EchoEkhi

To improve account security and reduce unsolicited password reset emails, AO3 has changed our password reset process. The “username” option has been removed from the "Reset Password" page for logged out users. You can now only reset your password by entering your email address into the password reset form.

This does not affect how you change your password if you're logged in.

If you need help changing your password, contact Support for assistance.

On September 4th, there were 145,806 comments created on the site, of which 1,252 comments were comments with the intention to lead another user off-site.

906 of these were left by 536 different registered users, and 346 were left by guests.

Out of all these, 507 had replies to them within 24 hours and 745 did not.

I went through the 536 accounts' history on that day and at least 97 of them were highly likely to be scammers. Those 97 accounts left a total of 1,675 comments that day.

In aggregate the scammers probably account for around (1675+346)÷145806=1.4% of the total comment activity that day. While the number is not great in absolute terms, it's probably pretty good in comparison to other fanfic sites. And given the fact that a very small number of accounts send out a lot of the scam messages, it should be relatively easy to deal with as long as the accounts are reported by people

The youtube channel's name is Fiction Domain (has a profile pic of some guy from Demon Slayer, I think). Maybe someone out there knows how to deal with this crap.

We're still actively tracking their channels, but if they steal from FFN fics and that fic hasn't been cross-posted to AO3, there's not much we can do about it.

If you're an author, here's what you can do to help us:

Cross-post your old fics to AO3;

Opt-in to automatic notifications so we can notify you about these cases faster;

Spread the word with your fellow authors and get them to do the same.

We're always looking for additional volunteers to help us with the backlog, which now stands at 416 videos pending manual confirmation.

Context: https://www.tumblr.com/echoekhi/735065816966545408

Hi tumblr,

I woke up this morning to a new AO3 comment on my most liked Naruto fic, but this comment was unlike any other I ever got before. It will interest fanfic writers around, especially anime fanfic, so allow me to share it:

Hello, Unfortunately, your work may have been stolen by content farms to create YouTube videos here: (link removed by dianamoth) If you have granted permission for this uploader, you can safely ignore this comment. If not, you should report these videos here: DMCA Complaint Form. You should re-confirm the video contents before you file the report. Note that only the original creator for this work can submit a complaint. When filing the complaint, do not select the "Scheduled: Send a 7-day notice" option, or they won't get a copyright strike from you. We recommend that you do not negotiate with the channel before-hand. Remember to include a link to your work in the 'Title' section of the complaint form. Channels like this have stolen many more works from other writers as well, so your efforts in reporting this channel will also help other writers. See full list of stolen works here: Google sheets Your authorship was identified a while ago by an automated detector under Project Copy-Knight, but for anti-spam reasons we could not inform you until we have manually confirmed each case. If you would like us to notify you as soon as possible, please include this text in your profile page to grant us permission. To prevent a build-up of our notification comments on inactive accounts, we will only send you a maximum of 3 messages before you grant us permission. This is message 1 out of 3. This message was sent automatically, but your case was manually confirmed, and replies to this comment are monitored. If you need additional guidance on how to take the video down, just reply to this comment. To learn more about what we do, please read this article or join our Discord server. From Project Copy-Knight, EsquireBot

I did not give consent for my fic to be used on youtube, ofc.

I was skeptical at first (for context, I write OC Naruto fanfics, who the hell would be interested in that on youtube, I thought?), but this is a well-crafted message with all the ways to verify this claim, so verified it I did!

And hello and behold, this is all true! I'm Amazed. This Project Copy-Knight is quite impressive, it seems well organized and everything, so I thought I would share the word about their existence so that people are aware that this is a problem that exists, that someone is working on it, that it's not spam, and how you can participate, etc... I recommend checking the article they mentioned and the Discord server if you have time to help.

Also, for those who might be in the same case as I and want to report a youtube video (because obviously I'm going to, can you believe this? Just read further to see what a mess it is), let it be known that the link above only works if you have a channel. If not, you have to send your complaint to [email protected]

Now, if you're interested in seeing what a MessTM this is, continue reading about this video that stole my work.

There is a bottom feeder amongst us, fanfic writers and readers, this creature here is stealing work from other authors and placing it behind a paywall on patreon currently with 210 subs at $10 dollars a pop. Needless to say I'm more than a little ticked to discover my own fic Secrets of the Hidden Leaf was added yesterday. I've filed a DMCA claim, and reported their page. Please take a look to see if your work, or someone else you know's work is up there and report them so this leech can be shut down.

What Are “What If” Videos?

These videos follow a common recipe: A narrator, given a fandom (usually anime ones like My Hero Academia and Naruto), explores an alternative timeline where something is different. Maybe the main character has extra powers, maybe a key plot point goes differently. They then go on and make up a whole new story, detailing the conflicts and romance between characters, much like an ordinary fanfic.

Except, they are fanfics. Actual fanfics, pulled off AO3, FFN and Wattpad, given a different title, with random thumbnail and background images added to them, narrated by computer text-to-speech synthesizers.

They are very easy to make: pick a fanfic, copy all the text into a text-to-speech generator, mix the resulting audio file with some generic art from the fandom as the background, give it a snappy title like “What if Deku had the Power of Ten Rings”, photoshop an attention-grabbing thumbnail, dump it onto YouTube and get thousands of views.

In fact, the process is so straightforward and requires so little effort, it’s pretty clear some of these channels have automated pipelines to pump these out en-masse. They don’t bother with asking the fic authors for permission. Sometimes they don’t even bother with putting the fic’s link in the description or crediting the author. These content-farms then monetise these videos, so they get a cut from YouTube’s ads.

In short, an industry has emerged from the systematic copyright theft of fanfiction, for profit.

Project Copy-Knight

Since the adversaries almost certainly have automated systems set up for this, the only realistic countermeasure is with another automated system. Identifying fanfics manually by listening to the videos and searching them up with tags is just too slow and impractical.

And so, I came up with a simple automated pipeline to identify the original authors of “What If” videos.

It would go download these videos, run speech recognition on it, search the text through a database full of AO3 fics, and identify which work it came from. After manual confirmation, the original authors will be notified that their works have been subject to copyright theft, and instructions provided on how to DMCA-strike the channel out of existence.

I built a prototype over the weekend, and it works surprisingly well:

On a randomly-selected YouTube channel (in this case Infinite Paradox Fanfic), the toolchain was able to identify the origin of half of the content. The raw output, after manual verification, turned out to be extremely accurate. The time taken to identify the source of a video was about 5 minutes, most of those were spent running Whisper, and the actual full-text-search query and Levenshtein analysis was less than 5 seconds.

The other videos probably came from fanfiction websites other than AO3, like fanfiction.net or Wattpad. As I do not have access to archives of those websites, I cannot identify the other ones, but they are almost certainly not original.

Armed with this fantastic proof-of-concept, I’m officially declaring war against “What If” videos. The mission statement of Project Copy-Knight will be the elimination of “What If” videos based on the theft of AO3 content on YouTube.

I Need Your Help

I am acutely aware that I cannot accomplish this on my own. There are many moving parts in this system that simply cannot be completely automated – like the selection of YouTube channels to feed into the toolchain, the manual verification step to prevent false-positives being sent to authors, the reaching-out to authors who have comments disabled, etc, etc.

So, if you are interested in helping to defend fanworks, or just want to have a chat or ask about the technical details of the toolchain, please consider joining my Discord server. I could really use your help.

------

Fanfiction is transformatively derivative from the original books or shows, which falls under the fair use clause and therefore does not infringe on the original copyright. At the same time, the fic's author holds copyright over their fanworks, so fanworks are actually copyrighted.

What Are “What If” Videos?

These videos follow a common recipe: A narrator, given a fandom (usually anime ones like My Hero Academia and Naruto), explores an alternative timeline where something is different. Maybe the main character has extra powers, maybe a key plot point goes differently. They then go on and make up a whole new story, detailing the conflicts and romance between characters, much like an ordinary fanfic.

Except, they are fanfics. Actual fanfics, pulled off AO3, FFN and Wattpad, given a different title, with random thumbnail and background images added to them, narrated by computer text-to-speech synthesizers.

They are very easy to make: pick a fanfic, copy all the text into a text-to-speech generator, mix the resulting audio file with some generic art from the fandom as the background, give it a snappy title like “What if Deku had the Power of Ten Rings”, photoshop an attention-grabbing thumbnail, dump it onto YouTube and get thousands of views.

In fact, the process is so straightforward and requires so little effort, it’s pretty clear some of these channels have automated pipelines to pump these out en-masse. They don’t bother with asking the fic authors for permission. Sometimes they don’t even bother with putting the fic’s link in the description or crediting the author. These content-farms then monetise these videos, so they get a cut from YouTube’s ads.

In short, an industry has emerged from the systematic copyright theft of fanfiction, for profit.

Project Copy-Knight

Since the adversaries almost certainly have automated systems set up for this, the only realistic countermeasure is with another automated system. Identifying fanfics manually by listening to the videos and searching them up with tags is just too slow and impractical.

And so, I came up with a simple automated pipeline to identify the original authors of “What If” videos.

It would go download these videos, run speech recognition on it, search the text through a database full of AO3 fics, and identify which work it came from. After manual confirmation, the original authors will be notified that their works have been subject to copyright theft, and instructions provided on how to DMCA-strike the channel out of existence.

I built a prototype over the weekend, and it works surprisingly well:

On a randomly-selected YouTube channel (in this case Infinite Paradox Fanfic), the toolchain was able to identify the origin of half of the content. The raw output, after manual verification, turned out to be extremely accurate. The time taken to identify the source of a video was about 5 minutes, most of those were spent running Whisper, and the actual full-text-search query and Levenshtein analysis was less than 5 seconds.

The other videos probably came from fanfiction websites other than AO3, like fanfiction.net or Wattpad. As I do not have access to archives of those websites, I cannot identify the other ones, but they are almost certainly not original.

Armed with this fantastic proof-of-concept, I’m officially declaring war against “What If” videos. The mission statement of Project Copy-Knight will be the elimination of “What If” videos based on the theft of AO3 content on YouTube.

I Need Your Help

I am acutely aware that I cannot accomplish this on my own. There are many moving parts in this system that simply cannot be completely automated – like the selection of YouTube channels to feed into the toolchain, the manual verification step to prevent false-positives being sent to authors, the reaching-out to authors who have comments disabled, etc, etc.

So, if you are interested in helping to defend fanworks, or just want to have a chat or ask about the technical details of the toolchain, please consider joining my Discord server. I could really use your help.

------

Fanfiction is transformatively derivative from the original books or shows, which falls under the fair use clause and therefore does not infringe on the original copyright. At the same time, the fic's author holds copyright over their fanworks, so fanworks are actually copyrighted.

--

--

--