One key improvement would be more detailed editing control. While automation is great, users sometimes need to fine-tune scenes, timing, or transitions without regenerating the whole video.
Another area is more natural and diverse AI voices and avatars. Expanding emotional range, accents, and speaking styles would make ads feel more realistic and better suited to different audiences.
It could also improve consistency across multiple ads in a campaign. Right now, keeping branding, tone, and style perfectly aligned across several variations can require extra manual adjustments.
Finally, faster comparison tools (like easy side-by-side previews of different versions) would help users decide more quickly which ad performs best.
These improvements would make it even stronger for both beginners and marketing teams.
Creatify
Hey Product Hunt, I'm Yinan, founder of Creatify.
Quick backstory: we've spent 3 years building Creatify. Started with URL to video, then avatars, Ad Clone, and a full creative agent. 3M+ users have used the platform.
But one thing kept bothering me. Our users would make great ads, launch them, then go back to staring at dashboards for hours trying to figure out what was working. The creative side was solved. The media buying side was still manual, scattered, and slow.
So we built AI Media Buyer an agent that plugs into your ad accounts and actually does the work.
How it runs:
Connects to your Meta, Google, and TikTok accounts (OpenAI ad accounts too now 🤫). Reads every campaign, finds what's scaling and what's quietly bleeding spend.
It doesn't just alert you. It acts recommends what to cut, what to scale, and can generate new creatives from your winners and launch them in one loop.
It remembers your account and past conversations. Recommendations compound the more you use it.
Cross-references your ad data with Stripe and CRM, so you see real numbers, not just Meta's attribution.
The part I'm most proud of: this is the only AI media buyer with a built in creative engine. Most tools stop at "here's what happened." Ours finds a winning hook, makes variations, and launches them. No other tool does that.
One early user found a campaign they'd paused was actually producing their highest-value buyers Meta was undercounting it. The agent caught it.
I'll be in the comments all day. What's the most painful part of managing your ad accounts right now?
@nyn531 What always burns me when I hand paid spend to an automated system is the learning phase, where it quietly torches budget figuring out what converts. Can I cap daily spend and lock which creatives it's allowed to test, or does it need free rein to get smart? Curious how much manual override you left in for the operator.
Creatify
@artem_fedorovich Great question. It's a copilot experience. It will ask for your approval before making any actions to your account. You can create "routine" within the AI media buyer and a routine can run a playbook defined by you such as cap daily spend etc.
RiteKit Company Logo API
@nyn531 Foxy chose your launch out of today's batch 🦊 — giving small brands the media buyer they could never afford (that improves daily) is a clean, real wedge. More than an upvote: a free launch kit + your launch rewritten for where advertisers actually are → https://foxplug.com/g/81f2ebe3aced4ea6ab3b
FoxPlug, from one URL: video, native posts, citation-grade articles, GIFs, pitches to journalists on your beat, and getting cited where AI answers your category — and it keeps going every time you ship:
$29/month · https://foxplug.com
curious what happens when a winning hook on meta just flops on ticktok, does the creative engine adjust format or just reuse it as is"
Creatify
@wyatt_cameron Good question. The agent treats them as separate signals a hook winning on Meta doesn't get blindly pushed to TikTok, since it's reading performance per platform. When it generates variations from a winner, [VERIFY: it adapts format/style for the target platform aspect ratio, pacing, hook style] rather than reusing the exact asset. And if the TikTok version flops, that's data too it'll flag it the same way it catches anything bleeding spend, and iterate from what TikTok is actually responding to.
congrats. quick question: how do you balance automation with account-specific strategy? For example, can I set guardrails like target ROAS, audience exclusions, or preferred creative styles so the AI's actions align with our brand and long-term goals?
Creatify
@swati_paliwal Thanks! Yes, you can set constraints like target ROAS and exclusions, and the agent works within them when it recommends. On creative style, it learns from your account: it generates from your actual winners, so brand direction comes from what's already performing. And nothing executes without your approval, so strategy alignment gets checked on every action, not just at setup.
how does this handle attribution windows? meta's 7-day click vs what stripe actually shows never matches for us
Creatify
@elle_go Yeah, this mismatch is basically why the Stripe/CRM connection exists. Instead of trusting Meta's 7-day click as truth, the agent pulls what actually hit your revenue and reads performance against that. That's how the paused campaign story in the founder comment happened Meta's window was undercounting a campaign that Stripe showed was producing the highest value buyers. So recommendations get made off real numbers, not platform reported ones.
congrats on the superb launch! Curious how it decides something is actually a winner before it scales budget into it. What window and what signal does the agent trust before it commits spend?
Creatify
@artstavenka1 Thanks Art Stavenka 🙏 Two parts: it doesn't trust platform numbers alone, winners get verified against what actually hit Stripe/CRM, since Meta's 7-day click and real revenue often disagree. And it never commits spend on its own it shows the signal and reasoning, recommends the scale, you approve. The trust call stays with you, not a buried threshold.
"strength is clearly the closed loop between finding a dead creative and regenerating one in the same system, weakness is trusting one agent to touch spend across four platforms at once without a human checking first, that's a lot of trust to earn early
Creatify
@ian_maxwell2 Fair concern, and it's exactly why nothing touches spend without approval. The agent audits, recommends, and preps the action - cut this, scale that, launch these variations but you approve before anything executes. The closed loop is the automated part; spend changes keep a human on the trigger. Trust it more over time? Your call not a default.
Creatify
Hey PH 👋 Chaz here, CTO/co-founder at Creatify.
Wiring an LLM to ad APIs over MCP is a weekend project now. Making the agent good is the hard part, and that's where the work went.
Three things that ate my time: getting it to genuinely understand account structure — hierarchies, budget inheritance, what each knob actually does, because get that wrong and every downstream call is confident nonsense; wiring it into our creative agent so the loop actually closes (spotting a dead creative is easy, being able to regenerate it in the same system is the part most tools can't do); and getting it to shut up, so it hands you the call and the reason instead of six paragraphs.
If you've burned real budget on ads, come try to break it. Most interested in where it misreads an account. 🙏