Hinter den Kulissen von Think Different! Think AI!

Transcript

00:00:00Welcome to Think Different, Think AI, the podcast by Mark and Jens.

00:00:07Two tech-loving minds who not only talk about artificial intelligence but live it.

00:00:14Here you will find clear classifications, real practical insights, and a fresh look at what is possible.

00:00:20Understandable, critical, and always with a wink.

00:00:24AI to ponder, to chuckle at, and above all, to discuss.

00:00:33Welcome to a new episode of Think Different, Think AI.

00:00:38Well, what can I say? A while ago, we talked with our guests.

00:00:44Okay, we only had one guest; that needs to be corrected here.

00:00:47With René about the whole topic of how AI affects the leadership of employees,

00:00:53its impact on organizational interaction, and Jens and I also had an episode where we talked about

00:01:00whether AI has a claim to vacation. Now, you have probably noticed that I am speaking

00:01:07here alone. I can't even pass the word to Jens because even though Jens is a

00:01:12real person, he is on a real vacation, and we thought about what we could bring you in a

00:01:18further episode so that the waiting time until Jens is back doesn't

00:01:25feel too long. And we thought I would tell you a little bit about

00:01:29how this podcast comes into being. A bit of the technical background. Let's

00:01:38start with, for example, the cover. Jens is sitting down with Manus and making us one for ]} app and cod...

00:01:45each episode has an uplifting cover and basically gives Manu the instruction to adjust the respective cover accordingly

00:01:52each time, so for example the symbol with the tie for leadership

00:01:59partially with AI, and I with the corresponding pixel graphics from earlier, I feel

00:02:06always reminded of Monkey Island, generated with the corresponding episode numbers

00:02:13in the overview. We record ourselves using riverside.fm. It is a not very

00:02:20established podcast tool that allows you to set up a virtual studio.

00:02:28In this virtual studio, you can invite guests who can then join via app or

00:02:32browser. You can designate who has moderation rights,

00:02:37who is a guest, who is just a listener, and you receive a very intuitive,

00:02:42I would say, control panel to see if, for example, the microphones

00:02:48are well aligned. If you want to make a video podcast, you also get support here.

00:02:53What is also extremely advantageous about Riverside is the possibility of editing. Classic

00:02:59audio editing is really about having the waveform and then trying to cut in this waveform

00:03:04to remove the appropriate pauses, eliminate slips of the tongue. That can be more

00:03:11or less complicated. At least for people who don't want to deal with

00:03:15waveforms. And this is where Riverside is quite

00:03:20nice, because Riverside has the ability to say, I go to

00:03:24you, I get the complete spoken text per speaker

00:03:29transcribed, I can highlight it, and I can

00:03:32remove it. And just like that, a sentence might lose a slip of the tongue. I can, while

00:03:40I'm holding a monologue here, and if someone interrupts me or if I interrupt

00:03:45someone, I can also mute other audio tracks while someone speaks.

00:03:51And if you completely trip over your words and maybe miss a sentence,

00:03:57to make the conversation a bit more fluid again, I can not only delete sentences,

00:04:03insert silence, but I can also generate corresponding words, sentences,

00:04:12entire sections from Riverside using AI Voice Cloning. Once the recording is cut and basically in

00:04:18the box, I have the option to export each audio track individually or as a whole

00:04:24audio track. And at this point, we still don't have what we can quasi

00:04:30provide you as a podcast, because even if Riverside allows more to

00:04:35edit audio tracks and add texts and already has a small voice enhancement built-in

00:04:43and microphone control at the recording stage, the audio quality

00:04:49is sometimes still not so optimal, that's something we have to say. Furthermore, you can sometimes hear the guests still

00:04:55breathing a bit, you sometimes hear filler words, and to reduce that a bit,

00:05:01and to make the voices a bit clearer, the pauses that may have formed,

00:05:07become a bit shorter and the breathing sounds disappear.

00:05:11I then upload everything to auphonic.com. Auphonic.com offers to edit

00:05:18two hours of audio material in the free account. You can also send more material in exchange for a handful of coins,

00:05:23as we do. You can upload audio and video files.

00:05:30And it feels like a miracle machine when it comes to audio quality. Many of you may have

00:05:35heard of the AI tool Adobe Podcaster. Adobe Podcaster has the

00:05:42charm or the challenge that I'm not really

00:05:47convinced by the quality. Adobe Podcaster somehow makes things for me, well, it's trying

00:05:53to isolate the voice, but especially the breathing sounds, the

00:05:58unnecessary pauses, Adobe Podcaster struggles with that because it doesn't really

00:06:02do that at all. On Honig, you can choose from various presets, you can also

00:06:10replace with custom configurations, and it has really become established to say, delete

00:06:15unnecessary pauses, make the voices clearer, level everything up, and edit the

00:06:23audio file, and this can be done for both the entire audio file as well as for the

00:06:29individual audio tracks, in order to assemble them later. Most of the time I do this

00:06:33for the entire audio file when Jens and I are alone. In the episode with René I have

00:06:39slightly adjusted the workflow to achieve the best possible result for everyone involved.

00:06:43to get. I went over and processed each audio track, so the one from René, the one from Jens

00:06:48and the one from me individually through Phonic, but not the one I

00:06:55had let pause removal happen, because otherwise the audio track wouldn't be compatible anymore,

00:07:01but I just did the voice enhancement and then took a step I usually don't take,

00:07:06namely transferring everything into an application again,

00:07:10it's called Farad. Farad runs on iOS and you can align audio tracks together.

00:07:16So I first went through all the audio tracks individually, also pushed them forward, then connected them again via

00:07:22Farad and then ran the result through again, but this time with the note to keep everything that's audio as it is.

00:07:31But please delete all unnecessary pauses.

00:07:35Yes, and when the result is finished, it is often referred to as Podigy. Podigy is a distribution platform that accepts podcasts, meaning it accepts audio files and

00:07:46distributes them. Distributing means then, for example, sending them out to you on Spotify, Amazon Music,

00:07:52Apple Podcasts, YouTube. This means you upload the file there. There is an app or

00:07:58a web interface. You can, as mentioned, make it available as a new episode. You

00:08:04maintain a title, you maintain a subtitle, you maintain show notes, upload the image,

00:08:09that I mentioned earlier, that Jens Breit provides. And then you can check the box,

00:08:14that PolyG not only takes care of the distribution but also provides a transcription.

00:08:21That's the reason that you can also find it in any player or even in the web interface as well.

00:08:26you have a unified transcription of the podcast, where you can then search for like, where

00:08:33a certain word was mentioned, and copy texts.

00:08:36If everything is then appropriately included and we have guests, the outcome

00:08:43will be provided to the guest again for approval. Jens and I no longer require

00:08:48mutual approval. This means, if it’s just the two of us as guests, the whole thing

00:08:52then goes online. Going online means that Empodigy presses publish. Man

00:08:57could also have entered a date for when the podcast should be published. That would have

00:09:02made sense, Jens, if we had already fully recorded an episode, we could have just

00:09:06released it now. But okay. And then the episode goes online.

00:09:10Online does not yet mean that people automatically know it exists.

00:09:16Loyal listeners of our podcast, if you like it, please let your friends know.

00:09:20We gladly gather more subscribers, even though this is provided in their podcast player,

00:09:25but there are still many people who don't even know we have a podcast.

00:09:30At this point, we have various strategies on how to increase our reach.

00:09:36That is, both Jens and I create ads for the podcast at different times on our LinkedIn profiles.

00:09:43So, what happened in the episode? Sometimes we also write

00:09:47on LinkedIn in articles and link to the podcast, that's the classic approach. What goes a bit further

00:09:52is that we, scientifically, I also use Manus, not for the

00:09:59cover image, but for advertising. I say, Manus, go to the podcast's website,

00:10:04and read the transcript that is available on the website.

00:10:08We remember, Podgy provides that, identifies the important keywords, searches

00:10:13for those keywords on social media and LinkedIn, and composes comment posts,

00:10:20to give appropriate feedback on the articles he found with the keywords,

00:10:24to refer to our podcast.

00:10:28That means, for example, when we discuss whether AI usage is justified or not, we look

00:10:35at where AI is mentioned as a colleague in a podcast, I wanted to say,

00:10:41in a post.

00:10:43As mentioned, this is searched for by AI and then a thoughtful response text is written

00:10:49and left as a comment with the link to the podcast.

00:10:52It's going quite well.

00:10:54So we are actually getting a few listeners from this, which the statistics

00:10:58indicate.

00:10:59Unless there are just hens, I have to admit, hens are a bit of the download dip,

00:11:05as for the PolyG download numbers.

00:11:09Speaking of PolyG download numbers, PolyG not only does distribution, but also statistics,

00:11:16That means, for example, if I look at Zink Different, Zink AI, it has

00:11:23in fact, I'm just checking since the release of Boss Level AI, over 100

00:11:30downloads.

00:11:31The only podcast that performed even better, with over

00:11:35150 downloads, was AI as a Customer.

00:11:37We are quite happy with our downloads, yes, we still need about 100 downloads,

00:11:42then we will hit the 750 mark for ourselves.

00:11:45It's slowly growing. If I compare that with my other podcasts, I have

00:11:51other podcasts where I don't speak in person. For example, there's 404 Laughs

00:11:57not found. That's a podcast that is completely produced by Manus once a week, on Tuesdays.

00:12:03created on Wednesdays and put online, and I only hear about it through the generation

00:12:09of Manus, what they have come up with. More specifically, Manus generates

00:12:13the text, and Lesman Leps makes the conversation between an IT manager and his experience.

00:12:21And it's pretty funny at times when you listen to it, because nothing is based on

00:12:28real stories; instead, he researches various things on Reddit and other

00:12:33sources and tries to build a funny, non-repetitive story from that.

00:12:39Kopf und KI is for me a bit of an attempt to learn more about AI.

00:12:44I told Manos to please create an episode every week and publish it so I can get a bit better

00:12:52at understanding the topic of AI; it's quite amusing.

00:12:56Schattenakte, the case of the week, I once said, hey, let's see.

00:13:00Manos, if you can manage not just a comedy story like

00:13:064.04 with different speakers and help from 11 Labs, but try

00:13:11making a story that's mainly narrated by a commissioner about fictional

00:13:19cases. It was important to me with the prompting adept that it is fictional, so that no one

00:13:24gets fooled into saying, wait a minute, I know that case. Recently, I had

00:13:30a little conversation with the company that the topic of prompting is also something

00:13:33people should perhaps engage with more; so I created Prompt Intelligence

00:13:37as a podcast. I am also starting to notice the limitations of the old prompts,

00:13:42because the old prompts have problems with foreign languages. So if I let a podcast

00:13:49be narrated by 11 Labs and switch between German and English

00:13:53words, it gets mixed up. From that perspective, I need to change Kopf und KI and

00:13:58Prompt Intelligence eventually, when I have time, so that the prompts are somewhat

00:14:02change things so that he gets more guidance on how to pronounce English terms or not.

00:14:08Sync Different Sync AI, as you can hear, is completely manually generated. So I have

00:14:13never used the topic of voice generation in Sync Different Sync AI, except for

00:14:20our title. You have the intro and the text in Sync Different Sync AI. By now, you can

00:14:29actually make music with Elevenlabs. Back then, we still used Suno and Music-GPT

00:14:35and also let Elevenlabs record the voices for the intro. But that is

00:14:41also the only thing we generate with AI for this podcast. And another

00:14:46podcast, which has very few listeners, but which I personally really enjoy, is

00:14:52AI Revolution. AI Revolution is a purely English-language format. Anyone who subscribes sees a glasses frame similar to what Steve Jobs used to wear, but in the lenses reflect AI symbols, and it is also 100 percent scripted. With the note, just look at what happened in the last few days if you piece together enough for a public podcast and use something from the sentence structure that I previously explained very thoroughly.

00:15:22to train it to try to formulate sentences like Steve might

00:15:29have presented them on stage as well.

00:15:31That's also a very funny experiment and, I think, a very nice experiment,

00:15:37that shows you that AI, just through system prompting, is already capable of generating very engaging,

00:15:45what I think are, texts.

00:15:47And I actually listen to those with absolute joy every time.

00:15:51And if I compare, what are some of my favorite podcasts, apart from Zinkdrift

00:15:57and Think AI, because I truly appreciate the conversations with Jens, it is every time a

00:16:00enrichment, what we openly discuss and the topics we have, I enjoy on one hand

00:16:06the comedian podcast 404 and on the other hand AI revolution.

00:16:10Yes, and that's basically our setup.

00:16:13And with that, we hopefully gave something to think about for one or the other.

00:16:19I would be happy if one or the other also checks whether perhaps one of the

00:16:24automatically generated podcasts is also something nice.

00:16:26I would be very interested in what stands out to you particularly positively, maybe even negatively

00:16:30because I really want to continue experimenting with this podcast.

00:16:35I hope we can record another episode soon when Jens is back.

00:16:40I must honestly admit in preparing this episode, but I forgot to ask us,

00:16:43when he will be back. But I believe it can't be much longer. And therefore

00:16:48I hope that you enjoyed this little special, this little special edition and

00:16:54that you stay with us. Until the next episode, when it’s time to think different,

00:16:59think AI. And as I said, if you like it, please share it.

00:17:03Welcome to ThinkDifferent, ThinkAI, the podcast by Mark and Jens.

00:17:11Two tech-loving minds who not only talk about artificial intelligence,

00:17:16but live it. Here you will find clear classifications, real practical insights

00:17:21and a fresh look at what is possible.

00:17:24Understandable, critical, and always with a wink.

00:17:28A call to think, to smile, and above all to participate.