Dear Eric and PG Board,
Please find a copy below of about a dozen or more emails that Eric, Greg and I discussed back and forth for the needed corrections, replacement, and delisting for the old audiobooks on the site and the need to add and integrate the new AI audiobooks into the site.
To summarize our previous discussions regarding the 1048 human-read and computer-generated audiobooks at Project Gutenberg and the need to add the 70k AI audiobooks:
- Proposals for Improvement: I suggested several enhancements to the catalog of audiobooks, including cross-referencing audiobooks with their source ebooks, improving file naming conventions for better identification of chapters/formats, offering combined ZIP files or single-file versions for easier downloading, delisting the legacy audiobooks all together, and replacing them with the 70K or AI Audiobooks.
In November I emailed Eric several times to inform him: “Let me know how you want me to deliver and help to integrate the next batch of 70k mp3 into the PG website?”
I would prefer FTP access to iBiblio or pglaf.org, so I can upload them for review and inclusion.
- Greg’s and Eric’s Feedback : Greg noted that while these changes are technically possible through database edits or SQL commands, they would be difficult to automate. He expressed a desire for any work to be coordinated with Roger and Percival. However, he also noted that many of the legacy text-to-speech files are of low quality and suggested that our focus might be better placed on directing users to external partners like LibriVox or Microsoft.
To summarize Eric’s reply to this was he had no interest in adding Audiobooks to PG, and felt folks can just go to the external link. Eric also mentioned Technical Challenges, repeatedly mentioning that PG currently has no FTP access and no existing method to easily integrate these audiobooks into the current system. However, I guess he is not aware, but PG does have a robust high-speed FTP server. Our FTP mirroring protocol uses it every day to synchronize the Gutenberg books to our local mirror on our cloud server. Eric might want to check https://www.gutenberg.org/MIRRORS.ALL for all the FTP configuration settings.
I had also written about five different articles in the monthly newsletter and posted on X about the Ai Audiobooks development in detail.
I hope that the following forwarded emails, helps to clarify the numerous audiobooks conversations we have had.
Best regards,
John
------------- John Guagliardo Volunteer Support Project Gutenberg Gutenberg.org
*Project Gutenberg* Online Library of free eBooks *Gutenberg.org*
On Fri, Jul 31, 2026 at 5:52 PM Eric Hellman <eric@hellman.net> wrote:
Please resend the emails you mention. I have not received anything
resembling what you describe. Maybe lost in spam?
Eric
On 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
---------- Forwarded message --------- From: Eric Hellman <eric@hellman.net> Date: Fri, Nov 14, 2025 at 9:15 AM Subject: Re: Replace/remove of 1048 Human read and Computer-generated audiobooks at PG To: Support Gutenberg <central.gutenberg@gmail.com>
Call we schedule a call? There is no ftp on ibiblio, there is no existing method to integrate the audiobooks.
Eric
On Nov 14, 2025, at 1:20 PM, Support Gutenberg < central.gutenberg@gmail.com> wrote:
Eric,
I think we need to include the AI audiobooks. Especially since Newby’s wiki page says:
In 2023, the Project Gutenberg Open Audiobook Collection was named as one of "The Best Inventions of 2023" by TIMEmagazine, after the organization collaborated with Microsoft and the Massachusetts Institute of Technology to employ text-to-speech technology to transform 5,000 books into artificial intelligence-narrated audiobooks.
Let me know how you want me to deliver and help to integrate the next batch of 70k mp3 into the PG website?
I would prefer FTP access to iBiblio or PGLAF, so I can upload them for review and inclusion.
Best regards,
John
------------- John Guagliardo Volunteer Support Project Gutenberg Gutenberg.org
*Project Gutenberg* Online Library of free eBooks *Gutenberg.org*
On Fri, Nov 14, 2025 at 8:28 AM Support Gutenberg < central.gutenberg@gmail.com> wrote:
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.
John
------------- John Guagliardo Volunteer Support Project Gutenberg Gutenberg.org
*Project Gutenberg* Online Library of free eBooks *Gutenberg.org*
On Sat, Nov 1, 2025 at 11:42 AM Eric Hellman <eric@hellman.net> wrote: What I think is a good longterm goal is to have a general method to link to *typed* related resources. (by typed, I mean human-read audiobooks, generated audiiobooks etc. We do wikipedia links, for example. We can use the same method to add links to the MIT-microsoft collection, to librivox, and to other editions in our catalog which we do awkwardly in the notes field0. The database has well developed data fields to support this but we have no workflow to support it. I trust you recognize that doing this well is a significant project requiring collaboration between a number of PG teams.
The existing audio books currently at PG are a different matter. IMHO 90% of them should be de-indexed and ignored. We are shitty at presenting them. We can put a bandaid on them but bandaids tend to (and do!) come off. We can best serve our audiobook users by helping them find the audiobooks and then send them (with links!) to people like librivox that we are aligned in mission with and do a great job presenting audiobooks. Eric
On Nov 1, 2025, at 2:26 PM, Support Gutenberg <central.gutenberg@gmail.com> wrote:
Eric,
Here is the resend of the email. My list of suggestions and Greg’s comments.
Best regards,
John
------------- John Guagliardo Volunteer Support Project Gutenberg Gutenberg.org
*Project Gutenberg* Online Library of free eBooks *Gutenberg.org*
On Sat, Sep 20, 2025 at 11:01 AM Support Gutenberg < central.gutenberg@gmail.com> wrote:
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
Dear Greg,
Sorry to disturb you, but I have a few questions for whenever you have
On Tue, May 13, 2025 at 10:52:21AM -1000, Support Gutenberg wrote: 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
--
---------- Forwarded message --------- From: *Eric Hellman* <eric@hellman.net> Date: Mon, Sep 22, 2025 at 4:47 PM Subject: Re: 1048 Human read and Computer-generated audiobooks at PG To: Support Gutenberg <central.gutenberg@gmail.com> CC: Greg Newby <gbnewby@pglaf.org>
I had some students link up the librivox titles, and Standard Ebooks versions to the corresponding pg books for an aggregation site funded by the Knight Foundation. It worked nicely but we decided it solved no user problems and added little so it was mostly turned off.
Eric
On Sep 22, 2025, at 6:27 PM, Support Gutenberg < central.gutenberg@gmail.com> wrote:
Dear Eric,
Yes, Percival was hired in 2021 by WLF to clean up and maintain the Self.Gutenberg.org <http://self.gutenberg.org/> website. He was hired to fix up lots of problems created when our previous programmer got sick and unexpectedly hospitalized for about a half a year.
After 2022, when Percival's one year contract was up, the World Library Foundation kept him on. Since there is a considerable amount of code overlap between the PG author's community and our WLF sites, he is also now become our general Webmaster too.
The Self publishing portal and Gutenberg Central were created about 9 years ago. It costs WLF about 20K a year to maintain it. We appreciated PG helping out back in 2021 to pick up one of those years. WLF had lost all of our funding during the pandemic and needed the help.
Regarding the audiobooks, Librivox has provided us with a CSV for about 33K of meta records (download <https://docs.google.com/spreadsheets/d/1s8Tlab_cuPl8oqQqU8zHLp8fYWmPDbXanpxZzHg-kh4/edit?usp=sharing>). If PG wants to delist the previous audiobooks, there are some really nice ones the Librivox folks have made to replace them with. I recommend delisting rather than deleting the previous ebooks generations, and redirecting the links to the Librivox ones that were created from PG books.
Personally, I love listening to the Librivox ones, and think they would make a nice addition. I also really like the AI generated ones too.
If you foks rather not deal with the Mp3 audiobooks and Mp4 videobooks, WLF can help with it at Gutenberg Central. It may take a few months to resurrect the portal and bring its development current. But the idea of Gutenberg Central was to create a single OPAC and Full Text discovery portal for all the PG eBooks, Self-published titles, Librivox AudioBooks, and other tangential PG derivative efforts, such as Penn State Electronic Classics, Djvu Editions Classic Literature, and Daniel Mozz's AI VideoBooks (link <http://central.gutenberg.us/Members/DanielMozz>), etc... The development was discontinued due to the cost to keep it updated.
You can view our beta-development at http://central.gutenberg.us/Results?SearchEverything=tempest. The Beta site is slow and NOT fully developed but, this link can give you an idea of the direction.
Best regards,
John
------------- John Guagliardo Volunteer Support Project Gutenberg Gutenberg.org
*Project Gutenberg* Online Library of free eBooks *Gutenberg.org*
On Sun, Sep 21, 2025 at 3:39 AM Eric Hellman <eric@hellman.net> wrote: Percival works for you?
I know the #2 problem. It's bothered me too, but as we're not adding new audiobooks, fixing it has been a low priority. We have the same problem with other types of material - data, images, etc.
I think the solution for all these three is to generate README.html pages with links - we could treat these more like books. We could then stop the context-free file lists on the bib pages.
This would be done in ebookmaker. I think it's mentioned in the ebookmaker roadmap issue, but please comment there if you can: https://github.com/gutenbergtools/ebookmaker/issues/258
But also, the computer generated audio files are problematic for *several* copyright reasons and we should waste as little time on them as possible. I would just remove them, but Greg has emphasized that PG does its best not to remove things, and I think that's a sensible policy, too.
Problem #1 gets reshaped by this solution a bit.
Eric
On Sep 20, 2025, at 5:01 PM, Support Gutenberg < central.gutenberg@gmail.com> wrote:
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
Dear Greg,
Sorry to disturb you, but I have a few questions for whenever you have
On Tue, May 13, 2025 at 10:52:21AM -1000, Support Gutenberg wrote: 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
---------- Forwarded message --------- Greg Newby <gbnewby@petascale.org>, Ilana Kingsley <stinkypup@gmail.com>; John Guagliardo <central.gutenberg@gmail.com>,
Happy to! Here’s the Google Doc with the reviews, please feel free to add comments directly.
On Mon, Dec 15, 2025 at 5:41 PM
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 <https://www.hajim.rochester.edu/ece/>*, *CS <https://www.cs.rochester.edu/>*, *GIDS <https://www.sas.rochester.edu/dsc/>* of University of Rochester
Director – *Audio Information Research (AIR) Lab <https://labsites.rochester.edu/air/>*
President – *International Society for Music Information Retrieval <https://ismir.net/>*
Co-Founder - *Violy <https://violy.app/>*
Mail: University of Rochester, 720 Computer Studies Building, Rochester, NY 14627, USA Tel: +1 585-275-5302
Web: *http://www.ece.rochester.edu/~zduan/ <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/autho... Paper 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
-end-
--