AI video for Instagram Reels: I made a 6-second product ad in 10 minutes (step by step)

Text-to-video is finally useful for small brands. The shot-not-story prompt formula, Nova Reel vs Luma, adding text, logo and music in the editor, MP4 export — and the mistakes that waste credits.

KKryotta TeamProduct & research · · 3 min read
A steaming filter-coffee mug being filmed for a short product video
A steaming filter-coffee mug being filmed for a short product video

I'll be honest — I avoided making Reels for my side project for months because "video" sounded like editing software, music licensing and a weekend gone. Then I tried a text-to-video model on a Tuesday afternoon and had a usable 6-second product ad in ten minutes. Not a masterpiece. Usable. Posted it. It got more reach than my last five photo posts combined.

Here's the exact process, including the bits that didn't work.

Step 1: write the shot, not the story

Video models are great at one shot with motion, bad at plots. Don't ask for "an ad about our coffee". Ask for a shot:

Slow push-in on a ceramic mug of filter coffee on a wooden table, steam rising, warm morning light through a window, shallow depth of field, cinematic.

Concrete nouns, one camera move, lighting, mood. That's the formula.

Step 2: pick the model for the look

  • Amazon Nova Reel — fast, affordable, clean product shots. My default for "just make it".
  • Luma Ray 2 — more cinematic motion and realism; a bit slower and pricier. For the hero clip.

In Kryotta the video studio is conversational, so refining is a chat: "make it evening, lamp light", "slower push-in", "add rain on the window". Each turn re-renders the shot without you rewriting the whole prompt.

Step 3: the 6-second rule

Reels reward the first two seconds. A 6-second clip is enough for: motion (0–2s), product clarity (2–4s), text overlay CTA (4–6s). If you need longer, stitch two shots — but start with one great one.

Step 4: add text, logo and music in the editor

The generated clip opens in the built-in video editor (think a light Canva for video):

  • Text layer: "Fresh-ground. ₹349. Link in bio." Big, bold, high contrast, bottom third. Rotate it a few degrees if you want that hand-placed feel.
  • Logo: upload the PNG, corner, 60% opacity, on for the whole clip.
  • Music: upload a track you have the rights to (or a royalty-free one), trim to the exact 6 seconds you want with the waveform, set volume so it doesn't drown a voiceover.
  • Shape: a rounded rectangle behind the text makes it readable on any background.

Export as MP4 (H.264) — Instagram, WhatsApp and YouTube Shorts all accept it directly. WebM if you're posting on your own site.

What didn't work (so you don't waste credits)

  • Hands and text inside the video. Don't generate a hand holding the product or a label with readable text; add text as an overlay instead.
  • Multiple products in one shot. One hero object per clip.
  • Vague adjectives. "Beautiful, premium, amazing" do nothing. "Warm morning light through a window" does everything.

Cost, so you can plan

A Nova Reel clip costs about the price of a chai; Luma more like a filter coffee. Even ten iterations to get the perfect shot is cheaper than one hour of a videographer — and it's ready today.

A repeatable weekly Reel

  1. Monday: write three shot prompts for the week's product.
  2. Generate on Nova Reel; upgrade the best one to Luma if it's a hero post.
  3. Add text + logo + music in the editor; export MP4.
  4. Post with a Hinglish caption (yes, AI can write that too — see my earlier post).

Ten minutes a week. Try it once and you'll stop dreading "video".

K
Written by
Kryotta Team
Product & research

Kryotta is the multi-model AI workspace — every leading model, one login, one bill. Try it free →

Related reading