I ran the audio feature on a short essay I'd written, and it's astonishing to hear my thoughts being expressed by a "conversation. " While certainly amusing, I'm not sure yet how to make the audio feature actually useful to my writing process. Be glad to hear how others are using it to advantage.
Yesterday I made a prototype Vlog audio track drawn from a perplexity.ai session, downloaded the audio file and layered it CapCut underneath matching stock footage. NotebookLM did a great job synthesizing the uploaded source content and the voices were credible and listenable. I would love to see NotebookLM add the ability to edit a transcript of the generative AI podcast talk track to clean up occasional word glitches and even regenerate small sections of dialogue, but keep the remaining conversation intact. This would be similar to the text to audio repair features in Descript.
Super cool work Steven; happy to find you on Substack through Hard Fork! I'm with you on the lol humor but I think why notebookLM flipped the amazing bit for so many people was how easy audio overview podcasts are to listen too - big jump from Siri squared. I do a somewhat regular digest of everything I've written in the last few months so rather than a single post I'll feed the dozen or so I've done since the last digest and see how that looks
I just entered seven papers I wrote on the topic of "accountability" between 1987 and 2012 (one co-authored) in the audio conversation function and in a few minutes it generated an impressive 15 minute discussion of many of the points raised in those works. (tinyurl.com/27qeq326). I immediately thought about the "99% Invisible" year long project focused on a discussion of Robert Caro's massive The Power Broker (they are currently on episode 9 of a planned 12) (https://99percentinvisible.org/club/). It would be interesting to run a comparison of the "conversations"....
PS: I also wondered about the "ownership" (copyright?) of the "conversation". I assume that issue have been explored. It would be of immediate consequence since I just posted to link to the wav file asbove....:-/
NotebookLM will be a great onboarding tool for our volunteers. We've loaded several web pages and have a good conversation for them to start with. During the "podcast" volunteers are encouraged to write down questions that come to mind. When it concludes they "ask" NotebookLM. After a few iterations the web pages will probably be updated. We are considering importing the audio onto Descript for some minor edits, but no glaring errors yet that require it, though a very short version be attempted.
I ran the audio feature on a short essay I'd written, and it's astonishing to hear my thoughts being expressed by a "conversation. " While certainly amusing, I'm not sure yet how to make the audio feature actually useful to my writing process. Be glad to hear how others are using it to advantage.
Yesterday I made a prototype Vlog audio track drawn from a perplexity.ai session, downloaded the audio file and layered it CapCut underneath matching stock footage. NotebookLM did a great job synthesizing the uploaded source content and the voices were credible and listenable. I would love to see NotebookLM add the ability to edit a transcript of the generative AI podcast talk track to clean up occasional word glitches and even regenerate small sections of dialogue, but keep the remaining conversation intact. This would be similar to the text to audio repair features in Descript.
Impressed with the quality of the sum of the parts.
Really useful to turn rough notes into an engaging introduction of a topic.
(Although it might bankrupt the world with all those million dollar questions and make elephants an endangered species.)
Kudos to the team
Super cool work Steven; happy to find you on Substack through Hard Fork! I'm with you on the lol humor but I think why notebookLM flipped the amazing bit for so many people was how easy audio overview podcasts are to listen too - big jump from Siri squared. I do a somewhat regular digest of everything I've written in the last few months so rather than a single post I'll feed the dozen or so I've done since the last digest and see how that looks
I just entered seven papers I wrote on the topic of "accountability" between 1987 and 2012 (one co-authored) in the audio conversation function and in a few minutes it generated an impressive 15 minute discussion of many of the points raised in those works. (tinyurl.com/27qeq326). I immediately thought about the "99% Invisible" year long project focused on a discussion of Robert Caro's massive The Power Broker (they are currently on episode 9 of a planned 12) (https://99percentinvisible.org/club/). It would be interesting to run a comparison of the "conversations"....
PS: I also wondered about the "ownership" (copyright?) of the "conversation". I assume that issue have been explored. It would be of immediate consequence since I just posted to link to the wav file asbove....:-/
NotebookLM will be a great onboarding tool for our volunteers. We've loaded several web pages and have a good conversation for them to start with. During the "podcast" volunteers are encouraged to write down questions that come to mind. When it concludes they "ask" NotebookLM. After a few iterations the web pages will probably be updated. We are considering importing the audio onto Descript for some minor edits, but no glaring errors yet that require it, though a very short version be attempted.
Is there a public road map for Notebook, and the Audio Overview feature?
Insightful.
Hi Steven. Am I the only one that thinks the male AI host sounds like you???