Can You Make a High-Quality AI Product Video With Zero Budget? We Tried.

-


A friend needed a product video for their custom steel and timber furniture business — and we wanted to see if we could pull it off without spending a single cent. Here's exactly how it went.

The Brief

A friend of mine runs a small business making custom steel-framed furniture with dark-stained timber panels. Beautiful stuff — raw, industrial, built by hand. They needed a product video. Nothing crazy, just something cinematic enough to use on social media and their website.

The problem: no budget. No camera crew. No studio.

So instead of doing it the normal way, I decided to test something. Could we use free AI tools — with clever prompting and some patience — to generate a full cinematic video sequence from scratch?

The answer, as it turns out, is yes. And here's exactly how we did it.

The Stack (All Free)

  • Claude (claude.ai) — creative direction, shot list, and every single prompt
  • Google AI Studio — text-to-image generation (Imagen / Nano Banana)
  • Google Flow — image-to-video animation using Veo 3.1
  • CapCut — final edit, assembly, audio mix, and cinematic letterbox bars
  • YouTube Audio Library — royalty-free music track

Total cost: R0. Zero. Nothing. Nada.

Step 1 — Building the Shot List with Claude

Before generating a single image, I used Claude to plan the entire creative direction. I described the project — a welder building a steel-framed cabinet with dark timber doors and drawers — and asked Claude to build a cinematic shot list that told the story visually.

The brief I gave Claude:

"I want a short series of images starting with a welder standing in his workshop, then measuring and cutting steel, then showing the material on a fittings table, then the welding process begins, the grinding and cleanup process, then the woodwork, then the final product — a beautiful industrial frame with dark stained wooden cupboards and drawers. I want this series to be cinematic in nature with close-up, medium, and wide angle shots."

Claude came back with a full 11-shot breakdown, each with a camera angle, subject, lighting description, and mood note. Here's an example of the level of detail:

Shot 1 — WIDE — The Arrival

Wide angle establishing shot. A male welder in dark work clothes and a welding helmet pushed up on his head stands alone in a dark industrial workshop. He faces away from camera toward a steel workbench. Tools hang on a pegboard behind him. A glowing amber compressor light sits to the right. Overhead fluorescent tube casts a single beam of light. Dark, cinematic, moody. Contemplative. No sparks. Still.

Claude also defined a master style reference to apply to every shot for consistency:

Cinematic documentary photography. Moody industrial workshop setting. Dark corrugated steel walls, single fluorescent tube overhead light, warm amber practical lights. Color grade: desaturated shadows, cool blue ambient, warm spark and tungsten highlights. Shallow depth of field. Shot on 35mm film.

This was the foundation. Everything that followed was built on this creative brief.

Step 2 — Generating the Storyboard in Google AI Studio

With the shot list locked, I took each prompt into Google AI Studio using the Imagen model (accessed through the Nano Banana interface). I fed the master style reference as a system-level context and ran each shot individually.

Here are the 10 storyboard images that came out:

Shot 1 — The Arrival (WIDE)



Shot 2 — The Measurement (MEDIUM)



Shot 3 — The Cut (CLOSE UP)



Shot 4 — The Fitup (MEDIUM)



Shot 5 — Welding Begins (WIDE)



Shot 6 — The Arc (MEDIUM)



Shot 7 — The Grind (CLOSE UP)



Shot 8 — The Wood (CLOSE UP)



Shot 9 — Assembly (MEDIUM)



Shot 10 — The Reveal (WIDE)



The consistency across these images is genuinely impressive for a first run. The lighting, colour grade, and mood stayed remarkably coherent — largely because the master style reference was baked into every single prompt.

Step 3 — Animating in Google Flow

This is where it got interesting. Google Flow is Google Labs' AI video tool, powered by Veo 3.1. You get a free credit allocation when you sign up — I had 150 credits to work with.

Each video generation costs roughly 20–40 credits, depending on settings. So I had to be strategic about which shots to animate and which ones to leave as stills.

My settings:

  • Model: Veo 3.1 Fast
  • Aspect ratio: 16:9 (YouTube native)
  • Count: x1 (single generation per shot to stretch credits)
  • Duration: 5–8 seconds per clip

The 7 shots I animated (priority order):

Shot Why
Shot 1 — The Arrival     Hero opener, slow push in
Shot 5 — Welding Wide     Most cinematic, sparks in motion
Shot 10 — The Reveal     Money shot, final frame
Shot 7 — The Grind     Sparks and motion, great animated
Shot 8 — The Wood     Subtle hand movement, tactile
Shot 3 — The Cut     Cutting action
Shot 9 — Assembly     Satisfying fit-up motion

The motion prompts I used were kept deliberately short and specific — Veo responds well to clear, concise direction rather than long descriptive paragraphs. Example:

"Slow cinematic dolly push toward the figure standing at the workbench. Figure completely still. Dark industrial workshop. Fluorescent tube overhead. Moody and atmospheric. No sparks."

Step 4 — Editing in CapCut

With 7 animated clips exported from Flow, I pulled everything into CapCut for the final assembly.

The edit order:

  1. Shot 1 — The Arrival (video, 5–6s)
  2. Shot 2 — Measurement (still)
  3. Shot 3 — The Cut (video, 4–5s)
  4. Shot 4 — The Fitup (still)
  5. Shot 5 — The Drama (video, 5–6s)
  6. Shot 6 — The Arc (still)
  7. Shot 7 — The Grind (video, 4s)
  8. Shot 8 — The Wood (video, 4–5s)
  9. Shot 9 — Assembly (video, 4–5s)
  10. Shot 10 — The Reveal (video, 6–8s, fade to black)

The still images had slow zoom Ken Burns effects applied initially, but they felt out of place against the animated clips, so I cut them from the final edit — a reminder that not every idea survives contact with the timeline.

The cinematic letterbox bars were added as overlays — black bars top and bottom, approximating a 2.39:1 ratio. This does two things: gives the whole sequence that proper film look, and covers the Veo and Nano Banana watermarks in the corners (haha!).

Music was pulled from the YouTube Audio Library — a dark cinematic instrumental that builds toward the reveal. No copyright issues, no cost.

The Final Result

Here it is. A complete cinematic product video. Zero budget. Zero camera. Just AI tools working together.



What Worked

The prompt quality made everything. The time invested upfront with Claude in building a coherent shot list and master style reference paid dividends throughout the entire process. Every downstream decision was easier because the creative direction was locked early.

Veo 3.1 handles industrial/spark content really well. The welding arc and grinder sparks animated convincingly — this is not an obvious use case, but it delivered.

16:9 was the right call. Gives you YouTube native format, and the cinematic bars bring it to true widescreen without any cropping awkwardness.

What Didn't Work

Ken Burns on stills broke the flow. The slow zoom effects on still images felt artificial next to genuine video motion. Cutting them was the right call, even though it shortened the final piece.

Watermarks are a real limitation on free tiers. The cinematic bars fix covered them creatively, but it's worth being aware of this going in.

Credit budgeting requires discipline. At 40 credits per generation on default settings, 150 credits disappear fast. Switching to x1 output and being selective about which shots actually need animation is essential.

Want the Full How-To?

This post covers the overview. If you want a dedicated step-by-step video walking through every tool, every setting, and every prompt — drop a comment on the YouTube video or leave one below.

If there's enough interest, I'll put together a proper tutorial covering the full workflow from brief to final export.

The Stack (Quick Reference)

Tool Purpose Cost
Claude     Shot list, creative direction, all prompts     Free
Google AI Studio     Text-to-image generation     Free (with limits)
Google Flow     Image-to-video animation (Veo 3.1)     Free (150 credits)
CapCut     Edit, assembly, audio, effects     Free
YouTube Audio Library     Royalty-free music     Free

Total: R0

heavymetalbren is a welding, making, and building channel based in Cape Town, South Africa. Follow along on YouTube and join the community over at The Forge.

Location: Cape Town, South Africa

Comments

Post a Comment

Popular posts from this blog

I Built a Fully Offline AI Butler. His Name is Alfred.

-
Image
A few weeks ago, I didn't have a single local AI model running on my machine. Today I can say "Hey Alfred" out loud and a British butler answers me — summarising news, opening apps, telling me the time, and doing it all completely offline, on my own hardware, with zero subscriptions and zero data leaving my house. Here's exactly how I got there, and why you should probably do the same. Why run a local LLM at all? The honest answer: privacy and resilience. Every time you type something into ChatGPT or Claude, that conversation goes to a server somewhere. For most things, that's fine. But there's something fundamentally different about an AI that runs entirely on your own machine — one that works when your internet goes down, one that doesn't bill you per query, one that has no idea what you asked it at 2am. I've also been retrenched before because of AI automation. That experience made me want to understand this technology, not just consume it. R...
Read more

I Built a Free Workshop Calculator App — Here's How It Actually Happened

-
Image
I didn't set out to build a web app. I was trying to figure out how to space flat bars evenly across a welding table frame. Simple enough problem — take your total frame width, subtract the bar widths, divide the remaining space by the number of gaps. Except every time I tried to do it in my head or on a calculator, I'd lose track of what I was doing, second-guess myself, and end up measuring twice anyway. I Googled it. Nothing useful came up. There were forum posts with someone's handwritten formula in the comments, a few Excel spreadsheets from 2009, and a lot of people saying "just measure it out on the floor." Not exactly what I was looking for. So I built the thing myself. What GapCalc Actually Is GapCalc is a free, browser-based workshop calculator built for welders, fabricators, and makers. No login. No app to download. No paywalls. You open it on your phone in the workshop, and it works. It started as one calculator — the Gap Calculator — which does exactl...
Read more