- The AI Ready CMO
- Posts
- Synthetic Media - from WOW to HOW
Synthetic Media - from WOW to HOW
How brands moved from AI demos to AI deployment in their campaigns
Peter Benei & Torsten Sandor
June 24, 2025
In partnership with
In 2024, AI was the showstopper at the festival of creativity at Cannes Lions. The annual gathering of the creative industry showed insane demos of their AI-fueled ideas. But it was still more of a visionary showcase.
This year, AI is part of almost all creative campaigns. As Alison Wagonfeld said at the event:
This year is the wow to the how.
It is no longer a question of whether AI will influence creative business and marketing. The question is how.
How to use synthetic media to scale creativity?
How to localize content within hours, not weeks?
How to use avatars, agents, and generative video in campaigns?
By the way, for the latter, check how Mars “cloned” Jose Mourinho for a Snickers campaign and won multiple awards at Cannes.
This week, you will see a lot from Cannes, but through the special lens of AI-generated video. A year ago, it was in pre-baby shoes. Now, it is ready to claim the internet with almost instantly generated synthetic media.
Since creative production by default gets automated, what stays? The “creative” part. The “strategic” part. Ideas still matter, even more so now when all execution will be driven by AI, in an instant, flooding everything everywhere all at once.
Taste and creativity will top speed and execution. Always.
 Peter Benei | Co-author Consultant at Anywhere Consulting |  Torsten Sandor | Co-author Senior Director of Marketing at Appen |
A quick word from this week’s sponsor:
Turn AI Hype Into Business Impact
AI is more than just using ChatGPT.
It’s about understanding how the technology works—and applying it in practical ways to solve real business problems.
The AI for Business & Finance Certificate Program from Wall Street Prep + Columbia Business School Exec Ed helps you move beyond the hype.
You’ll demystify today’s most important tools and learn just enough Python to streamline workflows, automate tasks, and make smarter decisions.
No coding experience needed—you’ll get step-by-step guidance from world-class faculty.
👉 Save $300 with code SAVE300.
When AI avatars break records in live selling
Digital Doubles Outsell the Real Deal in Live Commerce
On June 15, China’s e-commerce trailblazer Luo Yonghao handed the reins to two hyper-realistic AI avatars. One modeled on himself, the other a bespoke co-host, for a six-hour marathon stream on Baidu Youxuan.
Fueled by Baidu’s ERNIE and Xiling digital-human suite, these virtual hosts showcased 133 products, spun out 97,000 characters of on-the-fly dialogue, and executed 8,300 distinct avatar gestures. They drew 13 million viewers and racked up ¥55 million ($7.6 M) in sales, surpassing Luo’s best live-hosted performance in just 26 minutes.
Footage from the stream
Why it matters for you
Conversion at scale: AI presenters can sustain viewer engagement and match human conversion rates without breaks.
Cost predictability: Swap variable talent fees for fixed licensing and compute costs.
Around-the-clock reach: Run streams with the same digital talent in multiple time zones and languages.
Next-gen commerce framework: Leverage live selling without human constraints, from fatigue to scheduling conflicts.
What to do next?
Pilot an avatar cameo: Integrate a short AI-hosted FAQ or product spotlight in your upcoming livestream to benchmark dwell time and CTR.
Secure IP early: Nail down usage rights for likeness, voice, and script before cloning any influencers or execs.
Build a cost comparison: Model annual spend on avatar licensing + cloud rendering versus live host fees and markups.
Map global hotspots: Pinpoint two non-English markets where digital hosts can bridge language gaps and auto-localize content for seamless engagement.
Veo 3 Everywhere
AI video creation goes native in Shorts and Canva
YouTube CEO Neal Mohan announced at Cannes Lions last week that Veo 3, the DeepMind–powered text-to-video engine delivering 1080p footage with native audio, will be built directly into YouTube Shorts. This will let creators generate polished micro-ads without leaving the app.
At the same time, Canva introduced a “Create a Video Clip” button for Pro, Teams, and Enterprise users, granting five Veo-powered videos per month, complete with synced audio and your brand-kit assets.
Neal Mohan at Cannes Lions
Why it matters for marketers
With Veo 3 in Shorts, any marketer can produce broadcast-quality clips in seconds and keep valuable watch-time and engagement data within Google’s ecosystem.
Canva’s new feature turns your familiar deck-making workspace into a seamless video-production pipeline. No external agencies are required.
What to do next?
Pilot a Veo-powered Short in your next campaign to benchmark AI-native clip performance against your current video assets.
Enrich each clip’s metadata—titles, descriptions, and tags—so Veo creations surface in Shorts search and recommendations.
Feed engagement metrics from your AI-generated videos into your CDP to sharpen audience retargeting and offer sequencing.
Preload your complete brand kit (fonts, colors, logos) into Canva to ensure every AI clip stays on-brand.
MiniMax’s Hailuo 02 Leaps to #2 on the Video Leaderboard
A new challenger overtakes the giants
Shanghai-based MiniMax has released Hailuo 02, a text-to-video model producing sharp and smooth clips that surpassed Google’s Veo 3 on industry benchmarks.
With better motion physics, richer color grading, and aggressively priced early API access, it’s suddenly the go-to alternative for teams watching output quality and budget.
Hailou’s examples
Why it matters for marketers
Model supremacy is now up for grabs.
The gap between best-in-class and close-second is measured in weeks, not quarters. New players are leapfrogging giants by focusing on overlooked performance edges like motion realism and post-production finish.
Committing too hard to one vendor too early may limit your ability to stay current or competitive.
What to do next?
Form a small “AI scout” group to test emerging video models monthly and report on performance and pricing.
Push for contract terms with 30-day exits and usage-based pricing from current providers.
Use a fixed creative test set to benchmark speed, cost, and safety filters across all tools under evaluation.
Midjourney v1 Video + TikTok Symphony
Social creative gets a supercharger
Midjourney just introduced its new “/video” command, transforming any image into a 10-second loop with fluid 60 fps animation. Audio support is coming soon.
At the same time, TikTok launched upgrades to its Symphony suite, baking image-to-video, text-to-video, and product-focused templates directly into Ads Manager.
Example footage from Midjourney
Why it matters for marketers
Video outperforms static everywhere, especially in social.
These new tools let you breathe life into top-performing stills and UGC, without needing a motion designer.
You’re no longer stuck choosing between quality and speed: teams can turn raw creative inputs into polished clips daily, using the tools they already use.
What to do next?
Convert your top-performing image carousels into motion assets using TikTok Symphony this week.
Build a shared doc with winning Midjourney prompts and Symphony settings for your creative team.
Schedule bi-weekly refreshes of your social ad creative to keep engagement high and feed new signals to the algorithm.
Our favourite tool picks for AI-generated video
HeyGen nails lip-sync and localization. Great for global video ads, fast.
Synthesia is built for scale and compliance, perfect for internal communications and training.
AI Studios is scrappy, fast, and marketer-friendly. Ideal for quick campaign content.
Luma AI × HUMAIN: A $-Billion Push to Industrialise Synthetic Media
Scaling culturally-adapted video across regions, fast
Saudi-backed HUMAIN is teaming up with Luma AI to launch a global joint venture to turn Luma’s Ray2 and Genie models into commercial workhorses for advertising, film, and e-commerce.
The deal, announced by Sir Martin Sorrell at the Cannes Lions, signals a serious push to industrialize synthetic media and unlock scalable, hyper-local creative faster than ever before.
Martin Sorrell at Cannes Lions
Why it matters for marketers
Localization has always been slow and expensive. This changes that.
With HUMAIN’s infrastructure and Luma’s generation models, marketers can now produce culturally-tuned variants with tailored language, visuals, and casting cues in a fraction of the time.
This means less localization budget, more distribution spend, and quicker turnarounds across priority regions.
What to do next?
Identify your top five markets where localized creative could unlock incremental growth.
Model cost and turnaround time using HUMAIN–Luma against your current localization spend.
Review and update rights agreements to make sure global synthetic variants stay compliant.
Inside the Perplexity Fellowship
How to build AI ops into your organization
As part of the Perplexity Business Fellowship Program, we joined a fireside chat with Eric Glyman, CEO of Ramp (recently valued at $13B).
Here are our 9 takeaways from the workshop for marketers:
1. Ramp Is a Productivity Company, Not Just Fintech
Their mission is automation-driven time and cost savings, not just managing spend.
» What marketers can learn: Reframe your value prop around outcomes, not just category labels.
2. AI Agents Are Replacing Manual Finance Tasks
Ramp is automating actions like hotel bookings and receipt handling via agents.
» What marketers can learn: Start embedding agents in repetitive ops: FAQs, lead routing, approvals.
3. They’re Rebuilding Interfaces Around LLMs
Rigid UI is out; conversational, agent-led UX is in.
» What marketers can learn: Design marketing tools for prompts, not clicks.
4. Their Roadmap Is Measured in Saved Time & Money
Every feature must earn its spot by reducing cost or workflow friction.
» What marketers can learn: Prioritise martech based on impact per hour, not feature count.
5. They Hire for ‘Mispriced Talent’
Ramp targets smart people overlooked by traditional filters: immigrants, young grads, second-place finishers. Underdogs with something to prove often outperform stars with ego.
» What marketers can learn: Recruit based on potential velocity, not just credentials.
6. They Filter for AI-Native Thinkers
Some interview tasks require LLMs to solve, ensuring fluency from day one.
» What marketers can learn: Make AI tooling part of your hiring filters, not an optional skill.
7. They’re Building Toward Agentic Commerce
Agents (human or AI) will transact, approve, and decide under secure rules.
» What marketers can learn: Start prototyping agent-led journeys in customer support and upsell flows.
8. Small, Committed Partners Beat Big, Passive Ones
Ramp grew faster by picking leaner partners who had skin in the game.
» What marketers can learn: Bet on growth-aligned collaborators, not just brand names.
9. Beginner’s Mindset Wins
They encourage junior devs to embrace AI tools early to compound their impact.
» What marketers can learn: Don’t wait to be an expert. Test, tinker, and scale as you go.
CMO Tips
How to clean and prep your data for AI-driven marketing?
Your AI is only as smart as your data. Here is a simple framework for improving your data for AI marketing operations.
The guiding rule: you don’t need to fix all the data issues at once. Start with one loop first, and roll to the rest later.
Audit. Map your data. Spot bugs. Make problems clear. You can’t clean what you can’t see.
Consolidate. Merge data into one source of truth. Keep what matters, trash what doesn’t. You don’t need 12 lead stages. You need 3.
Tag. Add metadata, such as stages, channel, persona, etc. Clean data is labeled and categorized, and tagging turns data into AI lingo.
Assign. Every data set should have a human owner. Ownership means updating, cleaning, and tagging. Without ownership, data decays.
Track. Set review cadences and monitor usage: test data hygiene and usability for AI. Tracking turns cleanup into a habit.
Rapid-Fire News
AI News You Need to Know
Quick scans on things that mattered last week.
1. Reddit Unlocks Community Intelligence
Reddit just launched a new platform to turn billions of posts and comments into structured insights for advertisers.
» For marketers: Positive user comments can now auto-appear under your ads. Early tests show a 19% CTR lift.
2. Amazon’s AI-Driven Workforce Shift
In an all-staff email, CEO Andy Jassy confirmed AI will replace some corporate roles to drive efficiency.
» For marketers: This is your reminder. AI isn’t just about growth. It’s changing team structures and skill needs.
3. Adobe Firefly goes mobile
Firefly’s new mobile app lets users create and edit generative images on the go.
» For marketers: Empower your creatives with complete production tools they can use from anywhere. No laptop required.
4. GPT-5 Is Coming This Summer
Sam Altman confirmed GPT-5 is launching this summer, promising a major jump in capabilities.
» For marketers: Expect stronger summarization, faster outputs, and richer creative assistance—plan your upgrades.
5. The Cognitive Cost of AI Reliance?
A new study claims heavy ChatGPT use might dull critical thinking, though the science is still unsettled.
» For marketers: Frame AI as augmentation, not autopilot. Human judgment still drives strategy.
Autonomous Leadership Advice
We published our Autonomous Leadership Manifesto. It helps CMOs to guide their teams through an AI-powered future.
Based on the manifesto, this is our advice for this week:
Great leaders don’t scale by doing more. They scale by designing better.
Your job isn’t to be the fuel of the machine. Your job is to architect the engine. Every time you step in to fix the same issue twice, it’s not a team problem. It’s a system problem.
Delegation without intentional design is just outsourcing chaos. The goal is to build workflows that rely on structure, not energy. Where the right outcome happens by default, and the wrong one hits friction.
As your system matures, your involvement should shrink, not grow.
Your leadership design isn't finished if you are constantly consulted, involved in firefighting, or needed to make basic decisions. The true mark of effective leadership is when your team performs consistently, even in your absence.
Systems should hum quietly, not scream for attention. Design the work once so it scales without applause or panic. Design the system so your absence is a feature, not a flaw.
That’s a wrap for this edition.
If this helped you spot one experiment to run, one tool to try, or one partner to question, our job’s done.
Forward it to a teammate who needs to see it, and make sure you’re signed up to catch the next drop on Tuesday.
Until next week,
Peter and Torsten