While still no substitute for the supercomputer operating between your ears, AI is emerging as a powerful ally in the creative process. Whether you’re looking to generate attention-grabbing headlines, killer visuals, or SERP-topping web content, getting the right tools in place can save you a lot of heavy lifting.
In this piece, we’ll round up some of the most exciting platforms and services we’ve come across, outlining their use-cases, performance (based on overall G2 ratings), and pricing, so join us as we explore the blurry frontiers of technology and imagination.
The advent of Large Language Models (LLMs) has brought a dizzying array of writing assistants to market. Given the ubiquity of ChatGPT, and the native AI writing assistants now available in market-leading tools like SEMRush, Hubspot and Notion, you might wonder what tangible value the more niche platforms are bringing to the party.
The answer appears to be training data. ChatGPT is trained on basically the entire internet (up to September 2021), making it a great general purpose tool. Specialised training data on the other hand, yields more specialised output.
Take Jasper for example - one of the most widely used writing assistants. Their LLM is trained on a massive dataset of blog posts, news articles and social media posts, yielding output more natural and engaging than more general models. Platforms like Copy.ai and Kafkai take a similar approach, each offering their own content templates and research tools.
We’ve had a go at seat-racing these platforms, but the only solid conclusion we’ve been able to arrive at is that there’s not a huge amount to choose between them. The market is immature, and we expect to see providers becoming more concentrated over time. In the meantime your choice of tool may come down to personal preference. Most of them offer free trials, so the best advice is probably to play around with them a bit and see what works for you.
If you’re primarily writing copy for SEO purposes, you may want to look at one of the more specialist platforms. Tools like Frase.io, MarketMuse and Growth Bar use AI to analyse SERPs and recommend content for you to publish on your domain to boost your ranking for certain keywords, supported by a LLM-powered writing assistant.
Surfer SEO takes a similar approach, but focuses more on optimising existing content. Their AI effectively reverse engineers the NLP that Google uses to rank sites, then uses that to power specific recommendations.
It’s difficult to get a handle on how well these tools actually work comparatively, but their users seem pretty happy, which is a good sign. In the medium-term there’s probably some danger of saturation: the more they’re adopted, the less of an edge they will offer. Then again, SEO has always been an arms race; AI-optimisation may just become the new entry requirement.
Image Generation & Editing
Loved and feared by designers everywhere, visual foundation models (VFMs) take text input and turn it into image output. The platforms are moving so fast that it almost feels pointless to write anything about their current level of performance. Suffice to say, the big players are:
MidJourney - Created by an independent research lab in SF, MidJourney is our current platform of choice. It’s hugely flexible, and capable of translating a vast array of prompts into accurate, compelling artwork, although getting it to do exactly what you want can be somewhat tricky. All you need is a Discord account to get started on the free trial - paid plans start at $10/ month.
Dalle - DALL-E 2 is Open AI’s latest VFM, and while fun, we haven’t found the platform as easy or delightful as MidJourney. That may be a preference thing, but it certainly doesn’t offer the kind of significant improvement that might motivate us to switch, and an overall score of 3.9 on G2 might suggest we’re not alone. You can try the platform for free - thereafter it operates on a credit system. If you’re only generating the occasional image, it may be a more affordable option than MJ. The platform also now powers image generation on Shutterstock.
Stable Diffusion - The latest serious entrant in the VFM premier league is Stable Diffusion by Stability.ai. You can try it for free as part of the Dream Studio suite, then you’re into credits, which seem pretty affordable. While as with Dalle, the output doesn’t offer any compelling improvement on MidJourney, it does offer some interesting peripheral features including Inpainting (manually editing the rendered image) Outpainting (extending outside of the original image), and Image-to-image prompts.
We’re yet to test these out fully, but will be keeping a close eye on development. Our advice is to watch the space for any major advancements, and to play around with what’s available in the meantime; they’re a lot of fun.
In terms of image editing (rather than generation) Photo Room offers a suite of free AI-supported tools, including background removal, retouching and object removal. They even have apps for iOS and Android, much to the delight of #grammers everywhere.
It’s worth noting however that the big players like Adobe are already catching up. It will be interesting to see whether Photo Room ends up being a genuine competitor, or an acquisition target.
One of the scarier developments in generative AI is the emergence of video foundational models. Similar to the VFMs above, these allow users to create videos from text prompts (scripts, essentially). The potential for misuse is obvious (particularly around elections), and you will probably have come across a bunch of deep fakes already (whether you knew it or not).
There are of course more benign use-cases, and the technology is something of an equaliser. Even in the digital age, video production is expensive enough to make it a hefty investment for most startups. Tools like Pictory, Synthesia and HeyGen lower this bar, offering a simple suite of editing tools on top of the generative capabilities.
Again we are yet to fully bottom these out, but as of today they seem particularly well suited to demo videos.
Some of us remember a time when all websites had to be hard coded from scratch. Templated builders like Wordpress, and later WYSIWYG editors like Squarespace and Wix changed all that, standardising the component-based, low or no-code CMS.
AI builders are the next logical step along that path, turning no-code solutions into no-design solutions. Early entrants to the space include Hostinger, 10Web and Durable, which will generate a website for you in under a minute, based on a few basic fields like company name and sector.
The output isn’t bad at all, especially when the editing is supplemented by the copywriting ability of ChatGPT (and co), and/ or a VFM like MidJourney.
Overall these platforms are a solid solution if speed is your imperative, but don’t expect the output to be featured at Cannes any time soon. If you’re looking to create a world-beating brand or digital experience, there’s still no substitute for a skilled, human designer (yet).
Moreover, it’s hard to see how these new entrants can really disrupt the market when the established web building platforms can just copy them. They have a huge amount of training data after all, and plenty of resources to throw at the problem. Wix and Framer have already started to get in on the game, it’s only a matter of time before the others follow suit.
If you’ve ever felt embarrassed at the quality of your presentations, or found yourself struggling through an uninspiring template, Beautiful.ai may be the solution you’re looking for. They specialise in presentation building across a few key verticals; marketing reports, sales collateral, and pitch decks. There’s also a free plan for students.
Sign up for product, growth and GTM development tips for innovators