Machine Intelligence Garage shines a spotlight on the creative industries.Posted 6 May 2021
Machine Intelligence Garage, Digital Catapult’s flagship AI accelerator programme, hosted its creative cohort demo day: Artificial Intelligence for Visual Arts and Media. During the event, 11 startups presented their market-ready solutions to potential investors, partners and companies looking to apply AI and ML technologies in the creative industries.
Machine Intelligence Garage, is an acceleration programme that addresses the challenges today’s artificial intelligence (AI) and machine learning (ML) startups face.
The programme runs regular calls throughout the year, including themed cohorts – this time centred on the creative industries. The UK stands at the forefront of the creative and digital industries, which both contribute over £100 billion to the economy. AI and ML technology can transform the creative industries, accelerate new possibilities and push the boundaries for visual arts and media. In this blog post we review the event, its outcomes and learnings from the market-ready solutions and panel discussions.
Creating a framework for responsible AI.
As host, Hugo Defaux, Innovation Delivery Manager, Digital Catapult and manager of the Machine Intelligence Garage programme, led the event. After a brief introduction to the programme Hugo was joined by Nathan Coulson, Machine Learning & AI Technologist, Digital Catapult who outlined how Digital Catapult has been developing its practice in applied ethics to ensure responsible AI remains at the core of any business, from startup stage to corporate level.
Keynote Speaker Myrna McGregor, BBC Lead, Responsible AI/Machine Learning reinforced the importance of responsible AI and how, as a public service broadcaster, the BBC is adapting its policies for internal teams and external partners.
The BBC has devised the Machine Learning Engine Principles – a set of high-level principles to guide the adoption and application of AI and ML technology, a framework similar to Digital Catapult’s Ethical Framework, to ensure that the BBC can create robust and responsible AI products and services.
“We understand that AI and machine learning is going to transform the way that we work and interact with our users and our audiences. So we want to make sure, as the BBC, we do this in a positive way, creating public and societal value that we feel confident about.”
Myrna McGregor, BBC Lead, Responsible AI/Machine Learning.
Innovative solutions that extend the capabilities of AI and ML in the creative industries.
From 3D models to digital humans and immersive experiences, 11 cohort teams presented their creative AI prototypes. The teams also transformed many costly and labour-intensive processes for the industry, including data management and search, motion capture, video editing and rights management:
- Humain – Humain’s mission is to make digital human creation easier, saving visual effects teams thousands of hours creating authentic on-screen characters. Humain demonstrated how machine learning technology could build a human likeness in just a few seconds, recognising the subtle differences and observable facial movements that identify us as individuals.
- Kaedim – Although the demand for digital 3D assets grows every year, they are expensive and labour-intensive to create, requiring highly-skilled artists. Kaedim demonstrated how its AI-powered software can produce 3D meshes from 2D input images in 1/100th the time it would take a human designer, generating multiple 3D models from a single input image.
- Deep Reel – The growing popularity of streaming services has also created a market for foreign-language content. While subtitles or dubbing can cross many language barriers, they often distract the audience. Deep Reel demonstrated how it uses AI to localise content and achieve flawless synchronisation of the on-screen actor’s lip movements to match a new language.
- Breathe Music – Breathe Music has created a solution that can recognise every manifestation of a song or composition, live, online or recorded – protecting the rights of music creators, publishers and investors in song catalogues. During the demonstration, the Breathe Music web app connected two contrasting versions of David Bowie’s Space Odyssey to the original recording.
- MetaLynx – Working with image libraries at scale can be costly and often prone to error or bias. MetaLynx demonstrated how machine learning could be employed to visually explore and intuitively understand large image databases, using active learning to select the right data, reduce labelling costs and identify bias.
- EnVsion – EnVsion has reimagined how videos can be created and edited at scale. When working with video, finding a specific segment, a spoken word or an actor’s appearance on screen can be very time-consuming. As demonstrated on its visual platform, EnVsion leverages computer vision and text to speech (TTS) to recognise and index any person appearing on screen, find spoken words or text and can organise scenes across an entire video library.
- FeedForward – Figaro (by FeedForward) is a B2B search and metadata engine for music catalogues. Powered by machine learning, Figaro enables music catalogues to manage their descriptive metadata and learns the relationship between the words used to describe music. As FeedForward explained, music resonates at an emotional level that’s difficult to put into words, so its intuitive music search enables users to search based upon elements of the music that are important to them.
- YourTour – YourTour creates AI-powered ultra-immersive, 360-degree video and interactive 3D environments. A demonstration using intuitive gesture detection and head tracking illustrated how immersive visual media can reach wider audiences through 360 degree videos on TV, tablets and other devices, without relying on an Oculus VR headset or 3D glasses.
- Zeta Motion – Professional motion capture solutions are expensive to install and require highly-skilled resources. Zeta Motion has transformed motion capture and 6D tracking with a cost-effective AI-based solution. Demonstrating a prototype motion capture glove, hand movements were accurately displayed in real time within a three dimensional space, while interacting with physical objects.
- AWAYTOMARS – A company set to revolutionise the fashion industry with its AI software. Through its website, AWAYTOMARS enables the co-creation of fashion collections, where anyone can get involved, from students to artists and professional designers. Community members can submit their designs to receive peer-to-peer or professional feedback and collaborate with some of the world’s leading fashion houses.
- Charisma.ai – Charisma is an AI-driven toolkit for creating interactive stories with believable virtual characters. During the demonstration, the user interacted with an onscreen character in natural conversation to change the narrative. Charisma can fill the gap between television episodes or series to extend audience engagement and deepen the relationship with the characters, while generating marketing and revenue opportunities.
Key barriers for AI adoption in the creative industries.
To conclude the event, Max Cleary, Partnerships Lead at Digital Catapult, led a panel discussion on the key barriers for AI adoption in the creative industries and how to bridge the gap between startup innovation and the real needs of the media industry, attended by:
- Jamie Allan: NVIDIA
- Kadine James: RYOT Studio
- Rowan de Pomerai: DPP
- Amandine Flachs: WildMeta
Key points from the discussion included:
- There are lots of barriers for startups hoping to work with large businesses. The broadcast industry is very focused on getting a programme out the door and people don’t have enough time to discover what’s coming next.
- In the past, some AI companies over-promised and sold solutions that have not lived up to the hype. Content technology providers and content companies need closer partnerships to ensure solutions match the business need.
- There are challenges around data rights. If models get trained using a content provider’s content, does the technology company or the provider own the IP?
- Startups need to validate their ideas, but need to be aware that their own explanations may not necessarily resonate with the industry – when AI and machine learning meet industry jargon, the two don’t necessarily match.
- Sharing your IP can help evolve your business. You might find someone that wants to take your work in a new direction to commercialise it.
- Immersive is going to transform media, with XR content taking centre stage.
- 5G will enhance storytelling and provide consumers with a new way to interact with content, creating opportunities for advertisers and publishers using the latest augmented and virtual reality.