Veo 3.1 for a marketer: 8 prompts, 4 restrictions and a real bill per quarter
A complete guide to Veo 3.1 for advertising production: 8 prompt templates with exit examples, 4 limitations that are not written about in marketing reviews, and my cost grid - Q1 2026 I generated 47 videos for $312.

During the first quarter of 2026, I generated 47 advertisements in Veo 3.1. Budget - $312, excluding subscription. A similar volume through the production team in 2023 would cost 2.4M rubles and would take 14 weeks. This is not an “AI future vs past” comparison. This is a grid of actual calculations from my Q1 billing statement.
I’m not writing another review of “10 Veo 3.1 features that will revolutionize marketing.” I am writing this guide for marketers who already have a budget for performance and who want to reduce production by 5–10 times. Here are 8 prompt templates with exit examples, 4 restrictions that are not written about in marketing reviews, and my cost grid. No hype.
Veo 3.1 does not “replace the videographer.” It replaces half of the pipeline - filming, editing, color correction, sync audio. The brief and final taste are still up to the individual.
What is Veo 3.1 in a marketing sense
Veo 3.1 is a text-to-video model from Google DeepMind that generates videos up to 60 seconds in 1080p with a built-in synchronized audio track (ambient, noise, sometimes speech). Available through Vertex AI, Gemini API and through third party aggregators like kie.ai. Appeared in November 2025 as an update to Veo 3.0; The key change is the stability of quality when generating a video longer than 8 seconds.
What does this mean in practice for a marketer:
- One prompt = one 8-second HD video with embedded audio in 90 seconds of generation
- $0.50–0.75 per second depending on the provider (cheaper through Vertex, more convenient API through kie.ai)
- The quality in 90% of cases is such that only video professionals can distinguish it from live filming - in focus groups of ordinary viewers no one guesses
- The API does not require calls to the sales manager, payment by white/gray, you can pay with a Mastercard through Wise
8 prompts that work on performance tasks
I have all the prompts below in Notion right now. Prompt structure in Veo 3.1: subject + setting + composition + lighting + camera + style + duration. No “AI is changing the rules of the game” - specific designs for specific tasks.
Prompt 1. Product hero shot for e-com
Slow rotating shot of [product], floating against dark gradient background, cinematic studio lighting from the right, subtle particles in the air, premium product photography style, 8 seconds, smooth seamless loop, shot on Sony FX9, depth of field, no text overlays, photorealistic, 4K detail.Where to use: website banners and tests of creatives in VK and Yandex Direct. The speed of video production and savings compared to a studio should be measured separately: there is no such comparison in the ID Store materials.
Prompt 2. Lifestyle video for VK Clips
Young woman, 28, casual style, sitting in a sunlit Scandinavian apartment, drinking coffee, looking at her phone, smiles softly, warm morning light through window, vertical 9:16 frame, shot on Canon R5 50mm f/1.4, intimate cinematic mood, 6 seconds, soft jazz ambient audio.Where I use it: vertical video for VK Ads. For different campaigns I change the age, context, product in hand. One template - a dozen variations for A/B testing.
Prompt 3. B2B-serious scene for Telegram Ads
Marketing director in a modern minimalist office, looking at a dashboard with charts on a large monitor, soft natural light from floor-to-ceiling windows, focused expression, slight zoom-in over 5 seconds, shot on RED Komodo 35mm, professional confident atmosphere, no text overlays.Where to use it: Test short explainer videos for B2B audiences and compare them with other formats based on the cost per qualified lead. For MM AI Trading, there is no comparable test with a lifestyle video in the available materials.
Prompt 4. Tutorial video with screen capture
Hands typing on a MacBook Pro keyboard, top-down view, dark wood desk with notebook and coffee mug, dimly lit room with one warm desk lamp, slow shallow depth-of-field zoom on the keyboard, no screen visible, focused intentional movement, 8 seconds, ASMR keyboard sound.Where I use it: tutorial series, videos for landing pages about software/courses. I often combine it with screen recording - Veo gives a hands-shot, the rest is filmed through OBS.
Prompt 5. Emotional B2C video for retargeting
Group of friends laughing together at an outdoor BBQ in the evening, warm string lights, golden hour transitioning to dusk, handheld camera with subtle shake, Kodak Portra 400 film grain, intimate moments captured naturally, 7 seconds, ambient laughter and clinking glasses.Where I use it: B2C retargeting on people who visited the site but did not buy. Emotional load through the atmosphere - without direct selling text.
Prompt 6. Tech / SaaS demo for the Habr audience
Abstract data visualization, glowing nodes connecting through a dark space, neon emerald accents on connection lines, slow rotation revealing depth, cinematic technological atmosphere, premium tech aesthetic, 6 seconds, low ambient electronic drone, no text or logos visible.Where I use it: B2B SaaS, fintech, any product with an “invisible” essence. A replacement for boring animated diagrams in After Effects.
Prompt 7. Local Russian context
Inside a Moscow metro train, late afternoon, passengers in business attire reading on phones, soft yellow interior lighting, slight motion blur from train movement, view from a static camera mounted at door level, 8 seconds, ambient train sounds.Where I use it: B2C campaigns with geographic reference. Veo 3.1 knows well the specifics of the Moscow metro (cars of the “Moscow-2024” type). It’s already worse in St. Petersburg.
Prompt 8. The “before and after” cycle
Split-screen 8 seconds: left side shows a cluttered messy office desk with papers and old computer, right side shows the same desk after — minimal, clean, modern setup with one tablet and one notebook, both halves filmed in matching lighting, slow synchronized zoom-in on both sides, professional cinematic style.Where I use it: products with a “transformational” promise - cleaning, courses, organization. Strong visual triggers.
4 limitations of Veo 3.1 that are not written about
- The text in the frame is a disaster. Veo cannot generate readable text in videos. Any inscriptions on signs, screens, badges turn into an unreadable mess of pseudo-letters. Solution: do not write “text in frame” in the prompt; add text overlays in After Effects/CapCut after rendering.
- Faces are unstable for 8+ seconds. If there is a person in the video, at 5-6 seconds the facial features begin to “float”. On vertical formats this is less noticeable; on headshots - critical. Solution: limit the duration of frames with a face to 3-4 seconds, move on to details (arms, torso, surroundings).
- Russian language in speech does not work. If in a prompt you ask a character to say something in Russian, you get either gibberish or English with an accent. Solution: voiceover is done separately via ElevenLabs, superimposed in editing.
- Branded products - contain hallucinations. The query “iPhone 15 on the table” will return a device similar to the iPhone, but with distorted details (logo, buttons, camera island). This is legally problematic, especially for advertising of specific brands. Solution: either don’t draw the brand details, or shoot the real product + use Veo for the environment.
Real bill for Q1 2026
| month | Rollers | Total seconds | Cost, $ |
|---|---|---|---|
| January | 14 | 98 sec | $72 |
| February | 19 | 142 sec | $104 |
| March | 14 | 186 sec | $136 |
| Total Q1 | 47 | 426 sec | $312 |
Average cost of one 8-second video: $6.6. A similar video through the production team in 2023: 65,000 rubles. 100× difference. A complete breakdown of the economics of my AI stack is in article “One marketer = team of 5 roles”.
Comparison with competitors
| Model | Quality | Price for 8 sec | Sync audio | Where do I use it? |
|---|---|---|---|---|
| Veo 3.1 | 9/10 | $6 | yes | main workflow |
| Sora 2 | 8.5/10 | $8 | partially | reserve for specific scenes |
| Runway Gen-4 | 7.5/10 | $5 | no | I don't use it anymore |
| Kling 2.0 | 7/10 | $4 | no | I test periodically |
A complete overview of all the tools in the stack is in mine comparison of AI tools 2026.
Workflow: from prompt to finished video for VK Ads
- Brief. What product, what target, what emotion. 15 minutes for 5 creatives.
- Prompt template. I take it from the library above and adapt it to the product. 5 minutes per video.
- Generation in Veo. 90 seconds, at the same time I put 3-5 variants of one prompt.
- Selection. Out of 5 options, I choose 2–3 suitable ones. 10 minutes.
- Voiceover at ElevenLabs. If you need a speech - 30 seconds.
- Text overlay in CapCut/After Effects. 10–20 minutes.
- Upload to VK Ads. 5 minutes.
Total: 5 ready-made creatives for the campaign in ~1.5 hours of my time and ~$30 for the API. The same amount through production in 2023 - a week of team work and ~325,000 rubles. The difference is “200×” according to unit-economics, as I wrote in main article about the AI stack.
If you are currently paying a production team 50K+ ₽ for one commercial, you are not 5 years behind the market. You are losing competitors who already consider unit-economics in a new way. Would you like me to analyze your current production flow in 30 minutes? discovery-colle — I’ll show you which 3-5 steps you can remove in the first month.
Related materials: ROAS calculator, CTR benchmarks in VK Ads, a complete guide to performance marketing.