![[$20 / $100 per week] TankieTube's first fundraiser/pledge drive! 🪙 📺](https://hexbear.net/pictrs/image/bef5a8cf-9026-43ef-875d-119f17a7b479.webp){kind=link}
Note: I may be slow to update the post title with the current goal progress, but these badges should be automatically updated [Edit: Hexbear proxies images so I have to make them hyperlinks instead].
Receives X USD | Goal Progress Y %
TankieTube is holding its first ever fundraiser! 
If you've found the site of value, then please consider becoming a sustaining member through Liberapay (if you have the means to do so)! 
It's our preferred method.
https://liberapay.com/TankieTube/donate
We've set an ambitious goal of $100 per week, which would cover much of the current server, bandwidth, and storage costs.
One-Time Donations
Liberapay currently doesn't support one-time donations, but you can effectively make one by signing up and then cancelling your subscription after the first payment.
Other ways to make a one-time donation, for your convenience:
1)
Show
PayPal/Venmo/Credit/Debit [No anonymity]
Pros: Simple, fast, no account creation required
Cons: Your name and address is revealed (to me); it's not private and anonymous like Liberapay
2)
Show
Monero (XMR) [Ultimate OpSec]
8AaqRrPGfzPfGmZbmBc43WQ4BEnTBujbKiNrMcCZWXKo84LfhH8hysNZxV6JhvESmp2XueM4nHg9J6tLf6pKpbH4KH6dTR4
Pros: Untraceable: Neither I nor Liberapay will know your identity. Can bypass US sanctions which affect PayPal and Liberapay.
Cons: Complicated
The majority of existing cryptocurrencies, including Bitcoin and Ethereum, have transparent blockchains. Transactions can be verified and/or traced by anyone in the world. This means that the sending and receiving addresses of these transactions could potentially be linked to real-world identities.
Monero is the only major cryptocurrency where every user is anonymous by default. The sender, receiver, and amount of every single transaction are hidden through the use of three important technologies: Stealth Addresses, Ring Signatures, and RingCT.
If you buy Monero on a public exchange such as Kraken, then transfer it to a private wallet you control before sending a payment.
I've considered GPU acceleration, but I've read that software is vastly superior in terms of quality for encoding video, and so for a video archival website that made more sense to me. I haven't personally done any benchmarks or comparisons though.
I plan to add my home server as a remote runner for more transcoding power. I used it briefly in the past but encountered a technical bug (it may have been fixed in a recent software update). A downside to using remote runners is that it bloats the network traffic[1], so it makes sense to make the localhost as powerful as possible.
If someone uploads a 1 GB video to the server which gets transcoded locally, it consumes 2 GB of traffic once the files are uploaded to object storage. If a remote runner is used to transcode the same video, the server hands it off to the remote and receives it back again, doubling the traffic to 4 GB ↩︎
Is there any way you could make it user selectable? Or grant it on a per user basis based on needs? I suppose not.
It's just I don't think movie watch night uploads of Batman and MLP really need the same care as uploads from combat zones or from resistance fighters or communist content creators and re-encoding them using GPU acceleration makes sense as long as you use a reasonably slow preset (still very fast on gpu acceleration).
Both Intel QSV and NVENC are quite good and even AMD's implementation while considered behind and inferior to those too wouldn't necessarily be so bad that most Hexbear movie watch night users would necessarily notice given we're talking about a less than 800 pixel window on the side of cytube on 1080p monitors. Though I admit my only experience is with NVENC, QSV, and software encoding.
Warning technical rambling
For archival purposes I would definitely recommend software encoding where possible. I use it myself for my own collection of definitely legally acquired major films and TV shows because it results in:
And because I'm re-encoding from already lossy encoded bluray and at times web type sources reducing the amount of degradation via repeatedly encoding something is important to me because I'm watching it on a big TV.
Though when we're talking about software encoding generally there's a big difference between medium/slow and slower than that presets and things like fast, ultra fast, etc. Medium and below, especially slow really get you the best results with drastically diminishing returns for slower, very slow, placebo settings. They're the best balance of quality speed settings which is why many quality p2p type groups use them.
If you're doing very fast encoding using software then at that point the difference between that and a slower preset run on NVENC or QSV (intel) isn't a very wide gulf as faster presets are also less file size efficient compared to slower ones and lower quality.
I think personally if I was in a situation of needing to do fast or especially higher than that (faster) encode presets via software that I would consider whether I couldn't for no additional money or less money switch to something that can utilize NVENC at slower preset. Obviously keep in mind storage costs and test to see what file sizes you get out of a few different types of test files to see if it might represent in the long-run a storage issue before making a change like that.
It would require writing a custom plugin for PeerTube. It's something I'm very much interested in doing, but it's slightly above my current skill set.
PeerTube uses the
veryfastffmpeg preset which leaves a lot to be desired IMO, so you may have a point about NVENC or QSV. I'm an AMD fan though 😕.If you didn't know plugins can add transcoding options, so of you want you can have a slower preset
ⓘ This user is suspected of being a cat. Please report any suspicious behavior.
I use remote runners (just the localhost at the moment) and they're configured slightly differently, but I know it's possible to change their settings as well.
The issue is I don't want to make the preset any slower until I have the option for users to choose whether they want to prioritize speed or archival quality. If I switched to
veryslowglobally, the transcoding queue would bloat.Look at this shit
Only configed to do 480p and original right...
yet I also have 720p!!!
How? You already know, apartheid
ⓘ This user is suspected of being a cat. Please report any suspicious behavior.
What instance is that?
LOL, apartheid. Does that plugin let users select a custom CRF or preset? Or just resolutions?
This is my local test instance, with my own custom plugin. It doesn't do a whole lot right now, but I'm prototyping and learning the system. I realise that resolutions is the wrong thing to change, and I should look at the CRF stuff
ⓘ This user is suspected of being a cat. Please report any suspicious behavior.
So you wrote that? Sick! We should combine heads on this later.
Sure!
Also... it seems hooking into the transcoding options and "apartheiding" (i.e. using different setups based on user) is going to be harder compared to the resolutions... but it might still be possible, if through a more cursed way.
ⓘ This user is suspected of being a cat. Please report any suspicious behavior.
My knowledge of hardware encoders is a few years old. I don't know if AMD has caught up. I know Nvidia was leading the pack with NVENC for a while but Intel caught up with QSV which is very close in quality.
I guess you'd have to decide what you value and if this would save you significant money or not and what trade-offs if any there are.
Also I just checked and it seems if you're streaming AV1 the difference shouldn't matter between Nvidia and AMD. The difference was really in lower bitrate 264/265 stuff. As AV1 is open I guess that makes sense. So AMF (AMD's answer to NVENC and QSV) should be fine if you're just doing AV1. Obviously still worth doing some comparison testing with fast action clips, animation, etc.
PeerTube uses x264 by default.
I want to switch it to AV1 (libre-dav1d) when I get a chance, although I have slight concerns about device compatibility.
Yeah that is an issue but mainly for older Apple devices.
Don't suppose it can do some kind of fallback? That would be too clever.(Oh wait yeah no, it transcodes once on upload and it would be wild to store 2 copies)Apple unfortunately only added HW decode support in 2023 (which is why I doom about AV2 because this takes a ridiculously long time). iPhone 15 pro and 16 and on standard all support it.
Android/Google added it in 2019 so compared to Apple most Android users should be fine.
Apple's support is also kind of fuck-y in that I don't think it works with normal containers only HLS like they have some real hatred of anyone but official streaming providers using it.
On the other hand most people with an iPhone/iPad probably have a computer as well they can use and any computer made in the last 15 years can probably software decode AV1 just fine.
Do the Apple devices fallback to software decoding?
I dont think so. Mdn says only hardware decoders.
ⓘ This user is suspected of being a cat. Please report any suspicious behavior.
From what I remember NVENC quality was behind QSV, but that could have changed since. With the testing I did, I didn't notice any quality issues, but maybe I'm just not too picky about it