Six months ago I published a post on VC Cafe titled, the state of creative automation, diving into vertical applications of AI on content creation of all sorts; from text (GPT-3 and the likes) to Synthetic video (powered by companies like HourOne).
Last week’s announcement of Dall-E 2, the Open AI initiative that creates realistic images and art from a description in natural language, is yet another example of how fast this space is evolving. Compared to its predecessor, Dall-E 1, released about a year ago, Dall-E 2 is capable of: 1) generating more realistic images with 4 times higher resolution, 2) make realistic changes and variations of an existing image and 3) change the stye of a picture (i.e. Pixel art) using natural language.
With great power comes great responsibility. The media mostly chooses to cover the negative implications of these AI technologies: be it fake news or personalised targeting (like in the case of Cambridge Analytica during the Us elections). These risks are of course legitimate, as is the case with any technology: You can use AI to discover new cures to obscure diseases or to create autonomous weapons. OpenAI is of course aware of it, and therefore the Dall-E 2 project is not yet publicly available. It’s up to the companies working in this space to adhere to a code of conduct. But put the ethics aside, assuming you leverage these new technologies for business purposes, what use cases do they unlock? What new markets can be created?
There’s a new wave billion dollar of AI startups coming, and in this post, I will outline some of the positive real-world applications of these creative automation technologies.
Advertising automation
Perhaps the biggest use case to date, advertising is the first target market for several of the creative automation technologies. Within advertising the use cases and mediums are varied:
- Text – companies like Copy.ai, Anyword, Copysmith, Assembly or Jasper, NLP and generative text algorithms are used to create ads or social copy.
- Images – companies like Rosebud AI, Persado offer advertisers images on demand for their campaigns. DataGrid specifically focuses on AI generated whole body models.
- Voice – synthetic voice algorithms are already powering audio ads on podcasts, using text-to-voice.
For example, imagine you are looking for creative for your next advertising campaign. You’d have to sift through image banks, hire an agency, spend a day photo shooting etc.
Or alternatively, take a look at the gallery below. What do these images have in common?
All the subjects in the images (the sky, an eye, a shoe, a cat and a person) don’t really exist – they were created with AI. You can test these out yourself at ThisXDoesNotExist.
Social content
Spend a few minutes on TikTok, and you’ll be amazed at the volume of content that is already ‘machine-generated’.
- AI generated text – companies like Israel’s AI 21 Labs offer free NLP generative text engines that craft blog posts or social posts taking the input of a single sentence.
- Synthetic voice – text to voice tech enables users to mask heavy accents or maintain anonymity
- Synthetic influencers – Instagram stars like Lil Miquela develop a following and are used for brand partnerships, commerce, music deals, etc. AI Rozy, a Korean virtual influencer that signed over 100 corporate sponsorship deals is a great example.
- Interactive greetings – many people use Cameo, the influencer marketplace, to send personalised greetings of D List celebrities to their friends. AI will essentially enable a synthetic cameo, using both real people (who allow the platforms to sell their likeness) or animated characters and Metahumans to enable people to send each other interactive greetings powered by AI.
Learning and Development
- Synthetic Video – AI generated video companies like Hourone (disclosure: we are investors via Remagine Ventures) completely removes the need for a studio or expensive production to create high quality video. Users can simply type text to create a synthetic video. Other companies in this space include. Reprahse or Synthesia.
- Synthetic Audio – a lot of people prefer audiobooks to real books for their convenience, as they are able to ‘read’ a book while walking or driving. But producing an audiobook can be timely and costly, often using the voice of the author and painstakingly reading and recording each page, sometimes requiring multiple takes and heavy editing. Text to speech APIs like Google’s, Speechify, Afloritmic and others can bring us Synthetic Audible, or Aduible for documents.
- Dubbing/ Voice – AI startups like Deepdub are helping internationalise content with dubbing. The startup takes the original voice sample of the actor (Think Samuel L. Jackson in Pulp Fiction) and adapts the audio to the new target language and accent (i.e. Spanish, Argentinean accent).
- AI music – The Jukebox neural net, another OpenAI application, can generate new songs in the style of a chosen artist. It’s one of many startups operating in the space of AI generated music, which include: Mubert (generate a sountrack for videos by genre), Amper, Memu, etc. There’s already even an AI generated song contest!
Customer support
- Video chatbots – next time you call your bank or your airline, rather than stay on hold for an agent you may be able to interact with a virtual chatbot or receptionist. A combination of NLP and synthetic video technology makes it possible already today
Virtual goods
- AI generated NFTs – NFT sales hit $17.6 billion in 2021, and despite a recent slow down in the NFT market, earlier this week, the Moonbirds NFT collection (by the likes of Beeple and Kevin Rose) sold 10,000 images of digital owls for $281 million in two days. Text to image AI like Dall-E 2 can algorithmically create variations of an image, enabling anyone to launch a unique NFT collection, be it art or digital collectibles/fashion/game items. Think about AI generated bored ape yacht club, or a synthetic Beeple.
- Democratising virtual goods commerce – Meta recently announced that users will be able to sell virtual goods like virtual fashion or avatar accessories in their Horizon virtual world. AI generated 3D models can essentially democratise the creation of virtual goods/NFTs from designers to freelancers, to anyone.
- Voice to 3D – Meta recently unveiled its ‘builder bot’ AI concept, in which users will be able to create virtual worlds and elements within a virtual environment, by voice commands (i.e. ‘add clouds’). Anything World, a Uk based startup is also working in a similar space.
- Avatar technology – Ready Player Me, and Genies (which raised $150M in funding earlier in April, reaching $1 billion valuation) are two examples of startups working in the avatar space. The use cases today focus mainly around online digital identity outside of games, but the vision for these companies is to be the digital representation of the user in the future metaverseses and virtual worlds. There’s also of course Metahuman by Unreal Engine/Epic.
- Synthetic voice in gaming – Sonatic removes the need for expensive voice actors by offering a synthetic voice service that is both hyper realistic and captures breath and emotion. Following a fight with throat cancer, actor Val Kilmer lost his voice. Sonatic was able to recreate Val’s voice using old recording samples.
This list of use cases is just scratching the surface. Founders are taking advantage of these new technologies to solve problems in niches that might one day become big market. We remain excited about the potential of tools in the creative automation space at Remagine Ventures. If you’re a founder in the early stages (pre-seed and seed) we’d love to talk to you. We can be reached at info@remagineventures.com.
I’m a former general partner at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google’s first physical hub for startups.
I’m also the founder of Techbikers, a non-profit bringing together the startup ecosystem on cycling challenges in support of Room to Read. Since inception in 2012 we’ve built 11 schools and 50 libraries in the developing world.