Floatplane DOES NOT provide users with better audio quality

[RamblingPenguin]

I managed to get a shot with some motion in it, https://imgsli.com/MTk5MzEz 

It still looks pretty much identical, it's possible that the video they got the shot from was already compressed though. 

[Divritenis]

Just now, jkirkcaldy said:

again, there is no information about how they downloaded it etc. 

So for FP I downloaded directly using the download option, 4K quality. Then extracted audo using Kdenlive as WAV. For Youtube, same tool that reddit author used - yt-dlp. There the audio stream was downloaded directly (48kHz webm)

[ProbablyNotFidel]

9 minutes ago, Divritenis said:

Seems that audio does indeed differ from FP to YT. From YT I can download 48kHz stream that essentially has no LPF. 44.1kHz stream has LPF at around 16kHz. FP is around 14.4kHz, same as in the reddit post.

 

FP

 

YT

 

Luke and his team should look into this. I'm not going to brand this as lying, probably more like something that has changed since initial comments on WAN about it. 

With as hard as those cutoffs look, it's almost like they're two completely different sample rates. Like 36 KHz and 28.8 KHz.

[jkirkcaldy]

1 minute ago, Divritenis said:

So for FP I downloaded directly using the download option, 4K quality. Then extracted audo using Kdenlive as WAV. For Youtube, same tool that reddit author used - yt-dlp. There the audio stream was downloaded directly (48kHz webm)

what was your youtubedl command? 

[Divritenis]

2 minutes ago, jkirkcaldy said:

what was your youtubedl command? 

 yt-dlp -f 251 https://www.youtube.com/watch?v=WVycLhE45kg

 

[wanderingfool2]

1 hour ago, AHPanda said:

Not lying, check above post from @Divritenis

Having a LPF doesn't mean much.  It can literally occur from the fact that if they encode it separately for floatplane and upload it without that.

 

1 hour ago, Divritenis said:

Seems that audio does indeed differ from FP to YT. From YT I can download 48kHz stream that essentially has no LPF. 44.1kHz stream has LPF at around 16kHz. FP is around 14.4kHz, same as in the reddit post.

That seems more like an exporting thing than a floatplane thing; as a 48kHz stream should present with a 24kHz signal...if you are missing extra that means it was taken out prior to uploading most likely.

 

What does matter is stuff like bitrate for the audio and the codecs used.  (Or if you could show that it was FP removing it, instead of it being uploaded like that).  My guess is it's more along the lines of them exporting it and having slightly different export settings.  A note as well, 14 kHz vs 16 kHz doesn't make much of a difference; a lot of adults will actually not be able to hear 14 kHz unless it's blasted at max volume and is the only frequency.

 

So yea, check what the bitrate is, and check the other parameters.  No point in trying to compare stats that could have been caused by uploading a different export file.

 

1 hour ago, RamblingPenguin said:

I managed to get a shot with some motion in it, https://imgsli.com/MTk5MzEz 

It still looks pretty much identical, it's possible that the video they got the shot from was already compressed though. 

There was motion, but it's still mostly a black screen and less motion everywhere else.  I can't recall, but if there is anywhere where there's like confetti or something like that, that would be good (but then again I think a lot of youtuber's already avoid shots that will become garbage looking)...then again at 4k YouTube's bitrate is pretty high anyways so it makes it so you have to essentially have a lot of changing stuff in it. (and even then it's a diminishing returns kind of proposition; where the differences can become a few pixel colors different that no one can really notice)

[japers]

1 hour ago, jkirkcaldy said:

what was your youtubedl command? 

The reddit OP listed the format numbers of each file he downloaded with yt-dlp in the post.

[japers]

So i downloaded the same videos OP did and floatplane is still higher video bitrate than youtube at 1080p. 
Here is the mediainfo and exiftool readouts from both the youtube version (top) and the floatplane version (bottom).

 

[LAwLz]

I don't feel like reading through all the replies so sorry if I am repeating what has already been said.

 

This issue has been brought up several times in the past on this forum. I remember, because I used to engage in those conversations and made suggestions for video encoding for Floatplane. I am also fairly sure that one of the Floatplane members said that sending high-bitrate video just wasn't sustainable economy-wise.

I am fairly sure Floatplane stopped (if they ever did) advertising higher video quality as well. I can't find any reference to that on their website.

 

 

I will try and find the previous threads about it.

 

 

 

Edit:

Oh wow... I just read the comments on Reddit. Holy crap a lot of those people have no idea what they are talking about, and for some reason, people are upvoting them. I am referring to comments like this one:

 

 

1143 Kbps is not at all "barely 360p bitrate". It is fairly standard for 1080 video streaming. Youtube's 1080 streams are typically at or below 1Mbps. 

It seems like a lot of people on the subreddit are super eager to call people who point out issues "karma whores", "trolls", "bad faith actors" and so on. It's like they are in heavy denial. It's pretty scary that it seems like several people are also parroting this "the bitrate can't be that low!" claim. It's like people latch on to things they want to be true and then never verify it themselves. 

 

 

 

 

Edit 2:

Found some of the threads I talked about. Here is one from about a year ago when someone complained about poor video quality on Floatplane. I asked for some more info but sadly nobody nobody provided it back then. Good to see people doing it now.

 

Here is the Floatplane developer saying that "sending massive bitrate isn't sustainable cost-wise". Although, the bitrate mentioned in the OP was 4 Mbps and 8Mbps for 720p and 1080p respectively. I think the bitrates suggested by OP (4 and 8 Mbps) are quite high, but for comparison, Netflix back when they had more primitive encoding and streaming (like Floatplane has today) then they used to have ~2,3Mbps or 3Mbps for 720p, and 4,3Mbps or 5,8Mbps for 1080p.

 

So Netflix used to have more than twice as high bitrate for their 720p content than Floatplane has today for their 1080p content. To put things into perspective.

 

If they want to advertise high-quality videos, then I think they need to bump up their quality a bit more than this, especially in the audio department.

 

They might want to update their encoder as well. Their encoding software is almost 7 years old at this point.

[Thepyrodex]

55 minutes ago, LAwLz said:

I don't feel like reading through all the replies so sorry if I am repeating what has already been said.

 

This issue has been brought up several times in the past on this forum. I remember, because I used to engage in those conversations and made suggestions for video encoding for Floatplane. I am also fairly sure that one of the Floatplane members said that sending high-bitrate video just wasn't sustainable economy-wise.

I am fairly sure Floatplane stopped (if they ever did) advertising higher video quality as well. I can't find any reference to that on their website.

 

 

I will try and find the previous threads about it.

 

 

 

Edit:

Oh wow... I just read the comments on Reddit. Holy crap a lot of those people have no idea what they are talking about, and for some reason, people are upvoting them. I am referring to comments like this one:

 

 

1143 Kbps is not at all "barely 360p bitrate". It is fairly standard for 1080 video streaming. Youtube's 1080 streams are typically at or below 1Mbps. 

It seems like a lot of people on the subreddit are super eager to call people who point out issues "karma whores", "trolls", "bad faith actors" and so on. It's like they are in heavy denial. It's pretty scary that it seems like several people are also parroting this "the bitrate can't be that low!" claim. It's like people latch on to things they want to be true and then never verify it themselves. 

Most people just don’t understand how things as complex as this work (I know I don’t really get it) and so the “assume” things that are wrong and run with it

[LAwLz]

3 minutes ago, Thepyrodex said:

Most people just don’t understand how things as complex as this work (I know I don’t really get it) and so the “assume” things that are wrong and run with it

It really isn't that complex though.

Download the video and then look at the file size. Then divide the file size by the number of seconds the video is and you get the average bitrate.

 

 

For example the "What happened to the Ryzen 7950X3D review" video someone posted MediaInfo from.

 

The video is 15 minutes and 38 seconds long. In other words, it is 938 seconds long.

The video is 140MB large. That's MB with a large B. When talking about bitrate we usually talk about Mb with a small b. To convert between the two you multiply by 8 (or divide depending on which you're converting to which).

 

140MB = 140*8 Mb = 1120Mb

 

1120 / 938 = ~1,2

 

The bitrate of the entire video is roughly 1,2Mbps, or 1200Kbps.

That's including the video, audio, metadata, container and so on.

 

 

It's very simple math that even a middle schooler should be able to do with ease. Just divide size by time.

 

[Thepyrodex]

17 hours ago, LAwLz said:

 

 

 

[ImorallySourcedElectrons]

25 minutes ago, Thepyrodex said:

So how much of that data is audio?  Well assuming 5.1 surround YouTube says that’s about 512kbps, so that’s 42% of the 1200kbps you are saying at 1080p in your math.  Do you think video streaming platforms prioritizes audio or video when forced to choose what goes through when most consumers are on small or shallow speakers from laptops/tv, or on smaller headphones that still have a limited range based purely of driver size limitation (that can be overcome but cost $$$ that the AVERAGE user will never spend) 

 

 What level/type of audio quality was encoded into the video?  
 

what was the sampling rate of the recorded audio?  What is bitdepth of said audio?  What format was it encoded into the video before uploading to YouTube?  Was that audio a lossless format?  Did the convention into a YouTube format  create a lossless audio playback such as FLAC?

 

was the mics used able to capture the full range like a dynamic condenser mic or is it a tiny lapel mic? 

 

 

 

Most YouTube videos are encoded with 128-384kbps for the audio (mono-stereo) unless 5.1 surround sound which is 512kpbs according to YouTube’s data.

 

but hey, I guess moving beyond “big file move over this time” was to much to ask people to understand that there’s way more to audio quality on playback that the “bitrate of video”.  
 

 

Lavelier mics tend to do something like 50 Hz to 16ish kHz, so could very well be the reason.

[Sqg555]

1 hour ago, ImorallySourcedElectrons said:

Lavelier mics tend to do something like 50 Hz to 16ish kHz, so could very well be the reason.

It's probably better phrased that lav mics are "flat between 50 and 16kHz". There is still quite a bit of information well above 16kHz from them, even though it's a bit quieter. It just gradually rolls off at 16k.

 

The brick wall cutoffs in the spectra aren't mic-related. They're undisputably LPFs from sample-rate-conversion or encoding.

 

I don't think there's much mystery here - a configuration got messed up somehow and it wasn't drawn to the attention of any of the LMG/Floatplane staff who are audio experts. 

[ImorallySourcedElectrons]

1 minute ago, Sqg555 said:

It's probably better phrased that lav mics are "flat between 50 and 16kHz". There is still quite a bit of information well above 16kHz from them, even though it's a bit quieter. It just gradually rolls off at 16k.

 

The brick wall cutoff in the spectra aren't mic-related. They're undisputably LPFs from sample-rate-conversion or encoding.

 

I don't think there's much mystery here - a configuration got messed up somehow and it wasn't drawn to the attention of any of the LMG/Floatplane staff who are audio experts. 

Yes and no, the wireless transmitters are typically the limiting factor, not the microphone itself. But it also depends a bit on if it's digital or analog transmission.

 

This is what my analog Sennheiser XSW (that's supposedly also limited to 16 kHz) does when I blow into a lavalier microphone:

 

But I've seen nice steep cut offs on digital ones, so it'd actually be interesting to see if it was caused by the source file or not.

[DoctaJones81]

Just to have the content here as well: From Reddit: Luke_Lafreniere's comment on "Linus and Luke lied about Floatplanes

[LAwLz]

7 hours ago, Thepyrodex said:

So how much of that data is audio?  Well assuming 5.1 surround YouTube says that’s about 512kbps, so that’s 42% of the 1200kbps you are saying at 1080p in your math.  Do you think video streaming platforms prioritizes audio or video when forced to choose what goes through when most consumers are on small or shallow speakers from laptops/tv, or on smaller headphones that still have a limited range based purely of driver size limitation (that can be overcome but cost $$$ that the AVERAGE user will never spend) 

 

 What level/type of audio quality was encoded into the video?  
 

what was the sampling rate of the recorded audio?  What is bitdepth of said audio?  What format was it encoded into the video before uploading to YouTube?  Was that audio a lossless format?  Did the convention into a YouTube format  create a lossless audio playback such as FLAC?

 

was the mics used able to capture the full range like a dynamic condenser mic or is it a tiny lapel mic? 

 

 

 

Most YouTube videos are encoded with 128-384kbps for the audio (mono-stereo) unless 5.1 surround sound which is 512kpbs according to YouTube’s data.

 

but hey, I guess moving beyond “big file move over this time” was to much to ask people to understand that there’s way more to audio quality on playback that the “bitrate of video”.  

Dude, what are you on about? 

I feel like you must have replied to the wrong person or something. 

 

What is your point and which one of my points are you specifically trying to refute? 

 

By the way, the YouTube sido is nowhere near 512Kbps. Nor did I even mention the audio in YouTube videos. I was talking about the file size of the Floatplane video someone downloaded. Someone said it couldn't possibly be so low as ~1Mbps but very basic math shows this to be true. 

 

 

Anyway, now that it's confirmed to have been an issue that they realized about 4 months ago maybe their rabid fanboys will stop trying to say this isn't an issue and that people speaking up are just making things up.

 

Hope they fix it soon. But then again, since so many people seem to defend them maybe they shouldn't fix it at all. Clearly people don't care and will happily not just pay but defend them regardless of for example false claims about quality. 

 

Something I'd like to point out that it's not very transparent of them to hide this issue for 4 months. If they knew about it then they should have informed users about it, if they really want to be transparent. 

[alelo]

13 hours ago, RamblingPenguin said:

I managed to get a shot with some motion in it, https://imgsli.com/MTk5MzEz 

It still looks pretty much identical, it's possible that the video they got the shot from was already compressed though. 

you can see that FP is slightly sharper, esp if you focus on the nose there it is most noticeable

[8tg]

17 hours ago, AHPanda said:

That was +90 years ago, nothing changed.

I bought a bunch of led bulbs in 2010 and have been using them since, thank you. Mostly because I bought them at $50 a piece and I’m not throwing them away until they cease to be repairable.

And within the last few years I think this has become pretty common, who’s buying incandescent lamps?

 

anyway, off topic

Anyone who bought into floatplane deserves to be scammed, we all knew it wasn’t even going to come close to its competitor(s)

[HenrySalayne]

16 hours ago, DoctaJones81 said:

I'll also say this, spoken audio doesn't need a much from a bit rate level as a full music bed. So this would be similar to the video comment I made above.

This is not true and it's a very common mistake. People confuse quality and intelligibility. The later is measured with the speech transmission index (STI). Since even low bitrate codecs offer good or excellent intelligibility, people came up with the idea that speech "needs lower bitrates". While speech might be intelligible with lower bitrates, the loss of audio quality is still noticeable.

If we take a look at the bandwidth human speech occupies, it ranges from around 80 Hz to 10 kHz, while music generally tends to be somewhere between 40 Hz and 16 kHz. This might seem to be a huge difference but it's actually only a change from 8.5 octaves to 10 octaves. So music requires ~20% more bandwidth than speech. For all practical purposes, this difference is irrelevant.

 

 

[HenrySalayne]

16 hours ago, wanderingfool2 said:

I started reading the guys post because I was curious...I think he must have downloaded the Floatplane clip wrong or something seems highly suspicious about how he's analyzing things.

 

His claims is that the Floatplane 1080p clip ran at 1,093 kb/s...that's quite low, especially if it's using a crf it would be noticeable in videos.  Given that I haven't really seen too many people complaining I think that he might have messed up how he downloaded the floatplane clips

I downloaded a few videos in 1080p and came to the same conclusion dividing the file size by the length of the video.

The recent "buckling spring keyboard" SC has a filesize of 150,889 KB and a runtime of 13:32 or 812 seconds. So we are talking about ~ 1480 kbit/s combined video and audio.

The aforementioned 7950X3D video has a filesize of 142,909 KB with a runtime of 15:38 or 938 seconds. So we are talking about ~ 1210 kbit/s combined video and audio.

This is absolutely in line with reddit's findings.

 

[AHPanda]

3 hours ago, 8tg said:

I bought a bunch of led bulbs in 2010 and have been using them since, thank you. Mostly because I bought them at $50 a piece and I’m not throwing them away until they cease to be repairable.

And within the last few years I think this has become pretty common, who’s buying incandescent lamps?

 

anyway, off topic

Anyone who bought into floatplane deserves to be scammed, we all knew it wasn’t even going to come close to its competitor(s)

I was referring to customer reactions not changing.

[exsin]

22 hours ago, AHPanda said:

Also, they don't charge VAT (which is mandatory) on Floatplane subscriptions in Europe.

 

LTT support (I will play nice and not show the mail thread) points the finger at Shopify when I complained about the full and utter lack of 3D-secure verification, which is mandatory in 45 countries, including all of Europe. 

 

I have since lodged a complaint with my banks' security department about this, who are taking it immensely seriously. According to European law, you do business with LMG/Creator Warehouse when you are buying merch, not Shopify. The retailer is ultimately responsible to adhere to all processes required by law in the country he chooses to sell his products in.

 

Good find about the VAT, I will lodge a complaint with the tax authorities in my country too, that's amazing @AHPanda.

[Kiina]

11 hours ago, DoctaJones81 said:

Just to have the content here as well: From Reddit: Luke_Lafreniere's comment on "Linus and Luke lied about Floatplanes

Decent response.

 

"First, we think our audio transcoding settings were copied from our target for 360p video to all targets by accident sometime in the past when the transcoding stack was completely rewritten from scratch and this change wasn’t noticed"

It always seems like a rewrite is a good idea, yet is often isn't. But it's kinda weird that they don't have any unit tests or even a manual audit to see if everything is working as intended especially after such a big change. Kinda crazy to rewrite the entire transcoding stack and have someone just go "LGTM" and merge it.

[LAwLz]

4 minutes ago, Kiina said:

Decent response.

 

"First, we think our audio transcoding settings were copied from our target for 360p video to all targets by accident sometime in the past when the transcoding stack was completely rewritten from scratch and this change wasn’t noticed"

It always seems like a rewrite is a good idea, yet is often isn't. But it's kinda weird that they don't have any unit tests or even a manual audit to see if everything is working as intended especially after such a big change. Kinda crazy to rewrite the entire transcoding stack and have someone just go "LGTM" and merge it.

What bothers me about this response is that they themselves noticed it 4 months ago, and didn't think to notify their customers about the degraded service. 

Going "Oops, sorry we'll fix it in the future" once caught is not a good look for a company constantly talking about how "transparent" they are. 

 

It's good that they are planning to announce these types of issues in the future to the public, but it's not like they couldn't just have sent out an email or put up a notification when the issue was detected and identified. They chose not to use the tools they already have to inform users about this issue. They only announced it after they were caught.

 

I hope a similar thing doesn't happen again.