An audiobook streaming service that focuses on timeless classic works in the public domain.
I do everything from building the app to the audio engineering.
One thing I'm especially proud of is the restoration I did on the "War of the Worlds" 1938 Radio broadcast. I'm really happy with how it turned out. I've made it temporarily free to listen to [1] in case anyone is curious. You should compare it with the original [2] and let me know what you think.
Since there's been a reasonable amount of traffic due to this comment, I thought I'd leave a 50% coupon ($11.75 for yr one) for HN folks in case anyone is interested:
I just updated the home page (main nav & body content) to add a "Browse All Books" button that takes you into the app to view the current titles. Appreciate the feedback.
I’m seeing a lot of AI tells in the cover art (like for War of the Worlds) and book descriptions. Didn’t see any discussion of where the narration comes from on the site. Are all your audiobooks restorations of classic readings? Do you hire narrators to record new ones? Or is there AI involved there too?
All the audio is by real humans but I definitely use AI help on the descriptions and images, as graphic design and copywriting are not areas I’m competent in, and as a side gig currently I only have so much time.
A good chunk of the initial audio has been curated and re-engineered/enhanced from librivox, however I’m also working with voice actors to produce originals. For instance I just release A Christmas Carol which is original to our platform (also see Metamorphosis and Alice and Wonderland). More are coming every month but it takes time to develop real audio recordings with humans.
I appreciate your constructive feedback and welcome more!
When I read your first comment, I immediately thought that the audiobooks are voiced by AI. I'm really surprised to learn the opposite.
So you take existing recordings created before 1929 and remaster them? Are recordings (of books published pre-1929) which were created after 1929 in public domain too?
I don't even want to ask about producing and voice actors.. Really nice idea and realization!
There's a really awesome site called Librivox [1], where volunteers narrate books that are in the public domain. Those recordings are also in the public domain as well (this is just part of the Librivox thing). The quality of those recordings (both the narration, and the actual recording quality) varies quite a bit and most of them aren't at a quality I'd expect people to pay for and thus aren't useable for me. I've spent hours and hours sorting through those recordings finding the best ones (from a narration perspective) and then improving the recording/audio quality on them. Those recordings have all mostly been made in the last 20yrs, so they're not old recordings of the books. So, the value I add to the Librivox recordings are: curation/selection, audio enhancement, and a much better delivery mechanism (IMHO).
I'm also simultaneously building out our own library of original audio content by working with voice actors to get them recorded and proof read (this is a very expensive and time consuming process, but also very fun). One of the hardest parts is honestly the proofing process. Once I get finished narration files I have to compare them result with the actual script (as there are always mistakes) and request edits. I use whisper.cpp to transcribe them and then git and a few other scripts to compare the transcript with the actual book text.
I'll also add that I _do not_ use AI Audio narration because it just doesn't sound good IMHO, and I personally hate listening to it. I regularly run experiments to see what the current state of the tech is and it's still pretty far from where it needs to be IMO. I also don't love the idea of AI swallowing absolutely everything.
Currently I have around 20 books with plans on releasing 1 a month (hopefully more). Have several in production right now and several in the bag ready to release over the first several months.
I literally just went live with some podcast adverts a week ago, so I only have 2 subscribers thus far :) so not much in the way of voting. It’s also limited to a preselected set of books (https://www.soundreads.io/voting).
If you’re interested in giving me any feedback on the selection, quality and app, I’d be happy to give you a 6mo trial (I can figure out a way of doing it that doesn’t require a CC or anything).
You mentioned readium in one of your comments. I’m curious why you didn’t use the readium spec and corresponding ecosystem. Coming in cold and just glanced at your blog post, and I see you’re using RN. There is a RN package that leverages readium under the hood for ebook reading (I know because I’m the author), don’t know if it’s a weird actually has the api you need.
EDIT: storyteller is super interesting. Dug around a bit in the code and see that you do seem to be using readium for some things, so I must just be missing some nuance.
Readium is fantastic, and Storyteller uses it basically whenever possible. But Readium is exclusively for reading EPUB contents, and doesn’t have any support for modifying or creating them, which is the primary purpose of this library!
In my case (and I suspect many others) the answer is Xcode. I develop native apps which means I need access to an iOS dev environment. iOS development is nearly impossible without Xcode, so the only way to do it is either pay a virtualization penalty (which is not appealing when Xcode builds already take a massive amount of time), have a specific machine just for iOS work (which would be super inconvenient, even with a virtual KVM), use a Hackintosh (which have problems), or pay Apples exorbitant fees for hardware.
I dream of a day when I can daily drive Linux and also do dev on iOS, but currently that’s not really possible without some unacceptable compromise. I’ve done quite a bit of research and have yet to find a viable solution (recommendations welcome).
This might not be an answer you like, but I find the virtualization penalty (running on proxmox [kvm]) to be acceptable. In synthetic benchmarks, the penalty is possible to measure but small (1-3% typically). Passing through a AMD graphics card and a USB port, I get a pretty good experience in using it and an excellent experience in administering it [backups/snapshots, restores, etc. are all super clean].
If I measure against "is it viable?", for me the answer is "absolutely, yes".
Despite its many flaws, I use Plex as the single place for all my video and audio content. If you pay for the lifetime pass, you get access to an app called Plexamp which is designed for playing back audio.
Cancel your subscription, download all your books from audible, put them on a NAS, run Plex, and use Plexamp for access.
I still buy things via Audible, I just immediately download them and load them into my Plex server.
We’ve built a similar pipeline architecture for our product. One key thing I’ll mention is that we’re using Shaka-streamer which is a python wrapper around Shaka-packager (which in turn is a wrapper around ffmpeg). We queue our transcode jobs into a redis queue, and use k8s to scale the transcode workers based on queue volume. Lastly, as a few folks have mentioned, we have an experimental on-prem transcoding cluster with consumer grade HW that is pretty cheap.
If you’re interested in working on transcoding I’d highly recommend taking a look at Shaka-packager/streamer.
An audiobook streaming service that focuses on timeless classic works in the public domain.
I do everything from building the app to the audio engineering.
One thing I'm especially proud of is the restoration I did on the "War of the Worlds" 1938 Radio broadcast. I'm really happy with how it turned out. I've made it temporarily free to listen to [1] in case anyone is curious. You should compare it with the original [2] and let me know what you think.
[1] https://app.soundreads.io/discover/item/war-of-the-worlds [2] https://archive.org/details/WarOfTheWorlds1938RadioBroadcast...