What happened when the daily Podnews podcast became a video

For one week in October, Liquid Studios turned every edition of the Podnews Daily podcast into a video: a landscape version for YouTube, a vertical cut and chapter markers, all built from the audio and newsletter we already make. It was part of a week-long pilot, working with the team as part of their product announcement. You can see some of the work in the press release.

What was it like to build? And how did it work in terms of the numbers?

Let’s - as the best AI agents would say - “dive in”.

The process

I worked with Liquid Studios with a few visual treatments for how it would work - including fonts and branding. We looked at bespoke animated intros/outros, as well as different ways to handle the individual stories. The Liquid Studios team used a few old shows for this.

Once we had chosen the look/feel of the template and the individual story flow, I then set a few editorial guidelines. Particularly, Podnews is produced in text newsletter form, and in audio, every day - so I rarely have images to go alongside every single story; but I also didn’t want any generative AI to be used - so in this case, Liquid Studios suggested that they would use an image library where I didn’t have specific photographs or graphics. Not got an image of a particular conference or awards ceremony? There are plenty of generic images that convey what’s being talked about, after all.

During the pilot week, Podnews was publishing at 4am Pacific Time - which I thought might be a challenge for Liquid Studios, who are based in Los Angeles. And it got worse - unlike other publications, Podnews is often still being produced and written at 3.30am. There isn’t any time for pre-production, or for finding additional images and text. It gives virtually no time for any production polish. Nonetheless, Liquid Studios said that they were able to turn round the video in less than an hour - so, they assured me, we could have something up in video by 5am. (I raised a quiet eyebrow.)

As a disclosure, Liquid Studios ran this pilot at no cost to Podnews - without any specific ask about any level of coverage. It’s probably worth noting that the Liquid Studios team were always keen to help, proactive with ideas, and more friendly than the team had a right to be, given my less-than-enthusiastic coverage of their previous work.

On a very basic level, we agreed to use a shared folder, which Podnews would put the text and the audio of that day’s newsletter as soon as we had it finished; and where the final video edits would appear. My production process also produces a version in Markdown format - read natively by AI tools; and a WebVTT captions file for the audio, so I made those available as well. And, as I went, I realised that while I didn’t have images for some stories, I did have images for others, and so put those in the shared folder as well.

The first cut was assembled automatically, then checked by a human editor at Liquid Studios before it was sent over to me. And, for all five days, we had the main landscape version of the newsletter produced by 5am Pacific Time as promised, quickly followed by the vertical version and some clips. By day two, Liquid Studios had also produced chapter markers to copy/paste into YouTube as well, spotting an opportunity.

My process was then to publish the video file onto YouTube - replacing the audio version that had gone up an hour before; upload the video to Spotify, add it to LinkedIn and Mastodon (our main social media outlets), and upload it to our own server for newer podcast apps like Pocket Casts and others to also use. (Apple Podcasts uses a proprietary video publishing process, which we don’t have access to).

And then, I’d yawn and go to bed: because 5am Pacific Time is 10pm my time in Brisbane, Australia. No wonder I was keen for Liquid Studios to keep to their promise around publishing time.

The numbers

Here’s how it compared with the five weekdays before (and a few after).

• YouTube views rose 7x over the week. Launch day had a press release behind it, of course, but even leaving that day out, the other four editions averaged almost 4x our normal daily views.
• More of those views were “engaged” views: 82%, up from 69%.
• YouTube’s algorithm noticed. Impressions in the first 24 hours roughly doubled (the median was up 121%). That’s about the clearest sign we get that the platform wants to recommend something. When the pilot ended, daily YouTube views went back to where they had been - so the lift came from the format, not from coincidence.
• People watched a little longer. Average view duration was up about 5% on comparable days. This week, the videos were about a quarter longer than last week though (just the weight of news), so the percentage of each video watched was lower.
• On LinkedIn, the five videos got more views than they did on YouTube. Podnews wouldn’t normally post on LinkedIn, so there’s no baseline to compare against: but it was interesting seeing this amount of views and comments.
• Spotify. Spotify plays were flat, and nor did the total impressions change. Either Spotify doesn’t have an algorithm that cares about video, or this wasn’t enough to change its mind. Podnews - and all industry-focused podcasts - seems to do particularly badly on Spotify anyway.
• Audio downloads. Newsletter and download numbers changes weren’t meaningfully different over time, so they don’t tell us much either way. Some of those “audio downloads” would have been video plays, depending which podcast app you used.

So, did it work?

This pilot certainly got Podnews more visibility. Had we kept at it for longer, we might have achieved more of an algorithmic effect. Some of the additional views and interest would have come from the press surrounding it, especially conversations online; but my suspicion is that it would have increased consumption up to a new baseline.

For me, putting together Podnews, it was not an appreciable additional workload. Indeed, I wrote a bash script that would have copied all the right files into the right shared folder automatically during my publish process, and was only stymied by Google Drive’s slightly odd way of connecting into the filesystem.

My initial scepticism about publish time was unfounded. The product was editorially of a high standard, followed the brief, and was always delivered on time. Like any new thing, there would be elements that I’d change if I did it again: but it looked and felt part of the Podnews brand family, and felt a good accompaniment to the audio. Certainly better than watching a man read into an autocue in a dark office.

So, why aren’t we continuing the pilot, if it worked?

Podnews’s podcast version, which this is, is an unusual product - it’s one that I hope achieves more discoverability of the newsletter, but it’s also something that doesn’t specifically pay its way. In October the podcast achieved 43,622 downloads in the month - not a bad number for industry news, but also not massive, either (the newsletter would have been sent more than 675,000 times by comparison). The podcast’s main job is to get people to subscribe to the newsletter: where they’ll find links to each story, of course, but also see the names of our supporters and advertisers.

But - would I recommend it? Yes, I would. If Podnews was a “normal” podcast, where the rewards for more plays are measured in cash, then it would be an easy choice. Why have a static image on a screen, when you could have good, brand-complementary visual augmentation like this - encouraging trial and converting into more engaged audiences? Especially if it means no additional work at all by the people making the podcast.

This is where AI is ideal - making the most of existing human creativity, not replacing it. If you can’t create full video for your show, then this sort of tool seems the sensible way to go.