Film each move or drill with a real coach, then use AI for everything around it: turn the written class plan into a 60-second script, add captions, a title card and a voiceover, and cut it to length. Avatars and generated footage are fine for intros and announcements but should not demonstrate exercises. Check the music licence before posting.
A class plan is already a script in disguise: warm-up, blocks, rest times, coaching cues, regressions. Most gyms and trainers never turn it into video because editing eats an evening. AI shortens exactly that step. The decision that matters is what you let it make. Let it write, caption and edit, and keep a human body on camera for anything that teaches movement.
Decide what the video is for before choosing a tool
Member videos tend to do one of four jobs, and each needs a different mix of footage and AI. Choose the job first, because a tool that is perfect for one will be wrong for another.
| Video job | Length | Footage | Where AI helps |
|---|---|---|---|
| Class recap ("what we did Tuesday") | 45-60 sec | Phone clips from the class | Script, captions, on-screen reps and rest times |
| Move tutorial ("the kettlebell swing") | 60-90 sec | Coach filmed from front and side | Cue script, captions, slow-motion markers |
| Home follow-along for absent members | 10-20 min | Filmed session | Timer overlays, chaptering, caption track |
| Announcement or welcome | 20-40 sec | Coach to camera, or avatar | Whole script, voice, translation |
Small gyms usually get the best return from the first two: they reuse work already being done. A personal trainer with ten clients might choose the tutorial job to cut repeated explanations, and send the link before each session. The general question of whether this kind of video is good enough is covered in this candid look at AI video for small businesses; the focus here is the class-plan workflow.
Step 1: turn the class plan into a video script
Start with the plan as you already write it. Paste it into an AI assistant and ask for a script sized to the video job, with the coaching cues kept exactly as you wrote them. The risk at this stage is the AI "improving" your programming: adding exercises, changing rest intervals or adding claims about results. Tell it not to.
Below is my class plan. Write a 60-second voiceover script for a
member recap video. Rules:
- Use only the exercises, reps, rest times and cues in the plan.
- Do not add exercises, weights, calorie claims or promises of results.
- Plain spoken English, short sentences, British spelling.
- Include one safety line about stopping if something hurts.
- Mark where each exercise appears with [SCENE: name].
- Finish with one line inviting members to ask a coach about options.
CLASS PLAN:
[paste plan]Say a boutique gym runs a 45-minute strength class. The plan has a five-minute warm-up, then three blocks: goblet squats, bent-over rows and a carry, each as 4 sets with 60 seconds' rest, and a core finisher. The returned script might look like this (illustrative):
[SCENE: warm-up] Start with five minutes of easy movement: arm circles,
hip hinges and bodyweight squats. [SCENE: goblet squat] Block one is
goblet squats. Four sets, rest sixty seconds. Chest tall, elbows inside
the knees. [SCENE: rows] Block two, bent-over rows. Same pattern, four
sets, sixty seconds' rest. Keep your back flat. ...
If anything hurts, stop and tell a coach.Read it against the plan, as you would proofread any copy. Typical fixes: a cue the AI paraphrased into something subtly different ("keep your back flat" when your cue is "long spine"), a rest time that drifted, and an exercise name changed to a more common one. Because it is a script, not a programme, fix it by hand and move on. The wider question of how trainers use AI safely for programmes is covered in using ChatGPT safely for client programmes.
Step 2: choose footage, avatar or both
This is the step that decides whether the video is safe to share. There are three realistic options.
Real footage, AI around it. A coach films each move on a phone, front and side, 15 to 30 seconds each, in decent light and with the whole body in frame. AI adds captions, titles, music and the voiceover. This is the safe default for anything that teaches a movement, and it is the option that still feels like your gym.
Avatar presenter. An AI avatar reads the script on screen. Tools such as HeyGen and Synthesia offer this. HeyGen's pricing page currently lists a free plan with three watermarked videos a month, each up to one minute, and a Creator plan at $29 a month that removes the branding and exports in 1080p. Synthesia lists a free Basic tier with a watermark and a Starter plan at $19 a month (a discounted first month is advertised). These figures were taken from the vendors' pricing pages in October 2026 and they change, so check them before you commit. Avatars suit welcomes, timetable changes, holiday-opening announcements and spoken explanations over footage. They do not suit demonstrations.
Generated footage. Text-to-video models can render a person doing a squat. The result may look convincing for two seconds and wrong for ten: knees that bend the wrong way, a barbell that changes size, a movement no coach would teach. Treat generated exercise clips as unsafe to show members as technique. They are, at most, atmosphere for an intro.
A rule of thumb: if a member could copy what they see, a real coach must be doing it. If a member is only being informed or welcomed, an avatar is acceptable if labelled. For member-facing explainers without a coach on camera, making marketing videos with AI without being on camera compares the options in more detail.
Step 3: edit, caption and time it
Most members watch with the sound off, on a phone, in a changing room or on a train. Captions are not decoration; they are how the video gets watched. Editors such as CapCut list auto-captions and text-to-speech among their features, and Canva's video editor has captioning tools within a monthly AI allowance that is larger on paid plans than on Free (check Canva's pricing page for the current Pro price and allowance). Whatever you use, generate the captions automatically, then correct them by reading the whole track once. Automatic captions regularly mangle exercise names ("Romanian deadlift" can become "romantic deadlift") and numbers ("4 by 8" becomes "for by eight").
Timing rules that keep videos watchable:
- Open on the movement or the result, not a logo. The first two seconds decide whether a member keeps watching.
- One idea per video. A recap shows the blocks; a tutorial teaches one move.
- Put reps, sets and rest on screen as text. Members forget numbers and screenshot them.
- End with one action: "Book Thursday's class" or "Ask a coach about a lighter option".
- Keep a consistent title card and colours so members recognise your clips.
For consistency of captions, voice and intros across every video you publish, keeping AI-made videos on-brand gives a reusable style sheet.
Step 4: the music licensing check
Music is the part gym owners most often get wrong. A track you like is not a track you may use. A video posted to promote classes counts as business use, so personal-use licences do not cover it. Platforms can mute or take down videos with unlicensed music, and rights holders can pursue a business directly.
A practical check, in order:
- Where did the track come from: your editor's library, a stock music service, an AI music tool, or a streaming playlist? Streaming playlists are not a licence.
- Read the licence for your plan. Does it allow commercial use and posting to social platforms? Does it cover paid ads, if you will boost the video?
- Does the licence need credit in the caption? If so, add it.
- For AI music tools, check the plan you are on. Some plans limit commercial use or require a paid tier; confirm on the vendor's page and keep a screenshot with the date.
- Platform rules vary. Some video platforms ask you to disclose realistic AI-generated content in their upload settings, so check the current setting where you post.
Keep a simple log: video title, track name, source, licence type and date checked. If a takedown arrives, you can answer it in minutes. If you would rather avoid the whole problem, many coaches use no music and let the coach's voice and gym sound carry the clip.
Worked example: one Tuesday class becomes three member videos
Say a small gym runs a Tuesday 6:30 pm strength class of 12 members. The head coach spends about 90 minutes a week on video, filming two minutes of clips on a phone during the class and nothing else. Here is the weekly flow (illustrative times).
- Plan to script (10 minutes). Paste the class plan into the assistant with the script prompt. Correct three cues by hand.
- Footage (3 minutes during the class). One coach films the three main lifts, side and front, 20 seconds each, with the camera propped on a bench.
- Recap edit (25 minutes). Cut the clips to the script, add the voiceover (the coach's own, or a text-to-speech voice labelled as such), overlay reps and rest, auto-caption, correct captions. Result: a 60-second recap for members who missed the class.
- Tutorial clip (20 minutes). The best footage of the goblet squat becomes a 60-second move tutorial with three coaching cues on screen, filed in a "how to" playlist.
- Announcement (10 minutes). A 25-second avatar or coach-to-camera note about Thursday's timetable change, with a clear label if the presenter is an avatar.
- Check and post (15 minutes). Watch each video with the sound off, confirm captions and the music log, post to the member group and the gym's social channels.
Total about 85 minutes for three videos, against the several evenings a full manual edit would take. These are illustrative times; your first few weeks will run longer. The pattern holds: coaches film, AI writes and captions, a person checks.
More examples across gym types
A solo personal trainer with remote clients. Each Friday the trainer films three short clips of the week's new exercises and sends clients a private link with the script as captions. The AI writes a cue list per client from the programme, and the trainer removes anything that does not fit that client's injuries. Never paste health details into a public tool without checking its data terms.
A yoga or pilates studio. The class plan lists sequences by name. An AI voiceover reading Sanskrit or technical terms may mispronounce them, so record the cue words yourself and let AI write the rest. A recap video shows the sequence on screen with the name of each pose.
A martial arts club. A drill demonstration needs two people and a clear angle; the AI role is the script, the belt-level labels and the captions.
A corporate wellness provider. A small provider runs desk-stretch sessions for staff at client offices. A set of 90-second stretch videos is a good use of an avatar for the introduction, with a real instructor demonstrating each stretch.
A class with accessibility needs. The same script can be turned into a captioned version with large text, or a slower-paced version with extra rest cues. Ask the AI to rewrite the voiceover for a lower-impact variant, then have a coach review the changed movements, because the safe option is a coaching decision.
Before and after: a script that sounds like a coach
AI scripts tend to sound like a brochure. Compare the first pass with a version edited to match how a coach talks (illustrative).
First pass: "Get ready to raise your strength journey with this dynamic lower-body block designed to challenge every member." It is full of words no coach says aloud, and it carries no information.
Edited: "Block one is goblet squats. Hold the weight at your chest, sit between your knees, and stand tall. Four sets, sixty seconds' rest." It is shorter, it matches the class, and a member can follow it with the sound on or off.
Ask the AI to rewrite in your voice by pasting two or three things you have actually said or written, then tell it to match their sentence length and vocabulary. Then cut every sentence that does not give an instruction, a number or a reason.
A shot list to film from, filled in
Before filming, ask the assistant for a shot list from the script, and take it to the gym floor on your phone. A filled-in version for the strength class might read as follows.
| Shot | Angle | Length | On-screen text |
|---|---|---|---|
| Goblet squat, set 1 | Side, whole body in frame | 20 sec | 4 x 8, rest 60 sec |
| Goblet squat, cue close-up | Front, waist up | 10 sec | Elbows inside knees |
| Bent-over row | Side, whole body | 20 sec | 4 x 10, rest 60 sec |
| Carry | Front, walking towards camera | 15 sec | 3 x 20 m |
| Coach to camera sign-off | Front, chest up | 8 sec | Ask a coach about options |
Film each shot twice, because one take will have a member walking through it. Use a stable surface or a cheap tripod, hold the phone sideways for landscape footage, and clean the lens. Poor footage is the one problem AI cannot fix afterwards.
A quick sum: is it worth the subscription?
Say a coach values their time at $30 an hour and previously spent four hours editing one recap by hand. With a script prompt and an AI editor the same recap takes about 50 minutes, saving roughly three hours, or about $90 of time per video. If the subscription costs $29 a month and the coach makes four videos, the time saved dwarfs the fee. But if the gym posts one video a month and would have skipped it anyway, a free plan or no new tool is the sounder choice. These numbers are illustrative; swap in your own hourly value and video count before deciding.
Safety, consent and honesty checks
Before any video goes out, run a short check. Does every demonstrated move match how you teach it? Does the video suggest outcomes ("lose weight fast", "fix your back") that you cannot promise? Does it include a safety line, and does the wording suit your insurance and waiver terms (ask your insurer or an adviser if unsure)? Did everyone visible on camera agree to be filmed and to have the footage used publicly, with written consent from members and any under-18s? Is any AI presenter or cloned voice labelled? A cloned voice of a real coach needs that coach's clear agreement and the platform's rules on voice cloning respected.
If a video mentions medical conditions, rehab or pregnancy, route the script through a qualified professional. The AI does not know your member, and a general cue can be wrong for a specific body. Keep the videos about the class you ran, not about treating anyone.
Views, rebookings and editing time after a month
Pick two simple measures for the first month: how many members watched the recap (views in your member group or platform stats), and how many missed-class members rebooked the following week. Also count how many questions coaches stop answering by hand because the tutorial exists. If videos take more than 90 minutes a week, simplify: drop the avatar, reuse one template, film fewer clips. If nobody watches, shorten them and move the action line to the first five seconds.
If the video work spreads into membership questions and follow-ups, the receptionist-style help in the gym AI receptionist cost tutorial shows what the next step might cost. For a view of cost against filming everything conventionally, this comparison of AI video and a shoot gives a way to count it.
Questions trainers ask about AI workout videos
Can an AI avatar demonstrate exercises safely?
Not reliably. Avatars and generated footage often show impossible joint angles or loads, and a member copying bad form can get hurt. Film yourself or a coach doing the movements, and use AI for the script, captions, titles and voiceover around that footage.
Do I need permission to use music in member videos?
Yes, unless the track comes with a licence that covers your use. Music in a video you post for a business is commercial use. Use tracks from a library whose licence you have read, or an AI music tool whose plan says commercial use is allowed, and keep a note of the licence.
How long should a member workout video be?
For a class recap or single-move tutorial, 45 to 90 seconds works for most members. A full follow-along session can run 15 to 30 minutes, but that is a different job, and filming it properly beats generating it.
Should I tell members a video was made with AI?
If an avatar or cloned voice stands in for you, say so in a line on the video or caption. Members trust a coach, so do not let them think a synthetic presenter is you. Footage of your real coach with AI-written captions needs no special label.
Further reads
- ChatGPT Fitness Coach Prompt: Client Programme Template for Trainers — A prompt template for the plan itself.
- AI Software for Personal Trainers: Scheduling, Check-Ins, Billing — Scheduling and check-ins around the videos.
- AI Prompts for Personal Trainers: Session Recaps and Renewal Texts — Session recap and renewal texts to pair with clips.
- Boxing Gym Software Checklist: Memberships, Class Caps and AI Help — Memberships and class caps for a boxing gym.
- How to Write Instagram Captions With AI That Sound Like You — Captions that sound like you for the posts.
- ChatGPT Apps for Fitness Studios: Seven Jobs and Their Limits — Seven fitness-studio jobs for ChatGPT's connected apps, the permission to set for each, and the steps that stay with you.
- Yoga Class Plan Template: A Term of Themed Classes With AI — A one-page weekly template, a ten-week term arc and the review checks that catch what AI-drafted sequences get wrong, for studio owners and teachers.
- Can AI Write CrossFit Programming? Scaling Options and Coach Notes — A coach's guide to AI-drafted scaling ladders and whiteboard notes: the prompt, a worked draft with its errors marked, and seven checks before anything goes up.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: HeyGen, Synthesia, CapCut and Canva pricing and feature pages (checked October 2026); each tool's own licence terms for music and commercial use, which you should read yourself.