Choose one of four faceless formats (a voiceover over your own footage or stock clips, animated text, a slide or screen walkthrough, or an AI avatar presenter). Then write a 60-second script with AI, add your own voice or an AI voice, assemble it in an editor like Descript, Canva or CapCut, and add captions. The first one takes about two hours.
"Not on camera" doesn't have to mean "not real". The faceless videos that work for small businesses usually still show something true: your premises, your hands at work, your product, your staff's voices. AI fills the gaps you'd otherwise need a presenter or a crew for, such as the script, the voice, the captions and some of the visuals. The format least likely to build trust in a personal-service business is the one most tools push hardest: a synthetic presenter talking to the camera.
Four faceless formats and the jobs each suits
| Format | What it is | Best for | Weak spot |
|---|---|---|---|
| Voiceover over real footage | Phone clips of hands, tools, premises or products, with narration | Services, trades, local businesses, recruitment | Needs 15-20 minutes of filming |
| Animated text and stills | Words, photos and simple motion, often with music and captions | Tips, offers, short social posts | Can feel generic if the images are all stock |
| Slide or screen walkthrough | A narrated sequence of slides, diagrams or a screen recording | Explaining a process, a price list, software | Dull if slides are text-heavy |
| AI avatar presenter | A synthetic person reads your script | Training, internal updates, multilingual versions | Weakest for trust in personal services |
Most businesses end up mixing the first two: real footage where it matters, text and stills for the gaps. The comparison of what these approaches cost against a filmed shoot is in AI marketing video cost versus a shoot.
Script first: a 60-second structure that works without a face
With no presenter to hold attention, the script and the first few seconds do all the work. A structure that suits most short marketing videos:
- Opening line (first 3 seconds): the problem or question the viewer has, said plainly. No "Hi, welcome to…".
- Why it matters (10 seconds): one sentence of consequence.
- Three points (35 seconds): each with a matching visual.
- Proof (5 seconds): a fact, a number you can stand behind, or a customer's words used with permission.
- One action (5 seconds): what to do next, with the same wording as the caption or link.
A prompt that produces a usable first draft:
Write a 60-second voiceover script (about 150 words) for a
[business type]. Audience: [who]. Goal: [one action].
Structure: opening line naming the viewer's problem (no greeting);
one sentence on why it matters; three points; one line of proof
from the facts below; one call to action.
For each line, suggest a matching shot I can film on a phone
without showing anyone's face, or a simple text graphic.
Facts I can stand behind: [list]
Words to avoid: [list]
Plain spoken English, short sentences, nothing I'd feel silly
saying out loud.Here's an illustrative draft opening for a car repair garage's video, and the fix. The AI wrote: "Hey there! Are you looking for reliable, affordable car servicing you can trust? We've got you covered!" That's a string of clichés in two sentences and no actual problem. The rewrite, from the owner: "That grinding noise when you brake? It won't fix itself, and it gets more expensive the longer you leave it." It names a real sound the viewer may have heard, and the shot suggestion (a close-up of a worn brake pad next to a new one, in a mechanic's gloved hand) needs no face at all.
Read every script aloud before recording. Anything you stumble over, cut or rewrite. AI-written scripts often run long for their stated time. At a natural speaking pace a minute holds about 150 words, so if yours runs over 160, trim it.
The voice: record it yourself or use an AI voice
Your own voice, recorded on a phone, is often the better choice: it's free, it's recognisably you, and local customers may know it. Record in a small room with soft furnishings (a car parked somewhere quiet works surprisingly well), hold the phone about a hand's width from your mouth, slightly to one side, and record each paragraph separately so a mistake means redoing ten seconds, not sixty. Editors like Descript can remove filler words and background noise afterwards.
If you'd rather not use your own voice, AI voices are good and cheap. ElevenLabs is the best-known: its free plan doesn't include a commercial licence, so business videos need at least the Starter plan (listed at $6 a month when checked in September 2026), which includes commercial use and instant voice cloning; higher tiers add professional cloning. Descript, HeyGen and Synthesia also include stock AI voices.
AI voices stumble on names, places and jargon. Fix pronunciation by spelling words phonetically in the script you give the voice tool, not the one you publish. An illustrative example from a care agency script: the voice said "Hilde" as "hild" rather than "hil-duh", and read "24/7 cover" as "twenty-four slash seven cover". The fixes were "Hill-duh" in the voice script, and writing "round-the-clock cover" instead. Listen to the whole track once at normal speed before assembling; mispronounced names are the most common reason viewers realise a voice is synthetic.
Visuals without a presenter: footage, stock, AI clips and avatars
In order of how much trust they tend to build:
- Your own footage, no faces. Hands doing the work, tools, the van, the front door, the product being packed. Ten to twenty short clips of three to eight seconds each, filmed in one session, will cover several videos. Hold the phone steady, film in landscape and vertical if you'll need both, and get anyone who appears to agree first, even if only their hands are visible.
- Stock footage. Useful for generic scenes (a motorway, a kitchen, a laboratory bench), and built into Canva, Descript and most editors. Use it sparingly; viewers recognise stock.
- AI-generated clips. Tools such as Google's Veo can generate short clips from a text prompt. Google AI Pro ($19.99 a month) includes 1,000 monthly credits for Flow, Google's AI video studio, plus video generation in the Gemini app within usage limits. Clips are short, so plan on using them as cutaways rather than whole videos. OpenAI's Sora app closed on 26 April 2026 and its API followed in September, so don't build a workflow around it. Never generate footage that looks like your real premises, staff or customers; that's where AI video crosses into misleading.
- AI avatars. Synthesia's free Basic plan gives 10 minutes a month with 9 avatars but no downloads; its Starter plan is $29 a month, or $18 a month billed annually, for downloadable videos. HeyGen's free plan allows three videos a month up to a minute each; its Creator plan is $29 a month ($24 billed annually). Avatars fit training material and translated explainers. For a marketing video in a personal-service business, whether to use an AI avatar as your spokesperson is worth reading first.
A home-care provider's recruitment video, week by week
An illustrative home-care provider needs more carers and wants a short recruitment video for social media and its jobs page, plus a separate explainer for families. Nobody in the office wants to be on camera, and clients can't be filmed: their homes and faces are private, and many couldn't meaningfully agree anyway. That rules out the usual "day in the life" video, and makes a faceless format the right choice rather than a compromise.
Week 1: script and shot list
The manager gives the script prompt the real facts: guaranteed hours, paid travel time, a mileage rate, training provided, the area covered in general terms, and three things current carers say they like (from staff surveys, with permission). The AI's 60-second draft opened "Are you passionate about making a difference?", which every care recruitment ad says. The rewrite opens with the thing applicants actually worry about: "Paid for your travel time between visits. Not every care job is." The shot list has no faces: a car pulling up, a hand ringing a doorbell, a kettle, a rota on a phone screen, a training session filmed from behind.
Week 2: film and record
Two carers volunteer to be filmed from behind and to have their hands shown; they sign a simple consent note covering the jobs page and social media. Filming 18 clips takes 40 minutes, all in the office, the car park and a staff member's own kitchen standing in for a client's. One senior carer agrees to record the voiceover, reading the script in her own words where she wants to. The team decides against an AI voice for this video: a real carer's voice is the most persuasive thing in it.
Week 3: assemble, caption and publish
The office manager assembles it in Descript: drop in the voice track, edit by deleting words in the transcript, lay clips over each line, add captions and export three versions: vertical for social, square for the jobs page, and landscape for the website. The families' explainer is done differently: a slide walkthrough of how care starts (assessment, care plan, first visit, reviews), with the same senior carer's voice and simple icons instead of stock footage.
Illustrative costs and time for both videos:
| Item | Cost | Time |
|---|---|---|
| Script drafting and edits (existing chat assistant) | $0 | 2 hours |
| Filming on staff phones | $0 | 40 minutes |
| Voice recording | $0 | 30 minutes |
| Descript Hobbyist, one month | $24 (monthly billing) | 3 hours editing |
| Total | $24 | About 6 hours |
The families' explainer does the quieter job of answering "what happens when we call you?", which is the question the office hears most. The recruitment video gets reused in job adverts and at the end of every recruitment email.
A car repair garage explainer from phone footage and a voiceover
For a garage, the natural faceless format is "show the part, explain the problem". An illustrative 45-second video, "Three noises you shouldn't ignore", uses the owner's voice and footage filmed in the workshop in 20 minutes. The script, after one edit for length:
[Shot: brake disc spinning, close-up]
That grinding when you brake? Usually the pads are worn down to
the metal. Leave it, and you're paying for discs as well.
[Shot: hand pointing at a worn drive belt]
A squeal when you start the car on a cold morning is often the
drive belt. It's a cheap fix, until it snaps.
[Shot: wheel, then a hand spinning a bearing]
A hum that gets louder with speed can be a wheel bearing.
Worth a look before a long drive.
[Shot: the garage sign]
Hear any of these? Book a free ten-minute check. Link below.Two checks before it goes out. The mechanic confirms each diagnosis line is fair ("usually", "often", "can be" are doing important work), and the "free ten-minute check" is a real offer the front desk knows about. Captions go on for every word, because a lot of people watch social video with the sound off, and the garage's phone number stays off screen, since the link does the job.
A laboratory explainer built from slides and animated text
Some businesses have almost nothing filmable that's safe or interesting to show. An illustrative food testing laboratory can't film its clients' samples, and most of its instruments look like beige boxes. The format that works is a slide walkthrough: "What happens to your sample", in six steps, each a simple diagram with one line of text, narrated by an AI voice.
The AI helped in three places: turning the lab's internal procedure into six plain-English steps, suggesting a simple icon for each, and writing the narration. It also got one thing wrong that the lab manager caught. Its draft said results were "verified by two scientists", which is how the AI assumed labs work; this lab's process has one analyst and an automated check. The rule for any explainer about how you work: the person who does the process reads the script before anyone records it.
Assembling, captioning and exporting
Three tools cover most small-business needs:
- Descript edits video by editing its transcript, which is the easiest way for a non-editor to cut a voiceover. The free plan allows 60 minutes of media a month and exports at 720p with a watermark; Hobbyist is $16 a month billed annually ($24 monthly) for watermark-free 1080p exports; Creator is $24 a month billed annually ($35 monthly).
- Canva suits text-and-stills videos and quick social edits, with a large stock library and templates.
- CapCut is popular for vertical social video, with automatic captions and templates. Check its commercial-use terms for any templates and music you use.
Whichever you use, export for where the video will play: vertical 9:16 for short-form social, square or 4:5 for feeds, 16:9 for websites and YouTube. Burn captions into social versions, keep lines to five or six words on screen at a time, and check them against the script, because automatic captions mishear names. The practical side of keeping all this consistent across videos, from caption style to intro cards, is in keeping AI-made videos on-brand, and short-form specifics are in making TikToks and Reels with AI.
Disclosure: when viewers need to know AI was used
Using AI to write a script or generate captions doesn't need a label. Showing viewers something realistic that isn't real can. YouTube, for example, asks creators to disclose content a viewer could mistake for a real person, place or event when it's made with altered or synthetic media, and doesn't require disclosure for clearly unrealistic or animated content or for AI used in production such as scripts and captions. Other platforms have their own labels and detection, and the platform-by-platform practicalities are in how to label AI-generated images and video.
Two lines not to cross in any case. An AI avatar or voice must never be presented as a real customer or employee giving a testimonial. And a cloned voice must belong to someone who has agreed in writing to it being used, for the purposes you're using it for.
Why faceless videos flop, and the fix for each
- Nothing happens in the first three seconds. A logo animation or "Hi everyone" loses viewers before the point arrives. Open on the problem, with a moving shot.
- Every clip is stock. Viewers can tell, and it tells them nothing about you. Even three of your own clips change how the whole video feels.
- The voice rushes. AI voices read at an even, fast pace. Slow the speed slightly, and add half-second pauses between points.
- Mispronounced names. Fix with phonetic spelling in the voice script, and listen to the whole track before assembling.
- The wrong shape. A landscape video posted to a vertical feed shows as a small strip. Export each version separately.
- A vague ending. "Get in touch today" gives the viewer nothing to do. Name the action and make the caption, link and voiceover say the same thing.
Start with one format and one video, publish it, and look at how far people watch in the platform's analytics before making the next. If viewers drop off in the first few seconds, rework the opening; if they drop off in the middle, cut a point. For a broader judgement on whether AI video is good enough for your business at all, is AI video good enough for small business marketing weighs up the trade-offs.
Further reads
- How to Plan a Month of Marketing Content With AI in One Hour — Fit video into a month of planned content in one hour.
- How to Build a Brand Voice Guide That AI Can Follow — Keep scripts sounding like your business, not a template.
- How to Spot Deepfake Voice and Video Scams Aimed at Your Business — The flip side of voice cloning: scams to watch for.
- How to Turn One Customer Story Into a Month of Marketing With AI — Turn one customer's story into scripts, with their permission.
- How to Write Facebook and Instagram Ad Copy With AI — Write the ad text that goes around your new video.
- How to Turn Customer Reviews Into Marketing Copy With AI — Pull script lines from what customers already say about you.
- How Much Does It Cost to Create an Online Course With AI? — Itemised costs for an AI-assisted online course, three worked budgets, the hours you'll really spend and the fees that start after launch.
- Marketing a Physio Clinic With AI: Posts, Tips, and Newsletters — A physio clinic's marketing routine built on AI: caseload-led topics, voice notes turned into posts and newsletters, claims checks and booking tracking.
- Best AI Social Media Tools for Small Businesses (2026) — Writing assistants, design tools, schedulers and video editors compared by the job they do, with checked prices, an example for each and three budget kits.
- How to Plan and Run a Webinar With AI: Script, Slides, Follow-Up — A three-week plan for a small-business webinar, with AI drafting the run-sheet, slides and follow-up emails, and a dog groomer's live demo worked through.
- How Much Time Can AI Save on Social Media Management? — A task-by-task breakdown of where AI saves time on social media and where it saves none, with a two-week log to measure your own saving.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Synthesia, HeyGen, ElevenLabs and Descript pricing pages (September 2026); Google AI plans page (Flow credits); OpenAI help notice on the Sora discontinuation; YouTube's official announcement on disclosing altered or synthetic content.