We will not make the video without you in it

You can now type a script and get back a video of a person who does not exist saying it, or of yourself saying words you never said. It works, it is cheap, and it takes about four minutes. LensFloat does none of it, and what follows is the honest reasoning rather than a principle we are posing with.

What those tools are genuinely good at

Starting with the fair version, because a position that misrepresents the alternative is worth nothing.

Generated presenters solve real problems. Translating one training video into fourteen languages without hiring fourteen presenters. A compliance module that changes one clause every quarter and needs re-recording each time. Product copy that updates weekly in a company where nobody wants to be on camera. In those cases the video is a delivery format for information, the presenter is a convention, and synthesising the convention is a reasonable trade.

If that is your situation, use one. We would rather say so than pretend every video is a personal address.

What gets lost when the person goes

Most of the videos we care about are not information delivery. Somebody explaining a thing they understand to somebody who does not, and the value is carried in how they explain it.

A person who knows the material pauses before the hard part. They lean in slightly on the sentence they think matters, get faster when they are enjoying themselves, and put an odd emphasis on a word because of something that happened the first time they got it wrong. None of that is in the script. It is the difference between hearing information and being taught by somebody who has it.

A generated presenter reads the script correctly and evenly. Correct and even is what you want from a safety announcement and what you do not want from a teacher.

The part that will decide this

Here is the practical argument, and we think it matters more than the aesthetic one.

While synthetic presenters are rare, they are impressive. As they become ordinary, being visibly real becomes the thing that carries weight. An audience that has learned to spot a generated face starts discounting everything that looks like one, and the discount lands on your video whether or not you generated it.

Being unmistakably a person, with a room behind you and a voice that does uneven things, is going to be worth more in three years than it is today. Anyone building an audience is building it on the assumption that there is somebody at the other end.

So what we build instead

The reason people reach for a generated presenter is rarely that they want one. It is that recording yourself is annoying: the takes collapse, the slides need driving, the file needs editing, and an afternoon disappears. Removing the person removes all of that at once.

Our answer is to remove the annoyance and leave the person. A script that sits at the lens and keeps pace with you, so you are not memorising or chasing a scroll. Slides and a board on the same stage, so nothing needs driving mid-sentence. The ums and long silences marked and shown to you afterwards, so tidying up takes a minute rather than an evening. A finished file at the end, with no editor to open.

Every one of those is aimed at the same outcome, which is that recording yourself stops being a project. If it takes one sitting, the question of whether to synthesise somebody in your place mostly stops coming up.

Where we draw the line

LensFloat has tools that use a model. They help you write and shape a script, and they mark the ums and the long silences so you can decide what to remove. Those assist the person doing the work.

What we do not do is stand in for you. No cloned voice, no face built from a photo, and nothing that produces a performance you did not give. The explaining is the part worth anybody's attention, so we make that easier to do and easier to finish, and leave it to you.

If the recording itself is what keeps going wrong, that is the fixable part. Start with why the sixth take is rarely better than the first, or the practical guide to reading a script while looking at the camera.