friendly categories are from YouTube's advertiser-friendly content guidelines and eligibility wording from YouTube Channel Monetization Policies, both as of June 2026. Credit figures and timings come from one VidNext project of 600 seconds, measured 18 June 2026. The 16-to-30-hour manual range is an estimate reported by creators, not a VidNext measurement.