A faceless video is a narrated video where the viewer never sees the person behind it. Explainers, history, finance, true stories and how-tos all work this way. The parts are always the same: a script, a voice, captions, pictures and an edit. What separates a good one from a cheap one is the order you do them in and the checks in between.
This is the method we use for rethink normal, a Canadian finance channel, where every video starts as a written script.
1. Write for the ear, not the page
A script is heard once, at the speed of speech. The reader can't go back a line. So:
- One idea per sentence. If a sentence has two commas and an "and", split it.
- Say numbers the way people say them. "About forty thousand dollars" is easier to follow than "$39,850".
- Mark the pauses. A new line or a full stop where you want a breath. Voices rush where scripts are dense.
- Read it out loud once. Anything you stumble on, the viewer will too.
Keep the script in sections, each with one job: the hook, the problem, the explanation, the example, the takeaway. Sections make every later step easier, because each one gets its own pictures and its own timing.
2. Record the voice first
Everything else is timed to the voice, so it comes first. You can record yourself, hire a voice artist, or use an AI voice. Whichever you choose:
- Check how it says names, places and numbers before you record the whole script. Fixing one word later can mean recording a whole section again.
- Keep one voice for the whole channel. Viewers notice when it changes.
- If the voice costs money per minute, check the script is final before you record. A paid recording of a draft is money spent twice.
3. Time the captions to the voice
Most viewers watch at least part of a video with the sound low or off, so captions do real work. Word-by-word captions, where each word lights up as it is spoken, are easier to follow than blocks of text. Line them up with the recorded voice, not the script: people add and drop small words when they speak, and a good voice does too.
Read every caption once. Speech recognition gets names, numbers and brand names wrong more often than anything else.
4. Plan every scene before you make a picture
This is the step most people skip, and it is the one that saves the most money. Before making any picture, go through the script and write down, for each scene:
- The word it starts on. A picture should change when the idea changes, on the word that changes it, not every five seconds by habit.
- What the picture shows, in one sentence.
- Any words on screen, such as a number or a short label.
- Whether it has been shown before. If the same kitchen, office or person comes back, use the same picture again.
Why reuse matters: in one of our early videos, the plan asked for thirteen pictures of what was really the same scene. Planning first, and marking repeats, would have made one.
A plan written down also lets you spot gaps: a long stretch with no new picture, or three changes in two seconds.
5. Make the pictures in one look
A faceless video looks professional when every picture belongs to the same world. Choose one style and write it down: the colours, the kind of lines, the light, how close the camera is. Use the same words for every picture. If a character appears more than once, keep a reference picture of them and use it each time.
Keep words out of the pictures themselves and add them in the edit. AI pictures still get lettering wrong, and a wrong number on screen undoes a careful script.
Our guide on keeping AI pictures consistent from scene to scene goes further.
6. Put it together and watch it once, all the way
Lay the pictures on the voice using the plan, add the captions, and render. Then watch the whole video at normal speed, start to finish, without stopping to fix things. Note every moment that feels slow, jumps, or shows the wrong picture. Fix them all together afterwards. Watching in pieces hides problems that only show up in a full viewing.
7. Check before you publish
- Facts: every number and claim against its source.
- Rights: you have the right to use every picture, sound and piece of music.
- Disclosure: YouTube asks creators to label realistic content that was altered or made with AI. Check the current rule when you upload.
- Captions: read once more, start to finish.
Where vbe | video build engine fits
vbe is a Mac app being built to run exactly these steps in order: paste your script, see the cost and say yes, render on your Mac. It plans every scene, reuses repeated pictures, and shows the cost of each paid step before anything is spent. See how it works.
Written by Aman, founder of vbe | video build engine and maker of rethink normal. Published 24 October 2026.