When AI meets TV: Ai-Media

Beyond the hype name there's a thesis in there.

I don’t know about you but captions have slowly become a permanent fixture in the way I watch television.

Carmen Berzatto mumbling between screaming at his cousin or any other kitchen staff member in ‘The Bear’ has meant that it’s been a helluva lot easier to follow with the help of on-screen captions. While I’m far from the initial target market and not the primary reason to create captioning, the technology is becoming increasingly present in everyday viewing. 

Ai-Media Technologies Limited (ASX: AIM) (formerly Access Innovation Holdings Limited) was founded with the far more important belief that everyone deserves access to information.

The company’s technology is revolutionising the way people consume content across the globe. Ai-Media was founded by Alex Jones and Tony Abrahams, both passionate about advocating for those with disabilities. Alex, who was born deaf, teamed up with Tony sharing the common goal of ending the exclusion faced by people with disabilities. Captioning technology pioneer Ai-Media was born. 

About the Company

A hype name for the times, Ai-Media is a global provider of captioning technology that assists international brands in captioning their TV broadcasts, live streams, events, virtual meetings and more. Their customers include broadcasters, corporations, government and education providers. 

Ai-Media remains founder-led and was established in 2003. The company listed on the ASX in September 2020. It has a market capitalisation of $80 million at the time of writing and has had a bit of a rough ride since its IPO:

The business started out through the development of a cloud-based captioning, transcription and translations platform which utilised a combination of artificial intelligence (AI), machine learning (ML) and human input to convert speech to text. 

From source to screen its technology provides both live and recorded captioning using hardware, software solutions and human expertise.  

The company made a key acquisition in 2021 of US captioning technology company EEG enterprises. It was a significant purchase with a US $20 million up front payment, a US $5 million earnout and $10 million worth of shares issued at $0.876. This is essentially the business today with the key piece of the acquisition and jewel in the crown the iCap network. The acquisition of EEF allowed Ai-Media to become an end-to-end captioning and translation solution. 

So How Does it Work?

There are three key elements to A-i Media’s tech stack.

Encoder

The encoders can be either physical encoders (Encoder Pro), which plug in with a wire to the audio source from the customer. Or they can be virtual, like the Alta product or cloud-based, like the Falcon product.

Alta is a virtual API-powered IP video encoder, delivering automatic captions for broadcast-quality video. Falcon is a virtual live streaming caption encoder that integrates with the iCap network and iCap cloud for automatic captioning for live streaming events, conferences, and corporate communications. The Encoder hardware is the legacy option for broadcasters and governments. There are a number of legacy broadcasters still using the hardware option while governments can require an on premise solution for security purposes if the cloud is not an option.  

The encoding technology is incredibly important in speech to text, getting the audio into the speech recognition engine determines how accurate the captions are. E.g. A commentary booth is mic’d up with each commentator to avoid outside noise. Those devices connect to the iCap cloud. 

iCap Cloud/Network

Ai-Media’s premier asset is the iCap network.

With over 40 years of intellectual property brought over as part of the EEG acquisition the iCap network and iCap cloud acts as the captions delivery network. It takes the audio provided by the encoder and matches it with the most appropriate Automatch Speech Recognition (ASR) tool whether that be its own Lexi product or a human captioner. iCap then uses its encrypted technology to dig into a clients data to understand the context of what is being shown and creates its own data set dictionary. For example, if you’re captioning an NBA game and the LA Lakers are playing, it will review the team list to better reflect player name spelling and capitalization. Users can also set up a base dictionary with rules. The captioning is then returned to via the network immediately and embedded into the stream. 

iCap enables hourly charging subscriptions and also allows connectivity to thousands of certified third-party caption partners worldwide.

Automatic Speech Recognition (ASR)

The company continues to develop its advanced automatic captioning and transcription solution Lexi. 

Lexi provides highly accurate real-time captioning with minimal latency. It utilises artificial intelligence and machine learning to deliver captions with exceptional speed and precision. The system boasts an impressive 98%+ accuracy rate for broadcast-quality audio, with the ability to identify and label different speakers in a conversation, enhancing the clarity and usefulness of the captions for viewers. 

Lexi provides near human accuracy with its captioning at a fraction of the cost. You can take a look at the number of Lexi products and case studies here.

In short, the system works as follows:

The encoding software/hardware gets the audio from the content into the iCap system. The iCap network and iCap cloud ensures that you get the cleanest possible audio into the ASR system and matches it with the most appropriate speech recognition tool, e.g. Lexi, or a human captioner. The data is then sent back via iCap to the inserter and made available to the end user platform which could be YouTube, Facebook etc. 

Why is it Interesting?

The captioning process screams AI disruption.

And that’s exactly what Ai-Media is doing.

Via the EEG acquisition is disrupting its own original business model. It’s a very different business to the one originally listed on the ASX. With the AI capabilities of Lexi getting to the point where it is as effective as a human captioner for a fraction of the cost, it’s a no brainer for customers to take up this option. 

Previously a human captioner sat as the middleman to repeat what was being said and would train the model to recognise their specific voice print. What has changed now is that middle humans are no longer necessarily needed. There is still a place for human-captioners where there is singing, multi languages or significant background noise but AI is catching up quickly. 

While group revenue has been growing slowly over the last few years, the tech side of the business has been expanding rapidly behind the scenes. 

With a reduced input cost the gross margins for the technology business exceed 80% compared to that of the legacy services business which sits at around 43%. 

One of the core components of the thesis is the strength of the iCap network.

Originally there was no hourly charge for the use of iCap. As a result competitors were getting a free ride utilising the infrastructure. The fees charged allow Ai-Media to reinvest in iCap to sustain its competitive advantage. The strength of that competitive advantage is shown in that the company could maintain competitors on the system while adding a fee rather than being able to develop something on their own. Competitors like Red Bee, 3Play and Verbit have migrated across to iCap charging, highlighting their inability to develop a similar network themselves. 

Having customers on iCap then allows an easy upsell to Ai-Media’s own additional ASR technology like Lexi and its associated products. There are a number of US based broadcasters with the encoding hardware/software installed already which easily transition to Lexi.  

This is despite outside of being the first mover there may not be much of a differential between Lexi and competing ASR products. Tech giants like Google and Microsoft have their own ASR products in what is a highly competitive market. A product like Microsoft Teams essentially has its own network as it has access to your email, staff names, calendar and can quickly develop its own data dictionary. While a separate target market at this stage it is something that could quickly become a significant competitor if Microsoft expanded into that space. 

So won’t AI, Google, ChatGPT just ruin any advantage Ai-Media has?

iCap has multiple ASR engines that sit on top of the network to enhance the product. The more improvements in technology like ChatGPT actually improve the outcomes of iCap and Lexi. The more development in tech Goolgle or ChatGPT only improves the end product. 

So while the declining revenue in the services space is being destroyed internally by the switch to tech, competitors are losing out significantly too, without the benefit of the iCap network. 

Ai-Media’s strategy has been to expand its dominant iCap position in US broadcast and take it global. This will be an interesting space to watch. There is significant potential to build the network out globally while upselling its Lexi services. 

As mentioned Ai-Meda is founder led with Tony holding around 35 million shares following a recent on market purchase. The founder picked up an additional 5 million of shares with an acquisition price of 31 cents. Clearly he is seeing some value at the current price! The below table is prior to that acquisition with some key backers being recently retired board member Deanne Weir and the acquired business EEG. 

Financial Position

In its most recent H1 results, Ai-Media grew revenue by 10% on the prior period.

The move to high margin technology revenue is hidden here by the decline in legacy services revenue. Technology revenue grew by 38% while services revenue declined by 7%. The expectation is services revenue will continue to decline as the transition to the tech business continues. Customers like Channel 7 and Channel 9 are beginning their transition to be using LEXI and currently make up around 15% of the services revenue. At the end of the first half gross margins had improved from 60% to 63% with almost two thirds of the gross profit generated by the tech business despite representing less than 50% of overall revenue. 

Costs remained controlled and the total loss before tax reduced from $1.4 million to ~$500k. 

Operating cash flow positive for the half which was largely driven by a reduction in receivables and backing out depreciation and amortisation. The company continues to invest in itself which means increasing the number of engineers and investing in product enhancement. Fortunately the capitalisation of intangible costs has been reducing so there doesn’t appear to be any P&L engineering going on. To quote CFO John Bird:

“I'm a bit of an old-school finance guy, unless there's an absolute correlation, I'd rather not capitalise it. I'd rather take it through the books. So we're pretty conservative in that space.”

Ai-Media has a strong cash position with $11.6 million in cash at the end of December 2023. 

Intangibles make up more than half of the total assets which is something to keep an eye on. This is predominantly goodwill as part of the EEG acquisition. Inventory ticked up in the half as a result of investment in additional Lexi products like the disaster recovery box. Outside of lease liabilities the company had no debt at the end of the first half. 

While I believe the transition to tech is very interesting the company still has plenty of work to do.

I’ll be keeping a close eye on the continued improvement in revenue mix and the global growth of iCap. I believe if the transition can continue with overheads kept in check it is hopefully not too far away from a key inflection point. 

A shout out to friend of the stack (is that a thing?) Luke Winchester for putting this one on my radar.

I’ve taken a small research position in Ai-Media and I will look to add more soon. This is far from financial advice so please do your own research. If you’re still here, thanks for reading, I appreciate it! Would love any feedback particularly on this format or any additional thoughts/critiques.