Clojurians Log v2
clojuredesign-podcast
@neumann @nate I started to appreciate more and more the summaries of your podcasts; the key observations and messages. How did you go about that? Did you construct them manually or has AI helped you? I'm asking because I have recorded the first podcast episode on product-centered engineering (it's an internal company podcast), and I'd love to add those observations to the episode summary. using AI from subtitles would be one way, but re-listening to the episode and contructing that bullet list manually might have value on its own 🤔
@nate @neumann wow, thanks a lot for the insights and the process 💚 I experimented with AI summaries as well and my initial impression was exactly as you described as well. I also realized that I actually want to listen to the episode anyway and digest for myself what we've talked about. Thanks for the tip on Audacity; I haven't yet added intro/outro, so for the next one, I'll play with it. One unexpected thing that happened to me was that both of our microphones were quite bad, so I eventually stumbled upon Adobe Podcast Enhance Speech tool, which did wonders to improve the audio-quality. I think if the podcast takes off, I'll invest into higher quality mics. I also want to share that I stole your idea of audio-focused podcast 😄 The topic we discussed was about collaborative and visual system design (EventStorming and the like). I was tempted to show the Miro board, but eventually decided to try just talking about it; so that people can listen to the episode while commuting or running. Thanks a lot for that idea 🙏
Hey @marcel187 congrats on recording your own podcast, I hope it is well received. Early on, the quotes and summaries were done manually by listening to the episode and taking notes. Later on, we were able to extract transcripts (@neumann has the exact tool used), and we could pull quotes from those. Having a summary and list of quotes is incredibly valuable to help the content of the episode stick, as well as to help us remember key points when we are searching for an episode to share. It takes some time (manually or with AI help), sometimes taking longer than the recording itself, but it's extremely valuable.
@marcel187 Congrats on the podcast! That's so fantastic to hear you're creating something and sharing it—even if it's only internal! I started using AI models to transcribe so I could pull quotes from the transcript, but as @nate mentioned, the summaries are all hand written. I did try AI for summarizing, but it wasn't very good. It wasn't very precise. It didn't emphasize what I thought was important. It didn't follow our tone. It left important things out and included unimportant things. My specific process is: • use https://github.com/Purfview/whisper-standalone-win/ to transcribe the episode (with CUDA acceleration) • drop the tracks into Audacity and listen through the whole episode there • basic editing: cut out mess ups, compress and normalize audio, add the intro and outro, etc. • jot down notes about topics, themes, and concepts and try to organize them in a clear way • when I hear a quote I like, I copy it from the transcript and paste it into the notes. I often have to adjust some wording. To see the latest evolution of that process, check out the latest episode: https://clojuredesign.club/episode/118-pure-parts/.
After I got over the horror of listening to myself talk, I found that I liked listening to the episodes. I had observations in the moment that weren’t in my notes. Intro and outro are fun, but not essential. There are royalty-free music libraries out there. I believed we used one from a music bank Google created for podcasters. Yeah, nothing beats a good mic, but good to know about that Adobe product. I’ll have to try it for some other recordings I have. You can get a big studio mic, and they are nice, but a more recent option are small, magnetic, wireless lavalier mics. DJI and Hollyland make some inexpensive ones. If you use those, the deader the room, the better. Thats fantastic to hear about your experience with audio only. It’s a challenging limitation, but like you say, it unlocks so many options for listening. I’ve found that it forces us to focus on concepts and take more care explaining things.
Just my 5 cents to the topic of audio tooling: ardour has a plugin tailored for podcast editing. Apart from that, if you want to dive deep into audio production with FOSS, I definitely recommend taking a look at ardour:)
@aaronrebmann I've heard of it, but I've never tried it out. Thanks for the recommendation!
I'll be here to support 😄 the learning curve can be steep, but it's worth it (no problem for a clojurist of course)