Short answer: An AI training video generator should do four things: take your source, write and voice the lesson, let you check it, and keep it true when the source changes. Zıpır does this from a written document (procedure, manual, policy): a narrated, subtitled video where every sentence shows its source. It is not an avatar or camera tool.
What jobs make up a training video, and which one decides the tool?
Details
Short answer: There are four jobs, and the hard ones are usually not the editing.
Say you have a procedure or a product manual and you want people to learn it by watching instead of reading. Putting the video together is the small part. Splitting the document into scenes, writing the narration, and knowing a year later whether the video is still true after the document changed: that is the real load.
- Source (you). Which document, who will watch, what should they be able to do afterwards? You also decide that the document is current and correct; no tool can do that for you.
- Narration and voice (Zıpır drafts, you approve). Zıpır produces a scene plan from your document: in every scene the sentence that will come on screen and its place in the document. No video is made until you approve. No microphone or editing software is needed.
- Checking (you, with the source line). Every scene shows its source line underneath; when the video is done, a proof report is produced.
- Updating (you the document, Zıpır the video). If the document changes, you upload the current file and the video is made again.
When you look at an AI training video generator, the real question is which of these four jobs it takes over.
What should an AI training video generator do? Six questions to ask any tool
Details
Short answer: Test a tool on six questions before its feature list: does it show where a sentence came from, who approves before anything is made, do you start over to fix one frame, what happens when the document changes, how do voice and subtitles come out, and what happens to your document. Zıpır's answer to each is below.
The six questions work for any tool you are considering. Read Zıpır's answer, put it next to the answer of the tool in front of you, and if you cannot find that answer, ask the vendor.
- Does it show where a sentence came from? In Zıpır every scene has its source line underneath, and when the video ends a proof report lists every sentence with its place in the document.
- Who approves before anything is made? In Zıpır no video is made until you approve the scene plan. If a quote cannot be found in the source, its scene is not made at all, and a second model searches the source again for every sentence in the plan.
- Do you start over to fix one frame? In Zıpır, in the finished video you circle the frame you want to fix and write what you want; the change applies only to that scene.
- What happens when the document changes? You upload the current file and make the video again; every sentence is tied again to its line in the new text. The video does not update itself.
- How do voice, subtitles and output come out? Zıpır narrates the video itself and puts a subtitle track in the MP4. You can download the MP4 and use it on your own platform; a share link stays closed until you create one.
- What happens to your document? Documents are not used to train models; the terms of Anthropic, the text-model provider, forbid it too, and for the voice provider verification is in progress.
This is what the first question looks like in a video made in the live product from a 28-page law:
“Bu madde kapsamında verilecek eğitimin maliyeti çalışanlara yansıtılamaz.”
6331.pdf, s. 12, satır 1
In English (our rendering): “The cost of the training given under this article cannot be passed on to employees.” Source: 6331.pdf, p. 12, line 1.
What does Zıpır do for a training video, and what does it not do?
Details
Short answer: It turns a written document into a narrated video that shows its sources. It does not decide whether the document is right, and it does not put a presenter on camera.
It does:
- Reads the document (PDF, Word, Markdown, plain text) with its page and line information and produces a scene plan.
- Narrates the video, puts a subtitle track in the MP4 and shows the source line in every scene.
- Does not make a scene for a quote it cannot find in the source.
- When the video ends, produces a proof report: every sentence with its place in the document and a “found in the source” label.
It does not:
- Zıpır is not a tool for talking to a camera, for avatars or for live lessons; it makes a narrated video from a document and does the narration itself.
- Verify that the document is current or correct; it verifies only that the quote is in the document.
- Decide whether a training is adequate. In occupational health and safety that is the responsibility of the employer and the authorised specialist.
- Put clickable scenes inside the video, or replace hands-on training such as showing a movement at a machine.
Who is this for, and who is it not for?
Details
Short answer: It is for training whose content is already written down and taught the same way every time. It is not for knowledge that only lives in someone's head, or where your own face and voice are the point.
For you if you have a written document you can point to as the source: occupational health and safety and regulation training, procedure and policy training, the manual of a new device or piece of software. See Procedures and Technical manuals for the uses.
Not for you if the knowledge is only with your most experienced person. Write it down first; once it exists as a document, Zıpır can work from it. If you are a teacher making a lesson from your notes, the guide to preparing a lecture video was written for that. The example on this site is a Turkish law and its video is narrated in Turkish; if your documents are in another language, write to [email protected] before you commit.
How do I make a training video from my own documents? Four steps
Details
Short answer: You make it from your own document: upload the document, approve the scene plan Zıpır produces, and get a narrated, subtitled video in which every sentence shows its source. No voice recording, editing software or script writing is needed.
- Pick the document. One subject, one document; PDF, Word, Markdown or plain text. Use the official, current version. The code has text recognition for scanned PDFs, but it has not yet been verified end to end in the live product; a file with a text layer is the safest.
- Upload and choose a format. The product has twenty visual formats: Chalkboard, Whiteboard, Illustrated explainer, Balance, Brand colors, Panel presentation, 3D cutaway, Summary presentation, Isometric, Showcase, Clay, Document walk-through, Timeline, Map story, Paper cut, Pictogram, Mascot host, Cause and effect, Host-led cartoon and Branded workplace. See all of them on the formats page. You can leave the choice to Zıpır.
- Approve the scene plan. The plan lists the sentence each scene will show on screen and its place in the document. Nothing is made until you approve. For reference: the 28-page 6331.pdf (Occupational Health and Safety Law (6331)) became a 8-scene video of 1 minute 13 seconds in the Summary presentation format in the live product on October 10, 2026. Your document will come out with a different number of scenes and a different length, depending on its content.
- Get the video and open the proof report. A narrated, subtitled MP4 comes, and below it every sentence is listed with its place in the document. In the example video 16 of 16 on-screen lines were found in the document.
For each step in detail, see how to turn a PDF into a lesson video; for why the source stays on screen, how to make a video lesson that cites its sources.
What should I check before sending it to employees?
Details
Short answer: Four things: that the sentence is really in the document, that the document is still current, that no exception or warning has been skipped, and that the video serves the purpose of the training. The proof report gives you the first; the other three are your decision.
- Is every sentence on screen in the document? Look at the page and line in the proof report. In the example video 16 of 16 on-screen lines were found in the document.
- Is the document still current? The sentences on screen are quotes from your document, and the page and line in the document are written next to them. But if the document itself is wrong or out of date, so is the video: Zıpır does not verify the content of the document, only that the quote is in it.
- Has an exception or a warning been skipped? A one-sentence quote can hide that an exception is in the next sentence. Look at the line right under it in the document too.
- Does the video serve the purpose of the training? Does what the scenes say match what the employee has to do at work? Only you know that.
For which documents an occupational health and safety training needs and what Zıpır does not decide, see OHS training video: which documents do you need.
What happens to the video when the document changes?
Details
Short answer: You upload the current file and make the video again; every sentence is tied again to its line in the new text. The video does not update itself.
Because the video is made from the document as it was on the day you uploaded it, knowing which sentences a change touches matters: every sentence carries its place in the document, so the old page and line numbers tell you where to look. When you correct the document, you only upload the current file; the recording and the editing are not done again.
How much does it cost to make a training video?
Details
Short answer: Studio is $17 a month for 15 finished video minutes and Team is $59 a month for 60 minutes, in US dollars. Buying is open.
Subscriptions are billed in US dollars: Studio is $17 a month for 15 finished video minutes, Team is $59 a month for 60 minutes. Paying yearly, Studio is $163 and Team $566.
Extra minute packs are sold in US dollars (10 minutes for $17, 30 minutes for $45, 90 minutes for $115), are valid for 24 months from purchase. Buying is open; the pricing page has everything in one place.
To start, sign in at app.zipirapp.com or write to [email protected].
Frequently asked questions
Details
How do I make an AI training video from my own documents?
You make it from your own document: upload the document, approve the scene plan Zıpır produces, and get a narrated, subtitled video in which every sentence shows its source. No voice recording, editing software or script writing is needed.
What should an AI training video generator do?
Test a tool on six questions before its feature list: does it show where a sentence came from, who approves before anything is made, do you start over to fix one frame, what happens when the document changes, how do voice and subtitles come out, and what happens to your document. Zıpır's answer to each is below.
What do I need to make a training video with Zıpır?
A written document (PDF, Word, Markdown or plain text) and a Zıpır account. Making sure the document is current and correct is yours to do.
Can Zıpır make a video with an avatar or a presenter on camera?
No. Zıpır is not a tool for talking to a camera, for avatars or for live lessons; it makes a narrated video from a document and does the narration itself.
Can I make an occupational health and safety training video?
Yes, if you have a written law, regulation or procedure: the video shows the sentences of the document with their source line. But Zıpır does not decide whether the training is adequate; that is the responsibility of the employer and the authorised specialist.
Does Zıpır work as an AI lecture video generator or lesson video generator from my notes?
Yes, if the lecture is written down as notes: upload them like any other document and approve the scene plan. Zıpır does not record you teaching and has no presenter or avatar.
Could the video give employees wrong information?
The sentences on screen are quotes from your document, and the page and line in the document are written next to them. But if the document itself is wrong or out of date, so is the video: Zıpır does not verify the content of the document, only that the quote is in it.
Do I have to make the video again when the document changes?
No. You upload the current file and make the video again; every sentence is tied again to its line in the new text. The video does not update itself.
How much does it cost to make a training video?
Studio is $17 a month for 15 finished video minutes and Team is $59 a month for 60 minutes, in US dollars. Buying is open.
Related guides
Details
- How to make a video lesson that cites sources
- How to turn a PDF into a lesson video
- Training video from legislation: OHS and KVKK
- How to explain a dull legal document in a video
- OHS training video: which documents do you need?
- Zıpır or NotebookLM for a video lesson
- Video lesson from a law PDF, step by step
- Why a training video should cite its sources
- How to prepare a lecture video from your notes
- Proof: every claim of a video sits in its source
- What is Zıpır?


