Emily Glitch Video Analysis

that’s what it does it turns our sound output to you mic input so no noise is introduced.

It’s all software man! hardware is soooo last century :stuck_out_tongue_winking_eye:

Have you recorded it as a video? And if so, can you post it somewhere?

I can strip the audio out in about 2 minutes.

here is a link to all the audio so far I guess https://forums.etarc.org/t/phase-2-azure-voyage/1736/652?u=dragoon4205

Poly if you can do that, then grab it from the twitch stream…

Twitch streams I’ve seen only have one instance of the glitch.

I’ve turned one of the audio streams into a WAV file. You can get it here:

https://www.dropbox.com/s/zdd86b8594d3odl/EmilyGlitch.wav?dl=0

keyboard or mouse clicks - 46.40 - keyboard or mic clicks - Clyp
Emily 1 - Emily Glitch Audio 1 - Clyp
Emily 2 - Emily Glitch Audio 2 - Clyp
Emily 3 - Emily Glitch Audio 3 - Clyp
Emily 4 -Emily Glitch Audio 4 - Clyp
Emily 5 - Emily Glitch Audio 5 - Clyp

Thanks.

This one’s from YouTube:

This file makes a big difference as it clearly contains noise where it is supposed to be silent. I can easily get noise to show using amplification. Can you tell me how you obtained this specific file?

@solarparty I will see what I can do to get the 5 appearances of Emily. I do have the full source from the stream directly from Twitch, as well as the full audio stripped from it. As noted in one of my earlier replies, there is a volume change:

  • The full stream contains a repeating loop of audio, exactly 53 times.
  • The first 22 loops are at a slightly lower volume.
  • After about 47m53s, the volume goes up slightly to repeat another sequence of 31 loops

@Difficultlevel I would really suggest to stick to actual source material for best quality possible with least loss. Playing the stream, to record through your software, will result in quality loss as it introduces an additional step.

For anyone working on the audio parts, it is best to use the Twitch source. There are quite a few online tools, as well as software to do so. Audio can then be stripped using ffmpeg or VLC (convert) for example. Any conversion or extra steps will have negative impact on the quality. I am personally using Twitch Leecher:

I’m fairly certain you guys are dissecting a normal audio issue with twitch streaming, please check this video to see the similarity

I do not think this was caused by lag, as it happened 5 times total, whenever they started showing a glitched Emily. Although it may appear similar to lag, to me this seems done on purpose using a different technique after close inspection.

Here is a video with all 5 glitches in a row from the NY live event as streamed on Twitch on July 29th, 2017. I used a direct source download from Twitch and edited this with Sony Vegas Pro 13.

Video

Note: Do not use this video as source for doing audio science :wink:

Audio (dropbox)

Feel free to download the audio from all the 5 Emily glitches.
Twitch stream source > VLC conversion to:
WAV (PCM32)
FLAC

Approximate time-stamps in the full Twitch stream audio:

  • Glitch #1 1:24:52
  • Glitch #2 1:27:25
  • Glitch #3 1:31:03
  • Glitch #4 1:33:17
  • Glitch #5 1:45:15
2 Likes

I’m not saying it is standard lag, but rather technical interference on the broadcasters end whenever they introduce the Emily video into the stream, and it stutters. I don’t see the sense in making the audio stutter on purpose. I haven’t read all this thread, but I skimmed and haven’t seen any indication there is any message to decipher.

I agree that so far it does not seem likely there is any hidden message in the audio, although it does have some interesting patterns. I do however feel the audio stutter is purposely done some way or another, to accompany the glitching Emily.

2 Likes

Something very curious is going on. I downloaded the Twitch stream myself today, and extracted the audio to a WAV file using FFMPEG.

I amplified the whole thing by 10 dB, and saved it.

If I then examine my file with Audacity, residual audio data is clearly visible in the silences. however, it’s much more prominent in some of the glitched transmissions than others. It’s very obvious in the third one. If I then amplify some of the silences by around 35 dB, I can get the audio back to what seems the original level.

I’ve just tried the same thing on the WAV file you posted, and there’s nothing in the silences. Not anywhere. It’s like your file has been noise gated, and anything below a certain level has been completely removed.

Has somebody got a reliable binary/hexa sequence from this, and tried to see if it wouldn’t be a valid TCP/IP packet?

Note:

A TCP/IP decoder accepts my first four bytes, but it would take me ages with the tools I use to get the full hex string info from my WAV in Audacity.
My hex string starts with : 0x37 75 F7 6D, but also the whole thing could be shifted by one or 2 bits…

Note2:

Moreover I’m not convinced it’s binary, as the ratio 0 vs 1 is very unbalanced.

has anyone tried to separate all four then overlay them together maybe they just cut up the audio

Within the limits of my ability to measure it, the whole thing is based around 23 millisecond time divisions.

The short silences are 23 ms, the long silences are twice that, 46 ms.

The audio bursts between the silences are also in 23 ms blocks, being 1, 2, 3, 4, 5, blocks in length.

The data appears to be encoded into “words”. There will be a long (46 ms) silence, followed by audio bursts of varying length, separated by short (23 ms) silences. The “word” ends with another long silence, and the next “word” begins.

There are believed to be 5 different glitched transmissions. Of the 5 available to me, the pulsed codes hidden in them all appear to be different.

There ought to be some software method of extracting these patterns. Doing it manually is a large task.

yea I have two of them on my computer separated just trying to get the rest then I was going to over lay them to see if it did anything but your right it is a pain

I don’t know if this helps but in one of Emily’s Vlogs (the one with the chart) she did a survey to measure our collective abilities on different aspects of research, I believe she highlights audio/visual analysis, could there be a clue to these glitched videos of her in the results? Or the vlog in general?

The link to the vlog in question is here at 6.13