How to Create Beautiful Product Commercial Videos with AI — The Complete ImagineToVideo Tutorial
A traditional skincare commercial — the kind with soft window light, a model applying serum in slow motion, a hero shot of the bottle on a marble surface — costs an indie brand somewhere between $8,000 and $40,000 once you factor in the studio, model, gaffer, retoucher, and editor. It also takes a week. The same exact thing, generated in your browser, costs about a dollar and takes around five minutes.
That's not a hypothetical. It's the workflow we're going to walk through. By the end of this post, you'll have the complete two-model pipeline we use at ImagineToVideo for product commercials: ChatGPT (or GPT Image) to plan a 9-panel storyboard, then Seedance 2.0 on ImagineToVideo to render it as a finished video. Same logic as our K-pop dance tutorial — a static reference sheet drives a video model — but the storyboard format and prompt structure are tuned for product narrative instead of choreography.
Here's the storyboard ChatGPT generated for the running example, a clean skincare commercial:
And here's the final Seedance 2.0 render — the same woman, the same product, now actually moving through the commercial:
<video src="https://cdn.imaginetovideo.com/blog/content/zaraproduct.mp4" controls style="max-width: 480px; width: 100%; display: block; margin: 1.5rem auto;"></video>
About this series. This post is part of ImagineToVideo Video Generation Tutorials — a hands-on series where we walk through one specific creative use case per post, from prompt to final render, using the models built into ImagineToVideo. The series covers K-pop dance videos, product commercials, pet portrait animations, cinematic dialogue with native audio, and vertical narrative shorts — read them in any order. This episode is the one most ecommerce founders, DTC brand teams, and Shopify operators have been asking us for: how to generate a real, watchable product commercial without a film crew.
Why ChatGPT + Seedance Is the Right Stack for Product Commercials
A product commercial is not a dance routine. It's a story arc: a problem the customer feels, a discovery moment, a use moment, a transformation, and a hero shot. Nine frames is enough to land all of that. Sixteen — like we used for K-pop choreography — is overkill, and the extra panels actually hurt because they dilute the story beats.
ChatGPT and GPT Image are excellent at this kind of structured narrative storyboard. Give the model a structured prompt with a 9-panel layout and an explicit story flow, and it will render the same character — same makeup, same wardrobe, same lighting style — across all nine frames. No on-set continuity supervisor required.
What ChatGPT cannot do is animate. The output is a still grid.
That's where the second category of model comes in: reference-to-video (R2V). Feed it one or more reference images plus a motion prompt, and it generates a video that respects the references. Two are integrated into ImagineToVideo:
- Seedance 2.0 — supports up to 9 reference images, generates 5–15 seconds at 720p, and costs roughly $0.85 for a five-second clip. This is the one we recommend for product commercials. A hero shot is usually a single 5-second clip — Seedance is cheap enough that you can iterate on the lighting, the camera move, and the product placement without burning money.
- Kling V3 Omni — supports up to 7 references plus native audio and multi-shot sequencing, but costs about $13 for a five-second clip. The right tool when you want the entire 9-panel storyboard turned into a connected 30-second commercial with synced VO and music, not just a single hero shot.
The rest of this post is the Seedance path, with a Kling upgrade section near the end. If your final deliverable is broadcast-grade or destined for in-store retail screens at native 3840×2160, the Kling 4K Mode tutorial covers that path.
Step 1 — Plan Your Commercial in ChatGPT (the 9-Panel Storyboard)
Open ChatGPT (image generation enabled) or GPT Image directly. You don't need a base portrait this time the way you did for K-pop — for product commercials, ChatGPT is happy to design the model and the product styling from a single prompt. Paste in this template, edit the bracketed fields for your product, and run it:
9-panel skincare commercial storyboard, 16:9 layout, single image.
[CHARACTER] Young woman with soft natural makeup and styled hair, wearing
light elegant clothing. Same model, same outfit, same look across all 9
panels — do not redesign the character.
[SETTING] Modern sunlit interior, large windows, calm atmosphere, white and
pastel color palette.
[STORY FLOW — 9 panels in reading order]
1. Woman at her desk, looking slightly tired or concerned
2. Close-up of her face showing dry or dull skin
3. Discovering a clear skincare bottle, gently presenting it
4. Close-up: liquid pouring onto her hand
5. Applying the product to her face
6. After use — hydrated, glowing, smooth skin
7. Walking in a bright, clean space, refreshed and confident
8. Calm product beauty shot on a minimal surface
9. Final frame: soft smile at the camera, holding the product
[STYLE] Cinematic close-ups, soft focus, light reflections, subtle sparkles.
Soft natural lighting, minimal aesthetic, fresh airy mood. Elegant, visually
soothing.
[NEGATIVE] No text, no subtitles, no branding text, no logos, no extra
characters, no watermarks.
[OUTPUT] At least 1024×1024 final resolution. 2048×2048 if available.
Three things in that prompt are doing real work and shouldn't be removed:
The "same model, same outfit" line. This is the same trick we used for K-pop. Without it, ChatGPT will quietly change the wardrobe between panel 4 and panel 7 because the prompt mentions "applying" and then "walking" — it interprets the activity shift as a wardrobe shift. Pin the character explicitly.
The structured story flow. Listing all 9 frames in reading order — top-left to bottom-right, like a Western reading pattern — forces ChatGPT to compose the grid in narrative order. Without numbered panels, it will pick whichever frame "feels" like the hero shot and put it in the middle, which breaks the flow when you crop later.
The 1024+ output resolution. Same reason as our K-pop tutorial: ImagineToVideo's Seedance R2V endpoint rejects reference images smaller than 640×640 and returns a non-retryable error. ChatGPT's default output sometimes lands at 768×768. Demanding 1024+ in the prompt avoids the rejection downstream.
What you should now have is a single 16:9 grid with 9 panels showing the full commercial arc — exactly like the storyboard at the top of this post.
Adapting this template to other products. The same 9-panel scaffold works for almost any DTC category. For a fragrance: swap "dry skin" for "everyday outfit" in panel 1, "applying serum" for "spritzing perfume" in panel 5, and "glowing skin" for "confident night-out look" in panel 7. For a kitchen appliance: open on a chaotic countertop, midpoint on the product in use, close on a clean plated meal. The story arc — problem → discovery → use → transformation → hero — is the part that's universal.
Step 2 — Crop to Your Hero Frame (or Use the Whole Sheet)
You now have a single 9-panel image. There are two ways to feed it to Seedance, and the right choice depends on your deliverable:
Use the whole storyboard as a single reference. Best when you want a 5-second hero clip that captures the vibe of the commercial — the lighting, the model, the product styling — but not a specific narrative beat. Drop the full 9-panel image into the reference slot and let the model decide which moment to emphasize. In our testing, Seedance usually picks the application or the hero product shot.
Crop one specific panel. Best when you want a 5-second clip of one specific moment. For most product commercials, that's panel 8 — the calm beauty shot on the minimal surface — because that's the frame that goes on a product detail page or a paid-social ad. Crop it in Photopea, Figma, or Preview on macOS. The crop should be at least 640×640; re-export at 1024+ to be safe.
Crop two to four panels for a longer narrative. If you're building a 15-second TikTok ad or a 20-second YouTube pre-roll, crop the four story beats: panel 2 (problem), panel 5 (use), panel 7 (transformation), panel 8 (hero). Run Seedance four times — once per panel — and stitch the results in any free editor. Total cost: about $3.40 for the four clips.
For the example commercial in this post, we cropped panel 8 — the hero product shot — and ran a single 5-second Seedance clip. That's the video embedded above.
Step 3 — Bring the Product to Life on Seedance 2.0
Open the Seedance 2.0 generator (or the Seedance landing page and click Try Now).
Upload your cropped panel — or the full storyboard — to the reference slot. The UI will display the thumbnail. If it's rejected with a "pixel count too low" error, your reference is under 640×640. Re-export at higher resolution and re-upload.
In the prompt field — which has a 2000-character limit — paste in something structured. The pattern that works for product commercials:
[SUBJECT] A clear glass skincare bottle with minimal label, held by a young
woman with soft natural makeup. Her hand is steady, the product is centered.
[ACTION] She gently dispenses a few drops of clear serum onto her fingertips,
then smooths the liquid across her cheek in slow, deliberate motions. Her
expression softens from tired to refreshed. Light reflections move across the
glass bottle as her hand turns.
[CAMERA] Slow push-in close-up, shallow depth of field, soft focus background.
Subtle handheld micro-movement. Eye-level framing.
[LIGHTING] Soft natural light from a large window on the left. White and
pastel color palette. Light reflections on the glass, no harsh shadows.
[STYLE] Cinematic skincare commercial, minimal aesthetic, fresh airy mood.
Subtle sparkle on the liquid. Elegant, calm, visually soothing.
[DURATION] 5 seconds. 16:9 aspect ratio.
A few tuning notes from running this loop on dozens of product categories:
Resolution: stick with 720p for iteration. Seedance offers 480p and 720p. 480p is too soft for product detail — the label on the bottle blurs, the texture on the cream goes mushy. 720p is the sweet spot for Shopify product pages, Meta Ads, TikTok, and YouTube Shorts. If your final deliverable is a 4K hero on a brand site, nail the camera move and lighting at 720p first, then re-run the locked prompt in Kling V3 with the 4K toggle.
Duration: start at 5 seconds. Seedance supports 5, 8, 10, 12, and 15 seconds. For a hero product shot — a slow push-in or a calm rotation — 5 seconds is plenty. Longer durations give the model more room to drift, and product commercials are unforgiving about drift: the bottle shape changing mid-shot will tank the entire clip.
Camera: prefer slow over fast. "Slow push-in," "subtle rotation," "static with slight handheld" — these are the camera moves that read as premium. "Fast zoom," "whip pan," "rapid dolly" — these are the moves that read as cheap because they hide product detail behind motion blur. The whole point of a product commercial is that the viewer sees the product clearly. Move the camera slowly enough that they can.
Use Seedance 2.0 Fast for iteration. ImagineToVideo offers two Seedance tiers: quality and fast. Fast is 25–40% cheaper and noticeably quicker, with output that's nearly indistinguishable for short reference-driven clips. We do every iteration on Fast and only switch to quality for the locked final render that goes to the brand's product page or paid social.
Total cost for the example clip above: about $0.85. Seedance Fast would have brought it down to roughly $0.65. For comparison, a single 5-second product hero shot from a commercial production house starts at $3,000 and takes a week.
→ Generate your own product commercial on Seedance 2.0
Step 4 — Upgrade to Kling V3 Omni When You Need a Full Commercial
Seedance is the right answer for ~80% of product commercial work. The 20% case where Kling V3 Omni is the better tool:
- You need the entire 9-panel storyboard as one connected 30-second commercial. Seedance is single-shot only. Kling V3 Omni supports up to 6 connected camera shots inside a single generation, with consistent character and product identity across cuts. For a full TV-spot-style commercial — open on the problem, cut to discovery, cut to use, cut to hero — that's Kling.
- You need synced music and voiceover. Seedance generates muted video. Kling V3 Omni generates video and native audio in one pass. If you want a finished commercial with ambient room tone, product sound effects, and (with the right prompt) a cue to drop a music bed underneath, that's Kling.
- You need true 4K for retail screens or broadcast. Kling's pro mode generates native 3840×2160. Seedance caps at 720p. For a hero loop on a brand site above the fold, an in-store retail screen, or a TV spot — anywhere the viewer is closer than 6 feet to a 1080p+ display — the 4K upgrade is worth it.
Cost comparison for a five-second clip:
| Model | ~5s cost | Best for product commercials |
|---|---|---|
| Seedance 2.0 Fast ⭐ | ~$0.75 | Iteration, paid social, throwaway tests |
| Seedance 2.0 Quality | ~$0.85 | Locked final renders for product pages, ads |
| Kling V3 | ~$8.84 | Multi-shot commercials, 1080p delivery |
| Kling V3 Omni | ~$13.41 | Full 30-second connected commercial + native audio |
| Kling 4K Mode | ~$25–30 | Retail screens, TV spots, brand site hero loops |
The bolded rows are where this workflow lives by default. The Kling rows are where you go after the storyboard is locked at 720p.
→ Try Kling V3 image-to-video (entry point for V3 standard, V3 Omni, and the new 4K mode)
How the Three Models Compare on the Same Storyboard
We ran the same 9-panel skincare storyboard through all three relevant models on ImagineToVideo. The short version:
- Seedance 2.0 Quality held product identity rock-solid across the full five seconds. The bottle shape stayed the bottle shape, the label didn't morph, the liquid moved like liquid. This is the clip embedded above. For a single hero shot, this is the right tool.
- Kling V3 Omni produced a more cinematic image — better light reflections on the glass, more depth in the background — but cost 15× more for a clip that, side by side, isn't 15× better for a single hero shot. The advantage shows up only when you use Kling's specific capabilities: multi-shot sequencing, native audio, or 4K output.
- Kling V3 standard struggled because it doesn't accept reference images on the standard tier — only first/last-frame inputs. For a reference-driven product workflow, it's a non-starter. Use V3 Omni instead.
The takeaway: for the ChatGPT-storyboard-to-product-commercial workflow specifically, Seedance is not just the cheap option — it's the right option for a single hero shot. Kling V3 Omni is the tool you reach for when the deliverable is the full connected commercial, not just one frame from it.
Five Common Pitfalls (and How to Avoid Them)
After running this loop across skincare, fragrance, supplements, kitchenware, and consumer electronics, the failure modes that come up most often:
1. The product drifts mid-shot. This is the number-one product-commercial-specific failure. The bottle changes shape, the label warps, the cap disappears. Cause: the reference image is too small or the prompt doesn't pin product geometry. Fix: re-export the reference at 1024×1024 minimum, and add an explicit same bottle shape, same label, same proportions, same cap clause to the Seedance prompt.
2. Character drift between storyboard panels. ChatGPT will quietly redesign the model's hair or wardrobe around panel 5 if you don't pin it. Always include "same model, same outfit, same look across all 9 panels — do not redesign" in the storyboard prompt. If drift still happens, generate the model's portrait separately first, then feed that portrait into the storyboard prompt as a fixed character reference.
3. Lighting that reads as fluorescent. "Good lighting" in a prompt gets you generic flat lighting that looks like a stock photo. Product commercials need directional light: soft natural light from a large window on the left or studio key light at 45° with subtle rim light. Specify the direction. The single biggest jump in perceived quality on a product commercial comes from naming the light direction explicitly.
4. Camera moves too fast for the viewer to see the product. This is especially common on first-time generations because the prompt-writer over-corrects for "boring." A fast push-in feels exciting in your head; in the rendered clip, it blurs the product into a streak. Default to slow push-in, subtle rotation, or static with slight handheld. Speed reads as cheap on product. Slowness reads as premium.
5. Aspect ratio mismatch. A skincare brand site hero is 16:9 horizontal. A TikTok or Reels ad is 9:16 vertical. A Shopify product card is square 1:1. The model composes for the target aspect ratio; cropping a 16:9 generation into vertical at the end almost always cuts the bottle's cap or pump off-frame. Decide the deliverable format before you generate, and set Seedance's aspect ratio to match.
TL;DR — Your Product Commercial in Three Steps
- ChatGPT / GPT Image — generate a 9-panel commercial storyboard at 1024+ resolution, with explicit "same model, same outfit, same product" pinning. The 9 panels follow the arc: problem → discovery → use → transformation → hero.
- Crop — pick the hero panel (usually panel 8) for a single 5-second clip, or 4 panels (problem / use / transformation / hero) for a stitched 20-second narrative. Each crop ≥640×640.
- Seedance 2.0 — drop into /seedance, write a structured 100–200 word prompt with named light direction and slow camera move, render at 720p / 5s. Roughly $1 and three minutes per clip.
Upgrade to Kling V3 Omni when you need the full 9-panel storyboard rendered as a connected 30-second commercial with native audio. Otherwise, Seedance is where this product-commercial workflow lives.
→ Start your first product commercial on Seedance 2.0 → Or jump to Kling V3 image-to-video for full commercials + 4K
If your final deliverable is a brand site hero loop, an in-store retail screen, or a broadcast TV spot at native 3840×2160, the Kling 4K Mode guide walks through that path specifically — and our 4K AI video generator landing page is the fastest way to get there.
The Tutorial Series Roadmap
ImagineToVideo Video Generation Tutorials covers the creative use cases creators ask us about most. Read them in any order — each post is self-contained:
- K-pop dance videos — ChatGPT 16-panel choreography sheet → Seedance 2.0 dance video. (Published.)
- Product commercial videos — ChatGPT 9-panel storyboard → Seedance 2.0 hero shot. (You're reading it.)
- Pet portrait animation — turning your dog into a cinematic short with one photo and Seedance.
- Cinematic dialogue scenes with Kling V3 Omni native audio (lip sync + room tone).
- Vertical 9:16 narrative shorts — end-to-end TikTok-ready story clips from a single character reference.
Subscribe to the blog or check back as we publish more episodes. And if there's a specific product category you want us to walk through — fragrance, supplements, fashion, electronics — drop us a line.
Now go make your product move.
