Google Illuminate: Books and papers turned into audio • Hacker News Board

Re: Google Illuminate: Books and papers turned into audio

#21

Post by willwade » Tue, Sep 10, 2024, 5:44 PM UTC

oidar wrote: post #5

The voice models for this are very good. I'd love to have granular control over the output of a model like this locally.

Like SSML? See azure tts or google cloud tts, or ibm Watson or even old school system tts like SAPI voices on windows. But I hear you. In a VITS typical model system ssml isn’t standard. Piper tts does have it on the roadmap.

Re: Google Illuminate: Books and papers turned into audio

#22

Post by dgellow » Tue, Sep 10, 2024, 5:47 PM UTC

Really impressive. The podcasting spam we will get from this will be a pain, but really impressive demo

Re: Google Illuminate: Books and papers turned into audio

#23

Post by Jeff_Brown » Tue, Sep 10, 2024, 5:47 PM UTC

colesantiago wrote: post #9

So podcasts are now automated, anything with a speaker or a screen is now assumed to be not human. Is this supposed to be a good thing that we want to accelerate (e/acc) towards?

If can tell where content came from, it's fine with me. If a host of paid spammers or bots can astroturf an opinion and fool me into thinking they are a wide demographic, that's a problem. And it is -- but it predates LLMs.

Re: Google Illuminate: Books and papers turned into audio

#24

Post by nxobject » Tue, Sep 10, 2024, 5:48 PM UTC

A related experiment from Google: NotebookLM (notebooklm.google.com), which takes a group of documents and provides a RAG Gemini chatbot in return.

I wish Google would make these experiments more well-known!

Re: Google Illuminate: Books and papers turned into audio

#25

Post by ansk » Tue, Sep 10, 2024, 5:50 PM UTC

Imagine reading a math or programming textbook where each statement was true with probability 0.95.

Re: Google Illuminate: Books and papers turned into audio

#26

Post by Legend2440 » Tue, Sep 10, 2024, 5:53 PM UTC

smusamashah wrote: post #16

TranquilMarmot wrote: post #10

Earlier quoted context omitted.

From the "Help" modal: "Illuminate is an experimental technology that uses AI to adapt content to your learning preferences. Illuminate generates audio with two AI-generated voices in conversation, discussing the key points of select papers. Illuminate is currently optimized for published computer science academic papers. As an experimental product, the generated audio with two AI-generated voices in conversation may…

Wow. I did not pick anything in the voice as a clue that it's generated. So does it make it current best text to audio system?

I don’t know if Google’s specifically is the best, but these new GenAI-based text-to-speech systems blow away everything else.

Re: Google Illuminate: Books and papers turned into audio

#27

Post by Legend2440 » Tue, Sep 10, 2024, 5:55 PM UTC

alganet wrote: post #14

Cool tech. Now we know that very soon no one will be able to trust podcasts or video narration.

You shouldn’t have been trusting podcasts in the first place, Joe Rogan says plenty of false things no AI required.

Re: Google Illuminate: Books and papers turned into audio

#28

Post by AuthError » Tue, Sep 10, 2024, 5:55 PM UTC

TranquilMarmot wrote: post #8

freefaler wrote: post #3

Great idea. I wonder how long until we'd see a lot of "autogenerated" podcasts with syndicated advertising inside spamming the podcast space. Like with robovoiced videos on YT reading some scraped content.

Would you listen to an auto-generated podcast? Seems like removing the humans from the equation kind of defeats the purpose.

I would watch history pods for sure

Re: Google Illuminate: Books and papers turned into audio

#29

Post by danesparza » Tue, Sep 10, 2024, 5:55 PM UTC

I wonder how soon until this waitlisted service eventually gets thrown on the trash heap that Google Reader is on.

Building trust with your users is important, Google.

Re: Google Illuminate: Books and papers turned into audio

#30

Post by syntaxing » Tue, Sep 10, 2024, 5:55 PM UTC

I’ve been using the ElevenLabs Reader app to read some articles during my drive and it’s been amazing. It’s great to be able to listen to Money Stuff whenever I want to. The audio quality is about 90% there. Occasionally, the tone of the sentence is wrong (like surprised when it should be sad) and the wrong enunciation (bow, like bowing down or tying a bow) but still very listenable.