The cheapest way to make a SaaS promo video (without it looking cheap)

Most cheap SaaS promo videos don't look cheap because of the budget. They look cheap because of a handful of decisions that cost nothing to get right and nothing to get wrong.

I've spent fifteen years making promo and product videos, including for teams at Microsoft, xAI and Cloudflare, plus a long tail of startups who were counting every dollar. The startups taught me more. When there's no money for a shoot or a voice actor, you find out which parts of a video were actually doing the work. It's fewer parts than the expensive process implies.

So here's the ladder, bottom to top. Free first, because that's where most of the quality lives.

The zero-budget version, done properly

You need four things: a script, a screen recording, music, and a voice. All four can cost nothing.

Start with the script, and make it 30 seconds. Not 90. Thirty seconds forces you to cut everything that isn't the point, and what's left is usually the video. Write it as spoken sentences, then read it out loud with a timer. I aim for about eighty words and let the timer settle the argument. If you can't say what your product does in that space, the video isn't your problem yet.

A structure that survives contact with reality: one line naming the specific pain, not "workflows are broken" but the actual annoying thing your user does on Tuesdays. One line on what your product does instead. Then twenty seconds of showing it happening, then what to do next. No mission statement. No "founded in 2023 by two engineers who believed."

Then the screen recording. This is where free videos are won or lost, so I'm going to be tediously specific.

Record a window sized to the shape you're delivering, not your whole screen. Downscaled recordings usually leave small UI text looking soft, so check the export on your phone before you commit to anything. Hide your bookmarks bar. Hide your other tabs, so there's no "gmail (47)" sitting there. Turn on do-not-disturb properly, not hopefully. Use a clean browser profile with no extension icons crowding the toolbar.

Fill your app with real-looking data before you hit record. Not "Test Project 1" or "asdf". Real names, plausible numbers, dates that land on different days. Placeholder data is the first thing I notice in someone's demo, and I don't think I'm unusual.

Cursor discipline: move it deliberately, in straight-ish lines, and let it rest when you're not clicking. The thing that reads as amateur is the cursor drifting in little circles while the voiceover talks. Move, stop, click, stop. If your tool can smooth cursor motion, use it lightly.

And record one flow, not a tour. Pick the single sequence that makes someone go "oh, that's it?" — the thing their current tool makes them do the long way round, and yours doesn't. I've never been convinced by a menu walkthrough, and I've sat through a lot of them.

Music. There's genuinely good free and cheap music now, and it's the difference between watching a demo and watching a video. One track, no cuts, start it before the first frame of speech and let it run out at the end. Pull the volume way lower than feels right. If you notice the music, it's too loud.

Voice. Your phone's voice memo app is fine. What matters is the room. Record somewhere small with soft stuff in it: clothes, a bed, curtains. The classic move is real. Sit on the floor with a duvet over your head, phone about a hand's width from your mouth, slightly off to the side so your plosives don't thump. Say each line three times and keep the third. Take three is usually the least stiff. If it isn't, you're overthinking it, so walk away for ten minutes.

And don't read the line. Read it, look up, then say it like you're telling someone.

What actually costs money, and why

The free version above has real limits, and I'd rather name them than pretend.

You can't easily get motion design for free. Not real motion design: text that moves with weight, UI that assembles in a way that directs the eye, transitions that feel authored rather than applied. That's a skill. The tools that fake it have a house style, and once you've seen it you can't unsee it.

You can't get a great voice for free unless you happen to have one. Some founders sound wonderful on a phone recording. Some sound like a hostage video. If you're the second kind, that's worth paying for.

And you can't get taste applied to your specific product for free. Which screen to hold on, which to skip, when to cut on the beat, whether the logo belongs at the front or the end. That judgment is most of what you're buying when you pay someone.

Market rates, hedged because I haven't surveyed anything: a competent freelancer for a short product video usually lands somewhere in the high hundreds to low thousands, and an agency is generally several thousand and up, with a timeline measured in weeks. What surprises people is the calendar more than the price. On my own projects, the waiting between revision rounds is reliably longer than the work inside them. You send notes on a Tuesday and the new cut lands the following week.

Here's the part nobody warns you about. In my experience the third round is nearly always a script problem wearing an animation costume. You approved the script fast because it was just words, and now you're paying someone to re-animate thirty seconds because the second sentence made a claim you don't actually want to make. Spend your attention on the script, while changing it is still free.

I've been on the wrong side of that one. A founder pushed back on a line in a script I'd written for him and I talked him out of it, because I liked the line and it timed well against the music. He was right. The line promised something the product did on a good day, and we rebuilt half the animation around a replacement three weeks later. My taste, his product. He knew which one mattered and I didn't listen.

I've compared agency vs freelancer vs AI in detail — the short version is that you're paying for coordination, not craft.

Where cheap goes wrong

Cheap isn't the problem. Careless is. Some reliable tells:

Stock footage of people who don't use your product. A woman laughing at a laptop in an airy office has nothing to do with your API monitoring tool, and every viewer clocks it. If a shot could appear in a video for a completely different company, cut it.

Template intros. The three-second logo sting with the swoosh. It doesn't make you look established, it makes you look like you bought a template. Your logo can just appear. Or appear at the end.

Robotic pacing. I think flat pacing sinks more videos than ugly visuals do, and it's the harder one to spot in your own work. A voice that never pauses and hits every sentence with identical rhythm. Real speech has gaps. When you edit your voiceover, resist tightening every silence, and leave the breath before the important sentence.

Tiny UI text. Text that's perfectly comfortable on your monitor is often unreadable on a phone. Open the export on your phone and try to read it at arm's length. If you squint, zoom in, or crop into the region that matters rather than showing the whole screen.

Too many things. Six features in thirty seconds is zero features in thirty seconds.

Where AI sits on this ladder

AI video sits in a gap I used to have no answer for: above careful DIY, below hiring a person. It buys you the motion design, the voice, and a version of the pacing decisions, without the calendar.

I built PromoHyper, an AI promo video maker, to occupy exactly that slot. You paste your product URL, it reads your brand, writes and voices a script, scores it, and designs every frame rather than dropping your logo into a template. From $29, usually ready in minutes, in 16:9, 9:16 and 1:1. Multi-video packs bring the per-video price down: $147 for three, $290 for ten.

I'm not going to tell you it replaces a good motion designer who has spent a week understanding your product. It doesn't. What it replaces is the video you were going to make in three weeks, or the one you weren't going to make at all.

The order I'd actually do it in

Write the 30-second script today. Read it aloud. Cut it to about eighty words. Record your one flow properly, with a clean window and real data, four or five times until the cursor stops wandering. Add a quiet music bed. Record the voice under a duvet.

Then look at what you have. It might be good enough, and if it is, ship it and spend the money elsewhere. If the voice is the weak link, or the whole thing feels flat where a bit of motion would carry it, that's the moment to pay for something. And now you know what you're paying for instead of buying a video-shaped object and hoping.

Care beats budget. I've never once seen it go the other way, and I've been looking.