She's a snake, but has the mouth of a cow

"we love you nisha!" we all say in unison

jexson128 said:
"we love you nisha!" we all say in unison

No. No, I do not love Nisha. I ADORE Nisha!!!

Hold up, why was the other sound version removed?

This is a considerably worse sound edit comparing to the original post in my opinion, with the background music, most of the moans sounding as if the lady doesn't have a mouth full of meat and general noise pollution, it would be nice not to have the former version deleted, considering the soundless one is still up as a parent.

kowase said:
Hold up, why was the other sound version removed?

This is a considerably worse sound edit comparing to the original post in my opinion, with the background music, most of the moans sounding as if the lady doesn't have a mouth full of meat and general noise pollution, it would be nice not to have the former version deleted, considering the soundless one is still up as a parent.

It often seems like random chance which sound edits stay up and which ones are deleted right away.

mintymofo said:
It often seems like random chance which sound edits stay up and which ones are deleted right away.

No. The more concerning part is that anyone believes that instead of understanding how we do things.

The problem

In the case of videos, the number one thing to remember is that we do not evaluate audio. We only judge visual quality. The sound may as well not exist. Audio quality must be offensively poor for us to delete a video for that. To that point, e621 views audio edits as pure duplicates, which is a templated flag reason for users and deletion reason for staff (the "inferior/duplicate" one). Technically, according to old site doctrine, audio edits should be deleted as duplicates if they don't meaningfully change the visual information, but, you know... that hardline attitude would be just too hated lol. So we allow audio edits. However, that means they are still held up to the same standard as other third-party edits: the visual quality must be the same as from the official source.

And therein lies the problem. What usually happens, on Twitter, is the artist posts their video as a normal Twitter upload and Twitter takes a shit on the video quality. As e621 staff, we have to lower our standards for video to at least average Twitter quality and only delete the really bad ones (i.e., they'd never be fit for e621 using the Twitter version). Sound editors take the artist's Twitter upload, apply their audio, and reupload the artist's video to get shit on by Twitter a second time. This is automatically worse visual quality than the artist's original Twitter post, so e621 posts of those kinds of reuploaded Twitter videos are subject to heavy scrutiny and likely deletion. Further, sound editors often don't handle videos well anyway, so they may introduce a third round of visual quality loss.

The solution

Avoiding that quality degradation, however, is simple. Take the original artist's upload (can use the version already on e621 if it's decent), take the audio from the sound editor's version, and combine the two without re-encoding the video. Then our copy of the audio edit will have equivalent visual quality to our copy of the original video.

Here's a full copy-paste from my own notes for myself how to mux video and audio using ffmpeg, which is what we recommend for video conversion:

Stream Extraction and Muxing for Audio Edits

- can use ffprobe to find stream mapping
-> Stream #0:0[0x1](und): Video: h264 <- 0:0 is h264 video
-> Stream #0:1[0x2](und): Audio: aac <- 0:1 is aac audio

Extract without re-encode

ffmpeg -i 1788495283673829749_1.mp4 -map 0:1 -c:a copy audio.aac

- map 0:1 -> seen from ffprobe
- map 0:a -> all audio streams
- c:a copy -> copy audio without re-encoding

Extract with re-encode

ffmpeg -i 1788495283673829749_1.mp4 -map 0:1 -c:a libopus -b:a 128k audio.ogg

- single stream audio in new audio files is stream 0:0

Mux video and audio from two files

ffmpeg -i video.webm -i audio.ogg -map 0:0 -map 1:0 -c copy output.webm

This is actually really easy when it's all spelled out like this.

What happened here

So, finally, what happened with this video's audio edit? The original artist only provided the public with a normal Twitter post of this video, which is their prerogative as artist and specifically a part of their monetization strategy as mentioned on their Patreon. Fine. That video was posted to e621 in fair quality. Fine. That video got an audio edit on Twitter, which was posted to e621. The audio edit posted to e621 had notably worse visual quality than the soundless version of this video. Our automatic stance is delete. We (staff... which might only be like me or mairo) could have fixed it how I described above. I have fixed some of these sound edits up for deletion in the past. However, I saw a different sound edit, this one, retweeted by the artist, and the sound editor had shared the video in high quality in a Mega link. I asked the artist if this was okay (can be viewed as a Patreon leak), and they did not object because they provided the sound editor with the video at that quality.

This is now the highest quality version of this video on e621. I would replace our copy of the soundless post with this version with the audio track removed, but the sound editor added their watermark to the video so that would just be confusing. Three staff had already disapproved of the other sound edit for poor visual quality but did not push delete, so that video was finally deleted as inferior. I moved that video's favorites to this video, though, so people would at least have a sound edit in their favorites. I could have combined that edit's audio with our copy of the original video as outlined above, but I did not. Technically, the original video should also be deleted as clearly visually inferior to this one, and people can mute the audio if they don't want it. And combining this video with the other sound edit would be misattributing the sound editors due to the watermark, so that's not an option either. Funny dilemma, isn't it?

Edit: Fixed a stupid in my muxing command. Use -c copy after the -map options. How embarrassing lol.

abadbird said:
No. The more concerning part is that anyone believes that instead of understanding how we do things.

The problem

In the case of videos, the number one thing to remember is that we do not evaluate audio. We only judge visual quality. The sound may as well not exist. Audio quality must be offensively poor for us to delete a video for that. To that point, e621 views audio edits as pure duplicates, which is a templated flag reason for users and deletion reason for staff (the "inferior/duplicate" one). Technically, according to old site doctrine, audio edits should be deleted as duplicates if they don't meaningfully change the visual information, but, you know... that hardline attitude would be just too hated lol. So we allow audio edits. However, that means they are still held up to the same standard as other third-party edits: the visual quality must be the same as from the official source.

And therein lies the problem. What usually happens, on Twitter, is the artist posts their video as a normal Twitter upload and Twitter takes a shit on the video quality. As e621 staff, we have to lower our standards for video to at least average Twitter quality and only delete the really bad ones (i.e., they'd never be fit for e621 using the Twitter version). Sound editors take the artist's Twitter upload, apply their audio, and reupload the artist's video to get shit on by Twitter a second time. This is automatically worse visual quality than the artist's original Twitter post, so e621 posts of those kinds of reuploaded Twitter videos are subject to heavy scrutiny and likely deletion. Further, sound editors often don't handle videos well anyway, so they may introduce a third round of visual quality loss.

The solution

Avoiding that quality degradation, however, is simple. Take the original artist's upload (can use the version already on e621 if it's decent), take the audio from the sound editor's version, and combine the two without re-encoding the video. Then our copy of the audio edit will have equivalent visual quality to our copy of the original video.

Here's a full copy-paste from my own notes for myself how to mux video and audio using ffmpeg, which is what we recommend for video conversion:

Stream Extraction and Muxing for Audio Edits

- can use ffprobe to find stream mapping
-> Stream #0:0[0x1](und): Video: h264 <- 0:0 is h264 video
-> Stream #0:1[0x2](und): Audio: aac <- 0:1 is aac audio

Extract without re-encode

ffmpeg -i 1788495283673829749_1.mp4 -map 0:1 -c:a copy audio.aac

- map 0:1 -> seen from ffprobe
- map 0:a -> all audio streams
- c:a copy -> copy audio without re-encoding

Extract with re-encode

ffmpeg -i 1788495283673829749_1.mp4 -map 0:1 -c:a libopus -b:a 128k audio.ogg

- single stream audio in new audio files is stream 0:0

Mux video and audio from two files

ffmpeg -i video.webm -i audio.ogg -c:v copy -map 0:0 -map 1:0 output.webm

Bonus: Loop Input Video

ffmpeg -stream_loop 5 -i input.mp4 ...

-stream_loop -> number of loops after first playback, so one fewer than desired

Audio Too Long

-use -shortest

-maybe it's a frame too long causing an awkward pause

-video too long -> default behavior is to pad the shortest input to the longest input, so not a problem?

This is actually really easy when it's all spelled out like this.

What happened here

So, finally, what happened with this video's audio edit? The original artist only provided the public with a normal Twitter post of this video, which is their prerogative as artist and specifically a part of their monetization strategy as mentioned on their Patreon. Fine. That video was posted to e621 in fair quality. Fine. That video got an audio edit on Twitter, which was posted to e621. The audio edit posted to e621 had notably worse visual quality than the soundless version of this video. Our automatic stance is delete. We (staff... which might only be like me or mairo) could have fixed it how I described above. I have fixed some of these sound edits up for deletion in the past. However, I saw a different sound edit, this one, retweeted by the artist, and the sound editor had shared the video in high quality in a Mega link. I asked the artist if this was okay (can be viewed as a Patreon leak), and they did not object because they provided the sound editor with the video at that quality.

This is now the highest quality version of this video on e621. I would replace our copy of the soundless post with this version with the audio track removed, but the sound editor added their watermark to the video so that would just be confusing. Three staff had already disapproved of the other sound edit for poor visual quality but did not push delete, so that video was finally deleted as inferior. I moved that video's favorites to this video, though, so people would at least have a sound edit in their favorites. I could have combined that edit's audio with our copy of the original video as outlined above, but I did not. Technically, the original video should also be deleted as clearly visually inferior to this one, and people can mute the audio if they don't want it. And combining this video with the other sound edit would be misattributing the sound editors due to the watermark, so that's not an option either. Funny dilemma, isn't it?

i was just here to be at the stripped club jorking it (and lets just say...) and was not expecting to read a in depth essay on how the e621 approval system works
thanks e621

Snek