ElevenLabs · AI voice and audio tools
Feed your best and worst ad copy into a custom GPT that writes new ads
Newly built; guest calls it really cool, no results yet
What they did
- Stage
- scale
In their words
How ElevenLabs Turns Feature Launches Into a Growth Engine – Luke Harries
David BarnardAnd I think more apps should be trying. And then how do you think about taking those launch moments and using the ideas, the new features and things like that in your paid assets as well? Because it's not just about getting earned media, it's about using these new features and that creativity to get attention, even if it's paid attention.
Luke HarriesYeah, 100%. So, whenever we do a launch, part of the big bit is, okay, how does this translate into our Meta ads, into our TikTok ads? And so, we have a ElevenLabs app, which has all the best creative models, whether it's speech, sound, effects, music. And as part of that we release, for example, the image and video feature within it. As part of that, you then go, "Okay, what are the ads that we should do? What are the different hooks?" And as you're thinking and iterating on that core messaging for the X post, you're also thinking and integrating for the different ad copy. And one of the new things we've actually just built, which has been really cool is we built a custom GPT where we've analyzed what's all the top performing Meta and Google copy and what's all our bottom performing Meta and Google copy.
Luke HarriesSo, we normally take what's our brand X post and then we pass it through this custom GPT, which will then iterate on it into the high-performing and [inaudible 00:05:08] and Google copy. And then another big bit, which I'd be happy to dive into is also we want to create lots of different variants of images, of videos, of statics in order to see what will drive conversion.
David BarnardYeah. I did want to dive more specifically into video. I mean, between Reels and TikTok, and I mean, even a good X post with a video is going to generally perform better than just a thread without any kind of example. How do you think about in generating these assets, when to use AI creative, when to lean into UGC, and when to just team create assets? On your blog post you talk about, put the founder in front of a camera and loom showing off the …
Get tactics like this every Monday
New tactics from the week's founder interviews, each linked to where it was said.
More from this episode
- Turn every feature launch into a video, thread, landing page and fresh adsCalled the heartbeat of growth; no numbers given
- Test ad formats with agencies, then hire motion designers to make variants in-houseMotion-design ads are the bulk of ads and work really well; no numbers
- Hire an in-house creator to make YouTube videos teaching the productChosen over AI avatars for quality; no results given
- Use AI to make ads for each audience, like podcasters or app makersDescribed as the way to scale creative; no results given
- Treat each ad channel separately and grow its spend 10-20% a weekGoogle search is still the biggest channel; no spend or return numbers
- Drop the translation agency; have an LLM translate app text on every code changeMuch cheaper and faster than the agency and paid tool it replaced
- Translate ads into other languages with AI dubbing and prompted copyGuest says it gets fantastic results; no numbers
- Run holdout tests in each ad platform and discount reported sales by the resultsExample: 500 reported conversions were only 200 incremental
Tags: meta, google, ad-copy, ai, custom-gpt