Training Videos: DIY, AI Platform or Agency?

Sooner or later almost every company faces the same question: who actually makes the training videos? The department with a smartphone, a film agency with a camera crew or, increasingly, an AI platform that renders a finished video from a script?

The honest answer is: it depends. But on what can be said fairly precisely. This article draws on two voices who know the topic from different sides: a film production company that shoots training videos for clients, and a provider of learning platforms through which such videos reach employees.

Three routes, one deciding criterion

In its guide "Lernvideo erstellen 2026" (in German), learning platform provider reteach distinguishes three production routes and names the criterion that usually decides between them:

That sounds technical, but it is the heart of the matter. A video that stays unchanged for three years can cost more to make than one that is out of date after every process change.

When an agency pays off

In its article "Schulungsvideos erstellen" (in German), Frankfurt film production company muthmedia describes what an agency brings: a concept that connects learning objectives with visual and didactic means, plus crew, voice-over artists, lighting, sound and post-production, tailored to the audience, corporate design and platform.

That pays off wherever a video stays the same for a long time and has to make an impact: an onboarding film introducing the company and its values, safety briefings that have to stick, and anything with external visibility.

When in-house production is enough and where it stalls

muthmedia states the opposing view itself, with remarkable fairness:

The article names the limits just as clearly: a didactic concept, visual quality or emotional appeal is often missing, and the time needed for planning, shooting and editing is easily underestimated. reteach, on the other hand, says no studio is needed to get started: a smartphone, tablet or laptop with an external microphone is enough, and good sound matters more than a good picture.

Both statements fit together. The bottleneck of in-house production is rarely the camera, but the time for editing, voice-over and post-production, exactly the part nobody can take on in day-to-day work.

The cost almost everyone underestimates

When it comes to cost, reteach shifts the focus from production to what happens afterwards. The recording itself is almost never the expensive part. What is most often underestimated is maintenance over the years, meaning the question of who updates the video when the process changes:

reteach also recommends a rule of thumb that shortens many budget discussions: divide the total cost not by the number of videos but by the number of people trained over the planned lifetime.

Where AI video platforms reach their limits

AI platforms have made the third route popular: write a script, choose an avatar and a voice, let the video render. reteach sees this as a real option but names a limit that is decisive for manufacturers:

In manufacturing, the useful dividing line is therefore not between "in-house" and "agency", but between two kinds of content: what can be told and what has to be shown. For what can be told, script, avatar and voice-over are good tools. What has to be shown needs footage from the real workstation and a way to turn it into a usable video without a shooting day.

The fourth route: AI that works with real footage

Avatar platforms and Clypp both use AI, but in opposite directions. An avatar platform invents the picture: it renders a presenter from a script. It cannot show how a hand sets a torque wrench, where a part has to sit or what a correctly fitted seal looks like, because it has never seen your workstation. Clypp starts from reality: the person who has mastered the process films it at the workstation with a phone or tablet while doing it, and the AI does everything that used to make in-house production stall.

Pauses are cut, the noise of the hall is suppressed, and the recording is split into steps, each with an image, a title and an instruction. An AI voice-over is generated and translated into more than 100 languages, so the same instruction works for the whole team. Faces of colleagues can be blurred automatically, which often matters to the works council. And when the process changes, only the affected step is re-recorded. That is the maintenance problem reteach describes, solved at the root: an update costs minutes, not a new day of shooting.

In other words: Clypp combines the authenticity of in-house footage with the speed of an AI platform, without the editing effort that makes in-house production stall and without the avatar that cannot show the machine. At Rosenberger, more than 500 tutorials were created this way in the first six months. At RATIONAL, Clypp was in active use in production 14 days after kickoff.

How such videos are then sensibly split between learning platform and workstation is covered in our article on learning platforms and video instructions.

Frequently asked questions about training videos

How long does production take?

That depends mainly on the effort for concept, shooting and post-production. In-house production starts faster, but the effort for editing and post-production is easily underestimated.

Can AI avatars replace training videos?

Often for explanatory content, not for manual processes at machines. Those need real footage. Clypp brings AI to exactly that footage: it edits, structures, narrates and translates what was filmed at the real workstation.

Who should record training videos in manufacturing?

The person who has mastered the process. Others can take care of didactics, review and approval, but the hand movement itself is best shown by someone who performs it every day.

Which route is cheapest?

That depends less on the unit price than on lifetime and updating. Dividing costs by the number of people trained over the lifetime and factoring in the effort per change gives a fair comparison.

The question "in-house or agency" can therefore rarely be answered across the board. Sorting training content by how often it changes usually leads to a mix: a few carefully produced films for what lasts, and many short recordings for what keeps changing.

In a nutshell: an agency for the few films that have to last, an avatar platform for content that only has to be told. For everything that has to be shown at the machine, and in industry that is most training content, Clypp is the best choice. It is the only route in this comparison that combines real footage from the workstation with the speed of AI, delivers it in more than 100 languages and keeps it current by re-recording a single step.