Please resend the emails you mention. I have not received anything resembling what you describe. Maybe lost in spam?EricOn Jul 31, 2026, at 7:35 PM, Support Gutenberg <central.gutenberg@gmail.com> wrote:Dear PG Board,I just took a look at the webcast from the last meeting, want to make sure you folks are aware I’m partnir the AI development team. For the last couple years, I’ve been working closely with The Audiobooks project folks. You will notice I am one of the contributors for the IEEE paper and am listed as one of the contributing developers. I’ve attend the development meetings and am part of the AI audiobook team slack. Please let me know how I can be of assistance.I have sent multiple email to Eric and Greg with extensive notes and proposals on how we could best integrate the AI Audiobooks and how best to clear out the old legacy rubbish that is currently on the site. Perhaps he can share a copy of those emails and ideas with you folks too.Best regards,John
Eric,What are your thoughts on where to store these files?
There are terabytes of files on different PG servers owned by gbnewby. For example /data/gbn on inferno is 3.7T. It seems to be related to the microsoft audioproject. Are those files the master copies? Are they a mirror? Are they important at all given they are two years old?
I have cataloged in an excel sheet these 5k of mp3 files and 65k more of the next phase.
Dear Eric,
Here are the suggestions for how to improve audiobook presentation on Gutenberg.org.
Best regards,
John
---------- Forwarded message ---------
From: Greg Newby <gbnewby@pglaf.org>
Date: Tue, May 13, 2025 at 6:36 PM
Subject: Re: 1048 Human read and Computer-generated audiobooks at PG
To: Support Gutenberg <central.gutenberg@gmail.com>
Hi John. I'm always happy to correspond :)
Thoughts below:a
On Tue, May 13, 2025 at 10:52:21AM -1000, Support Gutenberg wrote:
> Dear Greg,
>
> Sorry to disturb you, but I have a few questions for whenever you have the
> energy to reply.
>
> I was reviewing Browse By Category: Audio Book, human-read | Project
> Gutenberg <https://www.gutenberg.org/browse/categories/1> and Browse By
> Category: Audio Book, computer-generated | Project Gutenberg
> <https://www.gutenberg.org/browse/categories/2> had a few comments and
> recommendations.
>
> Here are the suggestions for example record
> https://www.gutenberg.org/ebooks/22446
>
> 1. What do you think about combining audiobooks/# with having a reference
> to the ebook/# that the audio book was created from; e.g.,
> https://www.gutenberg.org/ebooks/30624, so the patrons can read along with
> the audio? Since https://www.gutenberg.org/ebooks/30624 was the text file
> used to create the audio version at https://www.gutenberg.org/ebooks/22446 ,
> what do you think about adding a single file version or zip file of the
> combined chapters of audio files to the details page that has the source
> file, or at least adding a cross reference to each other?
That can be done. It would involve editing the catalog database,
file by file (for around 1000 files) or of crafting the needed
SQL "insert" commands.
It's not going to be easy to automate, though Johannes might have
ideas for putting an LLM to work on the task. He did the links
to Wikipedia books for thousands of titles successfully.
> 2. On the details page for the audiobooks (i.e.,
> https://www.gutenberg.org/ebooks/22446 ), the list of files that are
> available are not identified as differentiated. They are all displayed
> with the same title name. The title phrase could include the file name or
> chapter name in the hyperlink reference <a> </a>.
Yeah, that's an artifact of mostly not having a nice index page
that links to all the files in order.
To make this work as expected, we'd need to add a file like
22446-h/22446-h.htm
linking to the individual files (which might be in ../ogg/22466-01.ogg, for example)
We can push a new 22466-h.htm easily enough, following our standards, and
it will "just work." But it needs to follow our standards so it's in the
right place for the catalog database to find & use it.
This would be a good task for Percival. Roger is an expert in what
needs to be done, and I'd want Percival's work to filter through Roger.
It's not hard, but it's picky - the PG back end needs files to have
specific names & directory locations for the automation to work as
expected.
> <a href="/files/22446/ogg/22446-01.ogg
> <https://www.gutenberg.org/files/22446/ogg/22446-01.ogg>" type="audio/ogg"
> class="link " title="Download">Ogg Vorbis Audio</a> *change to* title="
> Download">Ogg Vorbis Audio 22446-01
> <https://www.gutenberg.org/files/22446/ogg/22446-01.ogg></a> <a href="
> /files/22446/ogg/22446-02.ogg
> <https://www.gutenberg.org/files/22446/ogg/22446-02.ogg>" type="audio/ogg"
> class="link " title="Download">Ogg Vorbis Audio</a> *change to *title="
> Download">Ogg Vorbis Audio 22446-02
> <https://www.gutenberg.org/files/22446/ogg/22446-02.ogg></a> For ebook
> 22446 there are 4 file format types; speex, MP3, Apple iTunes, and Ogg.
> They all have one file per chapter, 12 chapters times 4 formats, which
> creates 48 links to files that have no specific reference to the title.
> This could be improved by including the file name, which includes the
> chapter in the name, so no additional cataloging would be needed. Just a
> minor change to the xml display page.
> [image: 5-13-2025.JPG]
> 3. What do you think about including a combined version of the version
> types, so I patron can play/download all 12 chapters with a single click?
Sure, but do you want to do so much work on these crappy-ass audio
files? They are mostly terrible.
Combining a bunch of MP3 files into a single file is pretty easy to
automate, and this task could be combined with #1 and #2 above.
I have basically abandoned the legacy audio files. They are messy
to deal with, but more importantly they are just not very good.
> 4. What do you think about the files when they open, they go to a new
> window/tab or iframe, rather than being sent to a page without navigations?
I don't have a strong view on this. Usually we keep people in the same
window for reading a book online, but I have no objection to opening
in a new tab or window.
What we don't want to do is get into the audio player business. We
need the player to be whatever people have on their computer/browser,
without Javascript or anything else that is needed for playback.
> If you are interested in moving forward with any of the ideas, I could work
> directly with the catalogers to help with implementation, so as not to take
> up your time or energy?
I encourage any ideas for improvements, and at the same time feel
like this is putting lipstick on a pig. The human readings are excellent,
and mostly available elsewhere.
The Microsoft readings are good, but we don't host them and I don't
want to - they no longer fall under our collection development policy.
The 700ish text-to-speech files are, generally, terrible. We're better
off sending people to Microsoft and Librivox, which is the point of
the audio blurb on the main page of https://www.gutenberg.org
That said, I'm happy to guide you & Percival to work with Roger and
me to make some improvements.
Thanks as always.
~ Greg
--
LARGE SCALE AUDIOBOOK SYNTHESIS USING LLMS AND TEXT-TO-SPEECH is Amazing!! Can you also share a Google doc with the reviews so we can all look at it together??On Mon, 15 Dec 2025 at 4:33 pm,Hi Everyone,I think the reviews are generally positive! I'm quite impressed by the quality of the reviews; they are substantial and well elaborated.Stella, could you draft the rebuttal mainly focusing on comments form the third reviewer?Also, note that this rebuttal will be read by the Area Chairs instead of the reviewers.Best,Zhiyao--------
Zhiyao Duan
Professor – ECE, CS, GIDS of University of Rochester
Director – Audio Information Research (AIR) Lab
President – International Society for Music Information Retrieval
Co-Founder - Violy
Mail: University of Rochester, 720 Computer Studies Building, Rochester, NY 14627, USA
Tel: +1 585-275-5302Web: http://www.ece.rochester.edu/~zduan/
From: ICASSP 2026 <papers@2026.ieeeicassp.org>
Sent: Monday, December 15, 2025 3:57 PM
Greg Newby <gbnewby@petascale.org>; Ilana Kingsley <stinkypup@gmail.com>; John Guagliardo <central.gutenberg@gmail.com>;
Subject: [EXT] ICASSP 2026: Author response requested by 22 December [Paper #14348]Dear Greg Newby, John Guagliardo:
Paper ID: 14348
Title: LARGE SCALE AUDIOBOOK SYNTHESIS USING LLMS AND TEXT-TO-SPEECH
Your paper has undergone the peer review process, and the program committee has requested a response to the reviews from the authors.
You may now prepare a response (400 words maximum), which will be reviewed by the Area Chair prior to making the final decision on your paper. The goal of the response is to inform the Area Chair of major misunderstandings or incorrect statements in the review reports. Therefore the rebuttal should focus only on major critiques. We recommend to group reviewers' comments into a few major categories; prioritize and focus on addressing major concerns, and also condense the response down to the essentials.
Please note that submitting a response is an optional step for you to address the reviewers' comments, rebut factual errors or to supply additional information requested by the reviewers.
https://urldefense.com/v3/__https://cmsworkshops.com/ICASSP2026/papers/author_response.php__;!!CGUSO5OYRnA7CQ!fC50--mCVka-Hb3gnDMi9f830ytdPSGzKq4i_GGP1E7BEAJsfLTfvbd7ZqOxzJ93hxVU-0KkpboEVrwO6WSE76akGL5Paper Number: 14348
Paper Password: 6E04C8B5
Responses are due on 22 December 2025, in order to review and send out the acceptance/rejection notifications on time.
Kind regards,
Christian Jutten, Markus Rupp and Tülay Adalı
ICASSP 2026 Technical Program Chairs
papers@2026.ieeeicassp.org