A short video does not always need a face on camera. Explainers, product teasers, mood clips, and many YouTube or Story openings work as a scene: a worktop, a street at dusk, hands and objects, a landscape that moves. If you do not have a clip to start from, you can describe that scene in text and generate a short, faceless video, then add a sound effect if the placement will play with audio.
The text to video tool on CanvaPro is prompt-only — there is no photo upload — with aspect choices of 16:9 (the default), 9:16, 1:1, 4:3, and 3:4. You click Generate Video, wait a few minutes, and download the clip when the queue finishes; the page does not show an in-page preview. For audio, the AI sound effect generator takes a prompt of up to 500 characters, usually returns in about a minute, and is meant for a short effect, not a song or a full voice track.
Write a scene, not a vague mood
Name a place, a time of day, a camera move, and what should not appear. “Top-down shot of hands pouring matcha into a ceramic bowl on a wooden table, slow steam, morning window light, no face, no readable brand type” is easier to direct than “aesthetic faceless video.” If you need a 9:16 Story, say so by picking that aspect, and describe a vertical composition — a tall window, a narrow worktop — so the model is not inventing empty side walls around a wide landscape.
Faceless still means faceless. If you do not say so, a figure may wander in. Mention “no people” or “hands only, no faces” when that is the point. Do not ask the tool to recreate a real news event, a real person, or a real crime scene. Invented footage of actual events is a poor idea and a confusing one for viewers.
The tool-page walkthrough to generate a short video from text covers aspect choices and prompt structure in more detail.
A practical sequence
- Pick the aspect for the placement: 16:9 for many YouTube cuts, 9:16 for Stories and Reels, 1:1 for some feed units.
- Write a concrete prompt with the scene, the light, the motion, and a clear “no faces” or “hands only” line, then generate and wait for the download.
- Watch the clip all the way through. Look for a face that crept in, typography that is almost readable, or a product that is not yours.
- Add a short sound if you need one — a pour, a quiet room, a bird at a window — with a prompt under 500 characters, then mix it in your own editor. Do not ask the SFX tool for a song.
Pick the aspect, write a concrete “no faces” prompt, then generate with the text-to-video tool and watch the whole download before you add sound.
What these clips are good for
They work as b-roll, as an opening mood, or as a background under a voice you record yourself. They are weaker as proof. A generated kitchen is not your kitchen, and a generated street is not footage of a place you visited unless you say so carefully. If you will run the clip as an ad, the usual review still applies: no invented claims, no implied endorsement, no likeness you do not have a right to use if a person appears by mistake.
Because there is no in-page preview, budget time to download and watch every take. A second generate with a tighter prompt is often cheaper than trying to repair a clip that already has a face in the corner.
If you already have a still to animate instead of a text-only scene, use the image to video tool and keep this prompt-only path for shots you cannot photograph.
A short checklist
- Choose 16:9, 9:16, or another listed aspect first, and describe a scene that fits that frame.
- Say “no faces” or “hands only” when that is the brief, and avoid prompts that recreate real events or real people.
- Download and watch the whole clip, then add a short effect — not a song — if the placement has sound.
- Treat the result as generated b-roll, and keep the prompt you used so you can run a cleaner take if you need one.
Voiceover, captions, and a small library
Most faceless videos still have a voice or a caption, even if no one appears. Record the voice yourself, or write captions in your own editor, and cut the generated picture to that timing. Do not expect the clip to be the length of a full script on the first try; treat it as a shot you may loop, slow, or sit under a longer line. If you need two shots — a pour, then a finished bowl — generate them as two prompts, not as one paragraph that tries to be a whole scene change.
Save the prompts that worked, with the aspect next to them. A small library of matching, faceless clips is more useful than a single lucky take you cannot remake. When a clip will run as an ad or a lesson opener, watch it once with captions on, because a stray bit of invented type in the picture can clash with your real title.
A short, well-described scene with honest sound will usually serve a faceless opening better than a long, vague prompt. When you are ready, start with one placement and one concrete shot, watch the download, and only then build a small set of matching clips.
A short pour, a click, or a room tone can be added after the picture is right; see how to generate a sound effect from text.