Table of Contents Hide
Stats Perform and Veritone last week announced a new product, Opta Voice, that employs AI solutions to automate the creation of “real-time play-by-play, pre-game, in-game and post-game updates.” The solution promises to automate content generation for sports apps ranging from live game updates to social media and podcast content.
In one example, the solution leverages Stats Perform’s sporting event data and natural language generation to provide an audio track of the ongoing in-game events. It then layers on the Veritone Voice synthetic speech engine to make that play-by-play sound like a human announcer.
New Synthetic Use Cases
Opta Voice is introducing a number of new synthetic media use cases. The real-time game cast is the splashiest. However, the recorded post-game and pre-game use cases are likely to become the most popular options.
More content typically translates into more engagement and more users, so publishers and app makers are incentivized to increase their production. Given that Opta Voice can draw directly from Stats Perform data, there are a lot of content formats that it can support. The company suggests social media posts, additional podcast episodes, and radio spots.
Access to the data and its automated conversion to various content formats should enable broader coverage as well as more frequent updates. Humans can only write so many social media posts and voice over a limited number of radio spots before they have to hire more of them. AI doesn’t have the same limitations. In the synthetic media space, AI enables hyper-automation, which means far higher production per hour.
Natural language generation (NLG) or AI-based content generation is not new. The Associated Press began using AI to write stories about quarterly earnings reports in 2015. The year prior, the company published an average of 300 stories about earnings releases of public companies per quarter. That average rose to 3,000 in 2015. By the spring of 2018, the figure was 4,700 per quarter. The AP even says these stories are more accurate than those previously written by humans.
That same technology was used to write over 5,000 NCAA basketball game previews in 2018. Automated Insights, the company that provided the NLG technology, was acquired in 2015 by Vista Equity Partners, which is also the parent of Stats Perform.
This content could just be published for reading as with the AP use case. However, adding audio increases where and when users can access the content and makes it more engaging. That’s where Veritone Voice comes in.
Cory Hill, director of product management at Veritone, said to Voicebot in an interview, “Veritone Voice is a set of APIs that enables you to either take text or speech and generate synthetic audio or voice from that content. So, imagine you are in a pipeline, and you have sports facts coming in saying, “player A from the left wing is on the left wing and is passing to player B.” We can take that content and, through some NLG, produce some really nice commentary around that without any human in the process.”
The content generated by Stats Perform is rendered as synthetic speech by Veritone Voice, so it sounds like a sports personality delivering the information. Hill added, “We can customize that experience to be the personality that you want to comment on your soccer match. You can also generate customized reports so maybe [you have a] fantasy team and you want a read out of those players…we can do all of those things and add commentary and sound effects in the pipeline as well.”
Patrick Fischer, Chief Business Officer at OneFootball commented, “We believe content that drives engagement and adds to our vertically integrated product will give our fans a reason to return to our app many times per day. Realistic and scalable audio content can play an important role in this strategy.”
Not Just Content – Monetization Too
“There are also dynamic ads that can be read in a different voice,” commented Hill. The product pitch is mostly around content generation. However, Veritone can also support ad monetization for the product through ad revenue and NFT sales.
Expanding Veritone Voice
Veritone Voice gained a boost earlier this year with the acquisition of VocaliD. Founded in 2014 by Northeastern University professor Rupal Patel, the company is known for high-quality synthetic voice engines, voice clones, and voice prostheses. This technology has been added to Veritone’s existing solution known as Marvel.ai. Voicebot’s Eric Swartz wrote at the time:
Buying VocaliD should boost Veritone’s capability as a synthetic speech provider. In particular, VocaliD’s vocal adjustment and audio mixing software will allow Veritone to more easily manage and scale its custom voices regardless of where they are deployed.
This solution could have an even bigger impact on regional and second-tier sports coverage. Sports fans have a voracious appetite for information, and outside of the major sports, there is only limited coverage. Opta Voice can fill this gap as well as supplement the coverage of major sports where there remain many more stories than working journalists can cover.
Follow @bretkinsella Follow @voicebotai
Synthetic Steve Jobs Interviewed by Deepfake Joe Rogan for AI-Powered Podcast Scripted by GPT-3
Veritone Acquiress Synthetic Speech Startup VocaliD
The 6 Segments of Synthetic Media – Video
iHeartMedia Picks Synthetic Voice Developer Veritone to Translate Podcasts
Read the full article here