Product video, built without a camera.
Explainers, feature launches and customer stories for teams whose product cannot be filmed yet. We animate the interface in code, and your script ships word for word, down to the line your legal team approved.
First draft in 24 hours. Flat quote before we start · revision rounds agreed up front, in writing.
Built for SaaS and fintech teams
Most of our work now goes to software and finance companies, so the three videos those teams keep asking for are the ones we build first. None of it gets filmed, which is usually the point: the demo account is empty, the real data is confidential, and the feature ships next month.
Product and feature explainers
The interface is animated in code rather than recorded, so the walkthrough can show a flow your product has not shipped yet, in the order you want it seen, with every number on screen one you approved.
Numbers, charts and quarterly updates
Figures are read from a data file and your brand lives in a template. When the quarter changes, we change the file and render again instead of rebuilding the video from scratch.
Customer stories and case studies
The problem, what changed, and the number that proves it. Cut long for a landing page and short for LinkedIn out of the same build, so sales and social get one story instead of two.
Video for regulated niches
Healthcare and finance, script to delivery, with the compliance pass built into the process: fictional names where they belong, disclaimers where a viewer sees them and copy checked line by line.
Marketplace listing video
Video for a product page that reads with the sound off and carries nothing the review list rejects: no price, no logo, no link, no contact line.
Training & SOP modules
One task per module, under a minute, cut to the recorded voice so the steps land on the word. A second language runs through the same pipeline, so a bilingual set is not double the price.
Ad teardown and a test plan
Before anyone writes a script, someone has to read the account. I go through the live ads in Meta’s Ad Library, name the angle the brand owns but never says out loud, and come back with three concepts. Each one carries a single claim, so a losing test tells you which claim died. Production is a separate step, priced apart.
Localization pack
One master into four markets, script timed to the shot and not word for word, subtitles in every language. Lip matched and voice over priced apart.
Recent work
Everything below is 100% generated: the presenter, the voice, the delivery. If you can tell, that's a revision we'd owe you.
When the words are regulated
Some categories do not let you improvise. Pharmacy, finance and insurance all have copy that somebody signs off on, and the video has to say exactly that, at a pace a viewer can follow. This spot is a satire of pharma advertising, which makes the constraint sharper rather than softer: the joke only lands if the format is exact. Supers, disclaimer pacing and voiceover all behave like the real thing while saying something else.
Case study video
The format B2B teams keep asking for: a client success story that doesn't read like a slide deck. Problem, what changed, and the number that proves it, built as a long version and a short cut from the same material.
Motion graphics, data and process animation
Some animation carries numbers and some carries a process: a timeline with two tracks running side by side, a path that stops while another keeps going, a diagram that has to be read in one particular order. Both are built the same way, frame by frame in code. Nothing here is a slide deck handed to a generator, and that is usually where this work goes wrong: a generator puts the turn in roughly the right place, and roughly is visible. Every line, label and figure is positioned and timed by hand, so a chart is exactly your chart, the brand stays identical across every version, and a revision is a value change rather than a re-shoot.
Product explainers
Most explainers need the product on screen, and most products are not ready to be filmed: the demo account is empty, the real data is confidential, the feature ships next month. So we build the interface itself and animate it in code. Nothing is recorded, nothing is a static screenshot, and every number on screen is one you approved.
One template, a series of videos
The first video in a series is the expensive one, because that is where the layout, the brand and the motion get decided. After that, a video is a file of copy and numbers: new figures, new lines, new screens, same build. A monthly update or a twelve part onboarding course stops being twelve projects.
Product shots without a product shoot
A listing needs the same object on white and the same object in a room, and usually the product is not in your hands yet: the first run is at the factory, the sample is with a reviewer, the colourway ships next month. We build the shots from your own product images, or from a spec when there is nothing to photograph yet. Same device, same finish, same screen, in a studio set and in the places a buyer would actually meet it.




Every frame here is generated, not photographed, and the device is a demonstration piece rather than a real vendor's product. For your listing we start from your own images so the object in frame is yours.
The same object, now moving
A still shows the product. A turn shows what it is: the ports, the card slot, the back nobody photographs. Below are two pieces of hardware that were never built and never shipped — a counter-top terminal and a floor-standing kiosk — each put on a turntable that does not exist, in a studio that does not exist either. The terminal in motion is the same device that was standing still one section ago; the kiosk goes further and changes its finish without changing its shape.
One long piece, three vertical cuts
A long recording holds one idea per minute and a feed holds one per swipe. These three cuts come from a single 52 second piece: the moment the number changes, the moment two people see different truths, and the result. Each one is rebuilt for vertical rather than cropped to it, so the frame sits where a thumb is not, and the line that carries the cut is typeset, not squeezed into a corner.
The same take, a different room
The presenter on the left was filmed in an office. The one on the right is the same take, moved to a boardroom forty floors up. Not a re-shoot and not a regenerated clip: the pixels of the person are untouched, so the delivery, the timing and every lip movement are the ones that were recorded. Only the room behind them changed.
One shoot, a second market
Same presenter, same take, now speaking Spanish — and the mouth agrees with the words. A subtitle asks a viewer to read while your product is on screen; a dub lets them just watch. The English original is two sections up, so you can compare the delivery rather than take our word for it.
Ten openers, one product
A media buyer does not need one good ad. They need the first two seconds in ten versions, because the opener is what the test is actually testing. Same device, same finish, ten different rooms it might be met in: a cafe counter, a bakery, a boutique before opening, a queue at a small shop, a food truck at dusk, a barber station, a market stall, a hotel desk, the contactless symbol itself, and the table once the payment is done. Swap the opener, keep the rest of the cut, and you learn which room your buyer lives in.
The same four, shot vertical
- 8 seconds each, silent by design, widescreen and vertical
- One product across all ten, not ten lookalikes
- Generated from a written brief, no shoot and no sample
- Delivered as separate files, ready to load as ad variants
- Vertical is shot vertical, not cropped from the widescreen cut
Every clip here is generated, and the terminal is a demonstration device rather than a real vendor's product. For your test we build the openers around your product and your offer.
Video for a regulated room
Healthcare and finance don’t buy a video. They buy a video that survives review. This sample is built the way that work is delivered. The practice is named as fictional inside the frame, the disclaimer sits where a viewer sees it, and no line of the script makes a claim about care.
Northbay Family Health is invented for this sample. On your material we start from your approved copy and your own legal line.
A listing video that passes review
Marketplace video has a short list of rules, and most rejections come off that list, not off taste. This one carries no price, no marketplace logo, no link, no contact line and no black bars. The brand is invented and says so on the end card.
Verda Pay is an invented brand and the terminal is a demonstration device. On your listing we build the same cut around your product.
A training module your staff will finish
One task, one module, under a minute. The steps change on the word, because the cut is timed to the recorded voice track and not to a stopwatch. The rule for the awkward case gets its own card, since that is the part people forget first.
The terminal and the brand in this module are invented. On your material we film nothing and rebuild the steps from your own procedure.
One master, several markets
Two tiers in one sample. English and Spanish run on the same take with the lips matched to the new line. German runs as a voice over on b roll, and the end card says so, because a voice that doesn’t match the mouth is worse than an honest label.
The presenter is ours and the brand is invented. On your master we start from your approved script in the first language.
A read of the account before anyone hands over the numbers
Most creative briefs go wrong before a render is involved, because nobody read the ad account first. This is a teardown of a live US sleep tech brand, built from public sources in one sitting. Fourteen active ads off the Ad Library, the brand’s own homepage, nothing else. No relationship with the brand, no access to their data.
The brand isn’t named here. They’re not a client and didn’t ask to be read in public. On your account the same teardown comes back with your name on it, and the parts I can’t see from outside are marked unknown rather than guessed. Read the full teardown →
A tour of a building that is not built yet
Renders are the only material a project has before it opens, and a tour built from them comes apart in the same place every time. Straight mullions bend, rooflines melt, and the camera wanders into rooms nobody drew. This one holds.
The building here is invented and the renders behind it are ours. On your project we start from the renders you already have.
What we make
Flat project pricing, first draft in 24 hours. You approve a sample before we polish anything, and the number of revision rounds is agreed up front rather than argued about later.
Product & SaaS demos
60 to 90 second walkthroughs with an AI presenter over your product: screens, features, offer. Built to sit on a landing page and convert.
Motion graphics & animated data
Explainers, kinetic typography and charts that move: your metrics, your product UI, your brand colours. Rendered from code, so updating a number is a one line change and not a reshoot.
Long video → viral Shorts
Podcasts, webinars and interviews cut into vertical clips that hold attention: hooks, word-level captions, pacing built for retention.
3D product animation
Photoreal CGI of your product in motion: rotations, exploded views, hero shots. No studio and no physical sample required.
Hook packs for Meta
5 to 20 hook variations cut from one master ad, so your media buyer tests openers instead of guessing them.
Spokesperson & UGC ads
A consistent AI presenter delivering your script to camera, for ads, promos and social. Vertical or widescreen, captions burned in.
Footage editing & background swap
You already shot it. We replace the background, move your presenter to a new location or clean up the frame, and the original take, timing and lip movement stay exactly as filmed.
AI product photos
Listing and lifestyle shots built from your own product images. Clean white-background sets for Amazon, styled scenes for social.
Architectural tours from renders
A walkthrough of a building that is not built yet, made from the renders you already have. Glass, rooflines and window frames stay straight while the camera moves, which is where most AI tours come apart. The camera stays inside what your renders actually cover, so nothing gets invented along the way.
Dubbing & localization
Your existing video speaking another language with the lips matched to it. Same presenter, same delivery, a new market. Priced per language, so one master becomes four versions without a second shoot.
Why it doesn't look AI
Because we don't hit "generate" and hope. Every video goes through the same four gates.
Script & compliance pass
We edit for spoken pacing and, for regulated niches like finance, run a banned-word audit before anything is generated.
Character lock
Reference frames pin the presenter's face, wardrobe and voice, so video #30 matches video #1.
Generation
HeyGen, Google Flow (Veo 3), Higgsfield, ElevenLabs. The stack is picked per job, not per habit.
Word-for-word QC
Every line is machine-transcribed and checked against the script, and audio is mastered to broadcast loudness. No mumbled words reach you.
What clients wrote
Both reviews are public on our Upwork profile, quoted here word for word. Click through and read them at the source.
“We needed someone who could come in and own the project while providing thoughtful feedback and direction. Alexey wasted no time and always came back with proposal, recommendation, and decision points for us. We completed on budget and on time. He let us know what fell in and out of scope. I felt like we had clarity the entire time on where we were in the project and we are very happy with the results. We would work with him again.”
“Alexey delivered a high-quality AI spokesperson video ahead of the deadline. The brief was detailed and he followed it precisely — realistic AI presenter, burned-in captions, clean 9:16 vertical format. He's local in Brooklyn which was a plus for us. Fast communication, zero revisions needed. Will work with again.”
Before you write to us
The questions clients actually ask, answered before you have to send the email.
Who owns the finished video?
You do. You get the final files and the right to use them commercially: ads, landing pages, social, paid media, wherever you need them. The synthetic voice and any music we use come with commercial licences, so nothing in the delivery is borrowed.
One honest caveat, because it costs you nothing to know it: in the US, purely AI-generated frames generally aren't protected by copyright, so nobody (us, you or anyone else) gets a monopoly on raw generated footage. What is defensible is the script, the edit and your brand, which is exactly where the work goes. This is how we operate, not legal advice.
Do I need to ship you my product or appear on camera myself?
For spokesperson, UGC and demo video, no. The presenter is a generated character we lock to your brand, so there's no shoot, no talent booking and nothing to send us.
For product photos and 3D animation we do need your product's own images or reference shots, because those have to be your product and not a lookalike. If you want your own face or a colleague's on screen, that's possible, but only with footage and sign-off from the person appearing. We don't build a presenter out of someone else's likeness.
I don't have a script. Is that a problem?
No. Tell us what you're selling, who it's for and where the video runs, and we write it. You approve the script before anything is generated, which is the cheapest point to change your mind. After generation, changes cost time.
If you do have a script, we still edit it for spoken pacing, and for regulated niches like finance we run a banned-word pass before production starts.
Can I run this as a paid ad?
Yes, that's what most of it is built for. Files arrive export ready with captions burned in, and you can have 9:16, 16:9 and 1:1 from the same build rather than cropped afterwards.
On AI disclosure: platforms set their own rules and change them often, so we don't quote them at you. We can add a visible AI-generated label to any cut on request. Tell us if your media buyer wants one.
Will the presenter look the same across a whole series?
Yes. That's the part we've actually solved and the reason clients stay: reference frames pin the face, wardrobe and voice, so video 30 matches video 1. We're mid-series on a 30+ video run for a US financial-education brand right now, same presenter throughout.
Can you localise a video we already have?
Yes. 18 languages, with the lips matched to the new track and a voice that stays recognisably the same speaker. You send the footage you own, we hand back the localised version. No re-shoot and no re-generation of the original.
How fast, and how many rounds of changes?
First draft in 24 hours. Final delivery is normally 48 to 72 hours, longer only for multi-video series, and we say so before you pay rather than after.
Revision rounds are agreed up front and written into the quote, so nobody discovers the limit halfway through. Every line is machine-transcribed and checked against the script before you see it, which is why most projects need no revisions at all.
A studio, not a marketplace gig
True Take Studio runs out of New York. We work US business hours, reply in plain English, and treat your brand's consistency as the actual job, not an extra.
The same pipeline we sell has delivered 300+ videos and carries a 30+ video series for a financial-education client today. Three Upwork contracts and $3K+ billed so far. We don't show client names here; message us and we'll send work relevant to your niche.
Send the script. Or just the idea.
Tell us what you're selling and where the video will run, and we'll come back with a plan and a flat quote.