Problem: The world of podcasts is changing. However, podcasts are virtually the same since they first became a medium of “audio-blogging'“ in the 1980s. Here are a few interesting statistics:
As of July 2020, there were more non-English podcasts than English podcasts. Non-English podcasts are growing fast, according to ART19's Chadd Hollowed, posting a set of graphs on Twitter. In April, May and June, there were more non-English podcasts launched than English ones. “This overturns a lot of assumptions about industry trends,” he says. (source)
People are more comfortable with content that is in their native language. 72.1% of people spend most or all of their time on websites in their own language. Meanwhile, 72.4% of consumers would be more willing to buy a product with information in their own language. (source)
There are now an estimated 660,000 podcasts in production (that’s a real number, not some comically inflated figure I invented to communicate “a lot”), offering up roughly 28 million individual episodes for your listening enjoyment (source)
And three final stats: Only 25% of internet users are speaking English; there are more than 500,000 active podcasts on the Apple Podcasts app alone; and the percentage of people in the U.S. who reported listening to a podcast in the last month is 26% (source)
Solution: This company would use the large database of podcasts to solve the infamously difficult problem of automated podcast translation through AI. It would position itself as a way to translate podcasts into entirely new languages to expand the audiences of podcasts by factors of magnitude. Thank you Mariana Briones and Diego Rejtman for inspiring this blog post today!
There are two main skews of podcast translation today: white-glove services (i.e. hire someone to translate your podcasts for you in UN-style) or automated translation (much less common and currently used by large tech companies). I’ll analyze a few of the major players and pros and cos in each space.
One example white-glove service is LenseUp. They, along with other prospective translation agencies, usually have 3 steps in the translation process.
The first step is audio transcription, where the transcriptionist manually transcribes audio to text (or software is used to put what is said into script form). Transcribing the audio well is key because this is the text that will be translated, localized, and edited.
The actual translation and localization of the text then takes place using the script. The language is translated line by line, and localization ensures the tone and message fit into the new culture.
A highly professional audio translation service will also have proofreaders to review the text in the new language.
The podcast is re-recorded in the new language with human voice (as LenseUp describes, “our actors are native, with sample voices available on request”) or synthetic voice: (as LenseUp describes, “new twchnologies [sic] and artificial intelligence allows us today to offer this service in different languages”)
The innovation in this space would be to complete steps 1-4 in milliseconds or seconds rather than weeks or months (which is the current timeline). At scale, this would equate to instant podcast translation.
Google Podcasts has been at the forefront of pioneering instant translation.
According to Google Podcasts Product Manager Zack Reneau-Wedeen, in the future, Google will have the ability to “transcribe the podcast and use that to understand more details about the podcast, including when they are discussing different topics in the episode.
“It’s important to say that this technology is still improving, and some of our vision here is probably a little more long-term than what we’ve talked about so far. Still, it’s an exciting motivator for us to try to make these experiences possible.”
Their long-term vision?
Google’s AI “listens” to every podcast published, converts all spoken word content into timestamped, searchable text, and indexes the contents of every episode. All the content of all episodes of all podcasts become searchable, sort of like a text article. And not just the entire episode: by analyzing podcast transcripts and/or publisher-created chapter markers, Google could begin to understand specific segments or topics within episodes. In the future, Google Search and Google Assistant could allow listeners to go beyond finding the right episode of a podcast. It could help them jump straight to the right section that is of interest to them.
Sometimes the best ideas aren’t entirely new, they’re just a better implementation of visions already out there. This business would be competing directly with the Google Podcast team and would serve as a replacement for existing white-glove services.
The market is huge, as reported by BusinessWire: “The Global Podcasting Market size is expected to reach $41.8 billion by 2026, rising at a market growth of 24.6% CAGR during the forecast period.” Perhaps this is why Spotify shelled out $100 million for Joe Rogan’s podcast and signed an exclusive deal with Michelle Obama.
Monetization: SaaS: selling access to these translation services.
Contributed by: Michael Bervell (Billion Dollar Startup Ideas)