The Content Particle-izer
This is spec for a new content production engine which allows a person with any level of attention span take in an appropriate version of the same story.
We were trying to determine the best length of a recorded conversation on a topic to serve as an individual episode where someone would listen to it in one sitting.
One question was: “What’s the average time a person exercises?” Another: “What’s the average drive time?” I was going to do some research on these averages when I realized the real question is what’s a person’s average attention span?
Because this is for a podcast, I started there and found one study measured the average podcast time is 38 minutes, 42 seconds.
A StoryCorps interview is 40 minutes and that’s based on 60,000 interviews among more than 100,000 participants in all 50 states and then archived in the Library of Congress
If we set 40 minutes as the benchmark for a full discussion or conversation, then the Content Particle-Izer uses AI to compress a transcript of that conversation down 8 discrete reductions so you end up with a single conversation reduced down to 8 shorter summaries — roughly halved with each reduction until you get down to 15 seconds at its smallest.
Then, a deep-fake audio generator uses the podcaster’s voice to regenerate the podcast into those 8 different time components so that anyone can choose whichever version they want to listen to.
A way to maintain the quality of the story as the audio is reduced, is to record the discussion in three parts: a beginning, middle and end, and then use the AI reducer discreetly against each of the three segments so as to make sure the 3 part structure always exists.
If you were charged with giving a presentation on a topic and we’re told you have twenty minutes to present, you would create and trim your presentation down to that time frame.
If you were told, no wait, you have to give the same presentation in 10 minutes — you could do it.
The more time, the more detail.
AI can reduce text passages down in varying level of percentages (i.e. 50% less words) while still keeping the core ideas intact.
Enter deep fake audio and 8 presentations can be created from every single presentation automatically.
Notes
Record the audio — 40, 60 or 80 minutes
How long can someone focus?
Most movies are 80 - 120 minutes
Average length of a podcast 38 minutes, 42 seconds
StoryCorps interview length 40 minutes
AI reduces the initial audio capture down in increments
Transcribe
Reduce down in half increments (the content algorithm is in the increment)
40 min
20 minutes
10 min
5 min
2.5 min
1.25 min / 75 sec
1 min
30 sec
15 sec
8 TIME VERSIONS of every story
Every story is told in 3 parts: A beginning, middle and end.
The AI content reducing is done individually on each part (i.e. 3 times per story) so as not to compress important story structure out of the reductions.
Recreate the audio based off of the reduced transcripts
https://wellsaidlabs.com/
https://murf.ai/
https://voiceovermaker.io/
Deep fake audio tools
Content Summarizer / Paraphraser Tools
https://quillbot.com/
https://www.prepostseo.com/
https://tldrthis.com/
Music component changes
Epidemic sound and stems
Membership
Choose your ideal audio length
I propose this as Content Engineering. Not Content Design because content design implies taking content and organizing and styling a static capture of information.
Content Engineering creates new and different lenses for seeing the same or all information of a given focus.
In this example, a universal reduction serves as a structure to the content and also to the person emitting the content. An outline of ideas and sub ideas to discuss with differing levels of focus (beginning, middle, end) and developing a beat by consistently repeating the conversation (weekly 40 minute discussions)
Links
Voice over AI audio generators