Guide: Creating Product Videos with Veo 3.1

Understanding AI-Powered Video Production

Video content testing for ecommerce typically involves creating multiple variations to identify winning concepts before investing in final production. Testing different hooks, product angles, and messaging approaches can require weeks of iteration, making it difficult to validate creative direction quickly.

Veo 3.1 enables rapid video ideation and A/B testing directly from product photos. This allows you to quickly test multiple creative concepts—different hooks, visual styles, and messaging angles—before committing to final production with your creative team.

This guide covers the workflow for using AI to rapidly prototype and test video concepts.

What You'll Need

Technical Access (Choose One)

• Gemini app: Best for single clips and quick iteration

• Flow (SceneBuilder/Flow editor): Best for multi-shot storyboards and bridging

• Veo Studio/AI Studio: Best for prompt testing and experimentation

Visual Assets

• Hero product photos with clean backgrounds

• 1-3 reference images maximum (hero product + lifestyle/hand/face shots)

• Tightly cropped, well-lit images

Key Veo 3.1 Capabilities

• Richer native audio: Generate ambience, SFX, and short voiceover with video (useful for ASMR/unboxing drafts)

• Image→video & frame bridging: Morph between start and end frames for smooth transitions

• Scene Extension: Extend clips from final frame to continue action

• Fast variant: Faster renders for iteration, full fidelity for finals

Technical Specifications

Clip length: Optimized for 4-8 seconds per shot. Plan multi-shot ads as multiple beats.

Aspect ratios: 9:16 for Reels/TikTok; 16:9 and 1:1 supported. Most default to 720p; 1080p available in many cases.

Reference images: Attach 1-3 reference images (hero product + 1-2 lifestyle/hand/face shots) to preserve continuity.

Native audio: Generated audio works for drafts; use professional VO for final brand ads.

Continuity tools: Use First→Last frame bridging or Scene Extension in Flow to join clips.

Step 1: Plan Video Beats

Break your ad into 4-8 second beats. Treat each beat as a separate generation.

Use ChatGPT or similar to create the beat plan with timing, action, and transitions specified for each segment.

Step 2: Structure Your Prompts

Use this three-part template for every shot:

Context: [One-line description - product + hook]

Visual: [Duration] [aspect ratio]. [Camera/lens]: [specific action]. [Lighting]. [Material/finish]. Use reference images: [filenames].

Audio + CTA: [SFX/VO direction]. End overlay: [text/CTA]

Prompting Rules:

• Always attach hero image and lifestyle/hand reference

• Be explicit with camera & lighting (specify lens, motion, light quality)

• Specify material & finish to avoid plasticky results

• Describe hands/interactions with detail (skin tone, nail style, scale)

• Keep spoken lines ≤2 short sentences for lip sync

Step 3: Generate Video Clips

These will largely depend on your product and the beats you have developed in Step 1. I am including these prompts to give you an idea about the detail/type of context needed in the prompt.

Example A: Unboxing + Texture Demo

Context: Quick unbox & texture demo — [Product Name]

Visual: 6s vertical (9:16). Overhead 50mm: female hand (medium skin tone) opens jar (0-1s), scoops product (1-3s), macro of product melting on palm (3-5s), gentle wipe showing finish (5-6s). Soft natural window light, warm highlights. Use reference images: hero.jpg, hand_ref.jpg.

Audio + CTA: Crisp box-open and scoop SFX, whisper VO: "so silky" (≤2 words). End overlay: "[Brand] — [Tagline]. [LINK]"

Example B: UGC Testimonial

Context: "I'm not paid for this — just sharing what I love."

Visual: 6s vertical. Selfie-style talking head, kitchen or bedroom natural light, handheld slightly shaky, close up. Camera: phone front, 35-50mm equivalence.

Audio & CTA: VO script: "First, the scent is spa-like, not overpowering. Second, it actually removes my makeup." End overlay: "Full routine link."

Example C: Ingredient Spotlight

Context: Why I love this — [Key ingredient] in [Product Name]

Visual: 6s vertical. 50mm macro label close-up (0-2s), texture swipe on finger (2-4s), smiling pinch of cheek showing bounce (4-6s). Natural daylight, clean background. Use reference image: hero.jpg.

Audio + CTA: Calm VO: "[Ingredient] — helps support the appearance of firmness." End overlay: "Learn more: [LINK]"

For the rest of the SOP + Video Examples of the prompts, click here.

• A crisp playbook to turn one idea into viral video with Veo 3 visuals and Sora 2 motion, from hook to remix.

• California becomes first state to regulate AI companions

• The Impact of Artificial Intelligence on Human Thought

• AI written articles now outnumber human ones

• The 2025 Deloitte Holiday Retail Survey has some interesting insights.

• The Tilly Norwood “AI actress” was marketing theatre, not a breakthrough, and human taste still sets the bar.

• Unitree H2 Destiny is born, a bionic humanoid that can serve you on your day to day.

  

• Wikipedia has lost about 8% of human traffic due to AI. Is the internet on the way to the grave?

• IMPORTANT: Facebook’s new button lets its AI look at photos you haven’t uploaded yet. Don’t let Zuck train on your data.

• OpenAI’s ChatGPT Atlas is a Mac browser with a ChatGPT sidebar and agent that acts on pages, powerful yet trust heavy and sticky to switch.

  

• Grocery inflation bites again, shoppers trade down and coupon, retailers push value lines and curated lists to hold baskets.

⛺ AI camp - Access multiple LLMs, assistants, and tools, for teams of all sizes.

⚙️ Hocoos: Create business-ready websites in seconds, filled with AI-generated content, captivating design elements, and eye-catching images.

🦔 Keepi - A personal knowledge assistant

💡 Visual Electric allows users to be inspired by its library of stunning imagery and prompts, and remix their ideas through iteration.

📧 SaneBox - Read the important emails in your inbox.

😍It was great, all I needed

🙂Good, not great

🙄It sucked