How to Make a Shopify Product Video with AI

You can turn any existing Shopify product photo into three polished video variants in under 30 minutes without a camera, a crew, or a studio. The workflow is three steps: clean and relight the still with Nano Banana Pro and Seedream 5.0, add motion with Seedance 2.0 or Kling 3.0, then cut the variants in the 8frame Studio. The total cost runs $1 to $3 per tested variant, compared to $500 to $1,500 for a traditional product shoot. If your store is still running static PDPs, you're losing the conversion gap that mobile shoppers have already decided is the baseline.

TL;DR

• The 3-step workflow: still cleanup (Nano Banana Pro + Seedream 5.0), motion generation (Seedance 2.0 or Kling 3.0), cut and caption in 8frame Studio
• Cost per variant: $1 to $3 on 8frame versus $500 to $1,500 for a traditional shoot; most stores test 3 to 5 variants per product before picking a winner
• Aspect ratios that matter: 1:1 for PDP hero, 9:16 for Reels and TikTok, 16:9 for paid display ads

Why Shopify stores need product video now

Mobile shoppers convert at roughly 50% higher rates on PDPs that include a short video than on static image pages. That number has been climbing every year as TikTok and Reels train buyers to expect motion before they add to cart.

The bottleneck isn't intent. Most store owners know they need video. The bottleneck is production. A single product shoot with a photographer, lighting, and a light post-production pass runs $500 minimum, usually more like $800 to $1,500 for anything with proper lighting. If you have 40 SKUs and want even one variant per product, that's $30,000 to $60,000 in production before you've tested a single cut.

The AI workflow doesn't replace every shoot. Hero brand videos still benefit from the real thing. But for PDP loops, first organic Reels, and ad creative testing, the economics no longer make sense for the traditional route. The output quality from Seedance 2.0 and Kling 3.0 in June 2026 is good enough for client-facing work on most product categories.

The 3-step AI workflow for Shopify product video

Step 1: Clean and relight the still

Start with what you have. Any existing product photo works, even a phone shot on a kitchen table. Two models handle the prep:

Nano Banana Pro removes the background, separates the product from shadows, and sharpens edges that become visible when motion is added. Messy shadows on a static photo look subtle. On a 5-second video loop they're immediately obvious. The background removal in Nano Banana Pro uses object-aware masking, so for complex shapes like stacked jewelry or translucent bottles, it handles edge cases that standard tools miss. Processing time: 8 to 12 seconds per image.

Seedream 5.0 handles the relighting step. You feed it the cleaned product image and prompt it to place the product in the scene you want: soft studio light for a premium feel, bright lifestyle daylight for casual categories, dramatic side-light for food. Seedream 5.0 is an image model, not a video model, so this step is still a still. You're building the source frame that the video model will animate.

A prompt that works well for this stage, tested on a ceramic French press:

Product still of a matte black ceramic French press on a warm marble surface, soft window light from the left, shallow depth of field background, clean white negative space to the right, no text overlay

Generation time: about 15 seconds. If the first frame doesn't match what you want, the second or third will. You're spending $0.15 to $0.25 per Seedream image, so iterate freely.

Step 2: Generate motion

Once you have a polished still, you're sending it to a video model as a reference frame. Two models cover most Shopify product categories:

Seedance 2.0 is the pick for any product where consistency matters: the product has visible branding, a label, a distinct shape, or a specific color that can't shift between frames. Seedance 2.0's multi-reference conditioning locks the product appearance and builds motion around it. For the French press example, the prompt specified a slow pour motion with steam rising. Seedance held the product geometry and label text frame-to-frame better than any other model we've tested. Cost: $0.45 to $0.65 per 5-second clip. Generation time: roughly 110 to 130 seconds.

Kling 3.0 is the pick for lifestyle motion: products in a scene, ambient camera movement, environmental context. It's faster (about 60 seconds per clip) and cheaper ($0.28 to $0.40 per clip). For a product shown in a room, on a table at a cafe, or worn on a person in a natural setting, Kling's output looks more filmed and less generated. It's also the better choice when you're running variant tests and need volume quickly.

The rule is: if the product itself is moving or the viewer needs to see specific product details in motion, use Seedance. If the environment is moving and the product is mostly static, use Kling.

Step 3: Cuts and music in the 8frame Studio

You'll exit with two or three raw clips from step 2. In the 8frame Studio, you trim them to the right duration for each placement, add auto-captions if you're targeting social placements, and export at the correct aspect ratio.

For a standard product launch, you want three exports from the same session:

• 1:1 at 15 seconds for the Shopify PDP hero loop (no audio needed, auto-loops)
• 9:16 at 7 to 10 seconds for Reels and TikTok organic posts
• 16:9 at 15 to 30 seconds for Meta and Google paid display

The whole Studio session takes 10 to 15 minutes once you have the clips. Total time from raw product photo to three exportable variants: 25 to 35 minutes.

Aspect ratios and where they go

Getting the aspect ratio wrong costs you in two ways. Wrong-cropped video on a Shopify PDP shows letterboxing or black bars that signal "this wasn't made for here." Wrong-cropped video in an ad gets auto-cropped by the platform in ways that cut off the product.

Here's the map:

1:1 (square). Shopify PDP hero section, product grid thumbnails, Instagram feed posts. This is the safest format if you're only cutting one export. Every placement that prefers a different ratio will center-crop a square without killing the product visibility.

9:16 (vertical). Reels, TikTok, Pinterest Video, Snapchat. The formats with the highest organic reach right now. If your store has any social presence, this ratio deserves its own cut. Don't crop a 1:1 to 9:16 post-generation. Generate the video at 9:16 from the start, because the motion composition changes.

16:9 (landscape). Meta feed ads, Google Display, YouTube pre-roll, and Shopify's email campaign embeds. Still the standard for paid inventory even as social shifts vertical. Most performance marketing agencies will ask for this format alongside the vertical cut.

When you run Seedance 2.0 or Kling 3.0 in 8frame, you set the output ratio before generation. It's not a crop. The model composes the shot for the ratio you picked.

Picking models by product category

Not every product category behaves the same way in video generation. These picks reflect what we've tested:

Food and beverage. Seedance 2.0 for anything poured, stirred, or with steam. The physics on liquids and vapor are noticeably better than Kling. Kling for plated food or ambient cafe settings where you want warmth and depth.

Fashion and apparel. Kling 3.0 for lifestyle shots: a jacket on a person in a streetwear context, a dress in natural light. The environment motion is more convincing. For flat lays or product-only close-ups, Seedance gives you better fabric texture consistency.

Beauty and skincare. Seedance for close-up texture detail: a serum drop, cream spreading on skin, gloss on a lip. The macro fidelity is better. Kling for full-face lifestyle context.

Electronics and tech. Seedance. Electronics have precise form factors, exact label positions, and port placements that shift if the model isn't locked to the reference. Seedance's multi-reference conditioning handles this better than anything else we've tried.

Home goods and furniture. Kling 3.0. Room-in-use context, ambient daylight through a window, products sitting in a styled space. Kling's environmental motion reads as filmed rather than generated.

Full walkthrough: a ceramic French press at $89

Here's the complete job, from a single product photo to three exports ready for launch.

The product. A matte black ceramic French press, $89, sold on a Shopify store. The existing photo is a white-background studio shot taken against a sweep, with even diffused lighting. Technically clean but visually flat.

Step 1 output. Nano Banana Pro separated the French press from the background in 10 seconds. Seedream 5.0 placed it on a warm marble countertop with morning window light from the left and a soft bokeh background suggesting a kitchen interior. We ran three Seedream variations at $0.20 each before landing on the frame we wanted. Total step 1 cost: $0.80. Total time: 6 minutes.

Step 2 outputs. We generated three clips:

• Seedance 2.0 close-up pour. The French press plunger slowly depresses, steam rises from the spout, and a thin pour hits a waiting mug just out of sharp focus. 5 seconds, 1:1 ratio. Cost: $0.55. Time: 125 seconds.
• Kling 3.0 lifestyle pan. Slow camera drift across the counter, French press in frame, morning light shifting slightly as if clouds are passing. 7 seconds, 16:9 ratio. Cost: $0.35. Time: 58 seconds.
• Kling 3.0 vertical social. Same counter setup, faster camera push-in ending on the French press label, 6 seconds, 9:16 ratio. Cost: $0.32. Time: 55 seconds.

Step 3 output. Trimmed in the 8frame Studio, auto-captions added to the 9:16 cut ("No filter. No script. Just coffee."), exported at correct ratios. Studio time: 12 minutes.

Total cost: $2.02 in model credits. Total time: 29 minutes. Outputs: PDP loop, paid display variant, organic social variant, all from one source photo.

Unit economics for Shopify stores

The math changes depending on how many SKUs you're running and what your current production budget looks like.

Traditional shoot comparison. A photographer, half-day studio rental, and basic retouching for a set of product stills: $500 to $1,000. Add video to that package and you're at $900 to $1,500 before editing. For one product. If you have a catalogue of 30 products, the per-product cost drops somewhat with a production day, but you're still looking at $10,000 to $20,000 to get everything shot once.

AI workflow cost per product. $1 to $3 per variant in 8frame. If you test 3 variants per product to find the winner and stop testing the losers, you're spending $3 to $9 to find your best-performing video for a given SKU. If you test 5 variants before picking, you're at $5 to $15.

Testing logic. Most DTC teams we see run 3 video variants per product on paid channels before promoting one to the full budget. The ability to generate 3 variants for $3 to $9 instead of $500 per video changes the testing math completely. You can afford to test every hypothesis.

Time per variant. About 8 to 12 minutes of active work per variant once you're past the first still setup (which is reusable across all variants). The generation itself runs in the background.

Volume break-even. If you have more than 5 products and any ad spend at all, the AI workflow pays back on the first product.

Where it breaks

We'd rather tell you the failure modes than have you discover them mid-campaign.

Glossy surfaces. Reflective packaging (glass bottles, foil pouches, mirror-finish electronics) picks up artifacts from the generation process that aren't visible on matte surfaces. The reflections shift in ways that read as wrong to the eye even when nothing specific is obviously broken. Workaround: generate with soft, diffused lighting to minimize specular highlights in the source frame.

Label distortion. Any product with legible text on the label (a supplement bottle, a wine label, a skincare tube) can see letter-level distortion when the product moves. Seedance 2.0 handles this better than Kling, but neither is perfect at high magnification. Workaround: keep label text out of close-up frame, or use a mostly-static product motion where the label doesn't rotate significantly.

Brand color drift. A product with a specific pantone or brand color (a signature blue, an exact orange) can shift slightly between frames or between model runs. If brand color consistency is critical, always run a color correction pass in the Studio before export, and flag this to any performance marketing team using the videos in paid placements.

Complex multi-piece products. Products with several separate components (a skincare set, a multi-piece kitchen tool kit) are harder to lock in multi-reference conditioning. Generate single-hero-product videos and create a compilation in Studio rather than trying to animate the full set.

FAQ

Does Shopify support video on PDPs natively?

Yes. Shopify's native product media section supports MP4 video files directly. You upload video the same way you upload product images, and it appears in the media gallery on the PDP. The recommended specs are MP4, H.264 encoding, up to 1 GB per file, maximum 10 minutes. For PDP loops, you'll want something 15 to 30 seconds that auto-loops cleanly. Most themes support video autoplay in the hero media slot without any additional app.

What's the best aspect ratio for a Shopify product video?

1:1 for the PDP hero (safest for cross-device display), 9:16 for Reels and TikTok organic, 16:9 for paid ads. If you're only generating one cut, go 1:1. If you have budget for three cuts, generate each ratio natively rather than cropping from another ratio.

Can AI-generated product video include text overlay?

Yes, and 8frame's auto-caption tool handles this in the Studio step. For paid ad placements on Meta, keep any text to under 20% of the frame area to stay within their text-in-image guidelines (the same rule applies to video). For organic social, there's no platform limit. The strongest-performing short-form DTC videos in 2026 are using minimal text, one to three words max, because the caption does the work.

How do I keep the product looking consistent across multiple video cuts?

Use the same Seedream source frame as the reference input for every video generation job in that campaign. When you run Seedance 2.0 with that image as the reference still, it anchors the product appearance. The bigger consistency risk is across different campaigns or seasons, when you're generating from a new still. Store your reference frames in a shared folder and reuse them.

Is AI-generated video allowed under Meta ads policy?

Yes. Meta's current ads policy does not prohibit AI-generated video. The relevant rules are about deceptive claims and product misrepresentation, which apply equally to real and AI video. Don't use AI video to show functionality the product doesn't have or to imply results it can't deliver. The "made with AI" label Meta started testing in 2025 appears on some boosted content but doesn't restrict delivery or targeting. Check Meta's current policy page for updates before a major campaign.

Run this workflow on your products

The 8frame product video workflow is ready to clone. Drop in any product photo and it runs the Nano Banana, Seedream, and Seedance steps in sequence with the prompts pre-configured for the five most common Shopify product categories.

If you want to go deeper on model selection before you start, the best AI video generator guide for 2026 covers every model we tested on the same brief. For UGC-style ad creative built around your products, the UGC ad workflow uses the same Kling 3.0 and Seedance foundation with person-in-frame reference conditioning.

The traditional shoot timeline is 2 to 3 weeks from brief to approved video. The AI workflow is 30 minutes. Test one product this week.