Videos
Turn a document into a video
There is an honest thing to say up front. A document you attach here is not transcribed into the video. It is read and summarised into a short description plus a list of key points, and it is that summary the director works from when it plans scenes and writes the script.
One analysis pass, then a summary
Every attached file goes through a single analysis pass that returns what kind of document it is, a one to three sentence plain language summary, the language it is written in, its main topics and its most important points as short sentences. Only the summary and the key points are folded into the brief. Everything else, the fields, the tables, the full text, is not carried into the video generation step at all.
Why it works this way
A video has roughly forty words of speech in it at short length. A ten page document has several thousand. Something has to choose, and doing that choosing as an explicit summarisation step is more predictable than handing a whole document to a director and hoping it picks the right paragraph. The trade is that a detail buried on page seven, unless it is genuinely one of the document's key points, will not make it into the script.
How to steer what survives
Two levers. The brief text is combined with the summary, so anything you name in the brief pulls the script toward it: lead with the retention figure, do not mention pricing, finish on the hiring plan. And the extra context assembled from your documents is capped at 20,000 characters, so attaching five long files gives each one less room than attaching the one that actually matters.
Failure is silent by design
If the analysis of a document fails, generation carries on without it rather than stopping. That is deliberate, so a broken attachment does not cost you the video, but it means a video can come back built only on your brief while you assumed the document was in it. If the result reads suspiciously generic and does not mention anything specific to the file, that is the thing to suspect first.
How it works, in three steps
Step 1
Attach the document with the brief
Up to five documents ride along, and they are read before anything is designed. PDF and plain text are the two document formats accepted.
Step 2
Say what to lead with
Name the two or three things from the document you want in the video. The summary decides what is available, your brief decides the order.
Step 3
Check it mentions the specifics
Read the generated script. If nothing in it could only have come from your document, the attachment probably did not land.
The full walkthrough with screenshots is in the guide Make a promo video with AI voiceover.
Limits worth knowing
- Only PDF and plain text are accepted as documents here. Word, Excel and PowerPoint files are not offered by the upload picker.
- The document is summarised, never transcribed. Exact wording, tables and figures do not carry through unless they are key points.
- Extra context from all your documents together is capped at 20,000 characters.
- A failed analysis is logged and skipped, so generation continues silently without that file.
Questions people ask
Will it quote my document accurately?
It works from a summary, so treat any figure in the finished script as something to check rather than something copied. The director is separately told never to invent a number, which reduces the risk but does not remove your need to read it.
Can I attach more than one document?
Up to five, and their summaries are concatenated into the brief. They share one 20,000 character budget, so two focused documents beat five sprawling ones.
Does the document appear on screen?
No. Documents are read for context only. Images are what appear on screen, and those are uploaded separately with a limit of eight.
What if the document is in another language?
The analysis detects the language from the real text, so a document in Spanish is summarised faithfully. The video is written in whatever language you set in the brief, which does not have to match.
Make your own video
The button opens the generator with this use case already described. Change the wording to match yours, generate, then edit anything you like.
Create a video with OneCraftRelated pages
Turn a one line idea into a video
A single sentence is a legitimate input, and it is worth knowing exactly how much is being invented on your behalf when you use one. The answer is: the arc, the scene roles, every word on screen, every spoken line, the voice, the tone and often the format as well.
Turn a PDF into a video
PDFs are read differently here from other files, and the difference is worth knowing. Rather than pulling out the text layer, every page is rendered to an image at roughly 200 dots per inch and looked at. A scan and a digitally exported file therefore go through exactly the same path.
Turn a script into a video
If you already have an approved script, the default behaviour will annoy you. The promo director rewrites for pace, aiming at five to eight words a scene, because that is what fits a two and a half second cut. Getting your own wording through is possible, but it takes a deliberate second step.
More finished work of this kind is on the video examples hub.