Replace Skype with Discord for Better Audio Quality and Web-Based Interviews | Hacker News Reader

Replace Skype with Discord for Better Audio Quality and Web-Based Interviews

macobserver.com

98 pointstambourine_man57 comments

Open articleView on HN

Wouldn't secure Wire[1] be a better replacement?

It's crypto-audited[2], open-sourced[3], has native mobile clients[4], Swiss based[5], and pretty well known[6].

[1] https://wire.com

[2] https://techcrunch.com/2017/02/10/messaging-app-wire-now-has...

[3] https://medium.com/@wireapp/wire-server-code-now-100-open-so...

[4] https://github.com/wireapp

[5] https://en.wikipedia.org/wiki/Wire_Swiss

[6] https://news.ycombinator.com/item?id=12148596

been banning mobile for most of my private life, and been using Wire for a couple of weeks now. It's much superior for audio/video quality than Skype. Also lean enough to run on a 5 year old phone.

Compared to Signal / WhatsApp gives me e2e encryption and I can run it also as a Desktop app on Linux (so I can still message my family/kids without having a phone switched on)

Signal doesn't give you e2e encryption?

it's not free?

Wire is free for personal use and paid for businesses[1].

[1] https://support.wire.com/hc/en-us/articles/202856854-What-is...

I saw "request a demo" so presumably it's not free and that's a problem. Also Discord works in most browsers so the person they are interviewing doesn't need to download anything

It's free for personal use, I responded to a nearby comment. In-browser is a good feature, I guess Wire can also get there in future, since both, Discord and Wire, are electron-based apps. But I believe call quality is not quite enough to replace Skype, I demand e2e privacy as well.

wire.com/download for the free/personal use downloadables

I've been listening to enough poor-quality phone/Skype podcast interviews, and figured that someone should build an interview tool that will setup a real time conference, exactly like Skype, but also record the audio in high quality at all remotes and then transfer the recordings afterwards and mix them together with correct timing, so it sounds like all parties are in the same studio. Perhaps even do some magic to mix out the latency, even though there would be some weirdness to work around when participants speak over each other and go "sorry, go ahead".

This is a common way to record remote interviews (since the mid-1900s, I think), and is called a "multi-ender" (or "double-ender").[1]

They're straightforward to do manually. All parties participate in a recorded "1-2-3-clap" at the beginning and end of the call, and the claps are later used for sync.

There are also many "assisted" solutions, including Cast[2], Cleanfeed[3], Ringr[4], and Zencastr[5]. Some folks find them useful, but the additional dependency also means more things that can go wrong. Anybody using them has learned (or will learn) that they must record the live/VoIP portion of the call as well, as backup.

[1] Wikipedia calls this a "phone-sync" (https://en.wikipedia.org/wiki/Phone-sync), which is not a term I've ever heard people actually use.

[2] https://tryca.st/ [3] http://cleanfeed.net/ [4] https://www.ringr.com/ [5] https://zencastr.com/

Excellent, I'm happy (and feel validated) that this exists.

As this would replace Skype (or whatever conferencing tool you're using), it's not really an additional dependency, is it?

> As this would replace Skype (or whatever conferencing tool you're using), it's not really an additional dependency, is it?

It is in the sense that those solutions are not only replacing the realtime part of the call, but additionally must do separate source-quality call recording and post-call file transfer sessions for each participant.

I've lost calls (or would have, if I hadn't recorded the realtime part) with more than one of those solutions because of unrecoverable errors on participants' sides. I'm reluctant to name them since (1) they in turn depend on browsers and their WebRTC implementations, and (2) in my experience different podcasters will swear by and swear at the same solution.

Here's how I'd order the reliability of common remote recording scenarios, from most to least reliable:

(1) Landline-to-landline

(2) Landline-to-mobile or mobile-to-mobile

(3) SkypeOut-to-landline/mobile

(4) VoIP-to-VoIP — Skype, Discord, whatever.

(5) Manual multi-ender — Can be done very reliably if remote participants are reasonably comfortable with recording audio. Still requires a POTS, mobile, or VoIP call for the realtime part.

(6) Assisted multi-ender solutions — Generally way less reliable than manual multi-enders. In my experience and from talking with other podcasters, expect 1/20 calls to fail.

These exist.

What doesn't, and what I'd pay nontrivial (but not vMix Call) money for, is a tool that separated out individual audio and video streams into separate NDI streams for video mixing and composition.

I've found Skype to have a great quality. It's normal phone calls that are terrible, especially mobiles and across continents.

I've Zoom with great results for the past two years at a few different companies

Second this. We use Zoom internally all the time. When a client's Skype/Lync system fails to work, we send out a Zoom link and everyone is like "Whoa!" every time....

We use Zoom weekly for our meetings across the world and it has never failed us!

Occasionally I deal with clients who insist on Skype and it will consistently give me headaches... I can't wait till it dies!

Zoom popped up really quickly and has a lot of adoption already. I honestly don't understand how this space is so competitive given the underlying tech and UX haven't changed much.

After trying every product I could get my hands on, I'm currently using Riot/Matrix for this purpose. All the same I'm still not really satisfied with latency/quality/reliability. Why is this not a solved problem? We've had videoconferencing since at least 1968[0].

[0] https://en.wikipedia.org/wiki/The_Mother_of_All_Demos

It is a solved problem nowadays thanks to the technology behind Jitsi Meet being open and easy to use, but it will take a while until it becomes used in all such products.

Unfortunately, to date, Matrix[1] suffers from a shortage of quality client apps. Riot.im[2][3], while the most advanced among others[4], is far from being convenient. Electron-based desktop client is subpar in terms of UX and performance, whereas the Android client is a disaster for the battery. However, I do hope Matrix will overcome these issues and take it's place as an underlying protocol for the Jabber/XMPP(/IRC) use-cases.

[1] https://matrix.org/

[2] https://riot.im/

[3] https://github.com/vector-im

[4] https://matrix.org/docs/projects/try-matrix-now.html

Disagree about Riot. I find the Android client very similar to WhatsApp and just as good except for a few minor issues: confirming private keys for E2E isn't done automatically as in WhatsApp (arguably better because it's more secure), the quote feature just copies text, and media files sometimes take longer to receive and will re-download. But Riot also has some big advantages like being able to send arbitrary file types and using the system emoji.

Additionally, the desktop web app for Riot is far and away better than WhatsApp Web IMO, which is horribly non-performant, bringing my web browser to a crawl and steadily throttling up my CPU over time. I consider it almost unusable.

As far as battery usage on Android, I do not share your experience. I am seeing 8% of total usage on a phone primarily used for texting and reading emails and news. It's less than the 9% from Google Play services, and slightly more than a newsreader app at 6%. Unfortunately I can't offer a comparison to WhatsApp.

I've never used WhatsApp, can't really compare. But we—team of ten tech-savvy persons—rolled back to Telegram in three days after trying RiotIM. Manual E2E keys sharing in Riot is not perfect, but definitely better than the lack of E2E at all in Telegram, however Linux desktop and especially Android performance was dreadful. At least that was the state at the beginning of the summer of 2017. How long do you use Riot? Can you confirm there have been significant changes since 2017Q2?

The new Skype user interface is a complete abdomination. Even simple things like the “away” status has been removed.

I’m actively looking for a replacement for the company.

Cleanfeed looks like a superior solution for this sort of thing: http://cleanfeed.net/

I haven’t used it, but came across it in https://news.ycombinator.com/item?id=15890816. I haven’t used Discord either.

I’m interested in actually informed opinions on the matter!

This sort of "assisted multi-ender" solution is a good alternative to a pure VoIP call, and elsewhere in the thread I posted links to more of this kind of remote recording solution.

The main problem with them is that they put more of an onus on the interviewee, which often isn't an option with high-value guests.

That, and they don't do video, and they don't do live. Podcasts are progressively moving to both of those (live as a Patreon incentive, for example, or Twitch/YouTube) and tools like these are limited.

There are limits to what external tools can do for video; at some point you either need to be a video mixer (and this you are not doing in a web browser, unless you want people like me to ask "hey, how do I configure interpolation on my HDMI intakes?" and you go "uhhhhhh" because you have no APIs for that), or you need to plug into a video mixer.

I mentioned upthread, but the best option for these cases that I can see is ripping each video and audio feed out of a VoIP solution and sending them to a video mixer via NDI. I'd pay decent money for that.

Hi, thanks for mentioning us (here and elsewhere)

Cleanfeed isn't actually a multi-ender, it was originally developed for live radio interviews. So it aims to give the best quality live and in that regard it's different to, eg. Zencastr which does the recording each side and syncs up the better quality audio later. Each method comes with its own pros and cons; my own experience is also that anything after the interview is not always practical.

'zbuf spams that around a lot, but frankly the audio from it is not appreciably different from Discord and the inability to use it in a DAW makes it, to me, an unserious tool.

Cleanfeed only seems to run on Chrome, no?

Yes. One issue with Cleanfeed and similar solutions is their very narrow and specific browser requirements due to WebRTC dependencies. Now that recent versions of Firefox and Safari support WebRTC, this should change soon.

Firefox got WebRTC support in 2013, about the same time as Chrome. See eg. https://hacks.mozilla.org/2013/02/hello-chrome-its-firefox-c... (though this blog's event was with beta releses of the browsers, stable releases got enabled-by-default WebRTC later in the summer for both browsers)

Zencastr is good for this too - but I'm looking for a video-based one of these.

Does anyone know something that does that

Can anyone recommend a solution that's both open source and works on Linux?

Use a regular phone?

Regular phones are about the worst alternatives. It's not encrypted, the audio codecs are terrible, the input and output of phones are rarely as good as a headset, recording features are limited and the list goes on.

Murmur and Mumble[0]. I have murmur running on a little VM and can handle thousands of gamers. It's not has happy clicky as discord, but it provides end to end voice encryption with great sound quality. You could run it on a RasPi provided you have the bandwidth. Be sure to use their 1.3.0 snapshot. Major improvements on user management, echo cancellation, game overlays, positional audio, etc.

I and many gamers have been quite happy with it.

[0] https://wiki.mumble.info/wiki/Main_Page

2nd Mumble. I've found the voice detection and noise cancellation to be better than Skype or TeamSpeak, possibly because it's more configurable.

We've been using Discord for our podcast for a while now. At this point we're only using Skype for a short video portion of the show.

We stopped using Skype recently for conference calls when they issued a software update that broke merging calls. I guess they just don't care anymore now that Microsoft owns them, they're more concerned with it looking pretty.

Just use Linphone and SIP; it allows to set the quality, the codec, the bandwidth, the resolution for both audio and video, while it's a native build, not electron.

SIP is a disaster. There are so many issues with configuration, firewall traversal, and general user experience that saying end users should be using it is just disconnected from reality. Skype nailed it for a reason—because all the dumb SIP stuff you need to deal with just to get it going is too much. You need to be able to install an app and just go. Especially with interviews and any time you’re dealing with people outside of you controlled little environment.

SIP is successful in the telecom space because it still requires specialized knowledge and dedicated teams of people to run it, and preserving that might be the reason it’s still such a pain.

I think SIP lives in Telco land for a similar reason Cisco does (or Kerberos in enterprises). Yes it's complicated, but that's just a funny way of spelling flexible. Most people just want to make calls and are happy with a locked down one-size-fits-all solution, so they just use Skype and it's good enough. Some people want to tune their own systems to get them closer to just right and aren't afraid of the complexity.

It has options for multiple stream transports, a proper SIP gateway can even allow two totally incompatible clients to talk to each other by relaying and converting the calls. There's no reason half of your call can't be webrtc to someone's browser (asterisk SIP gateway supports this out of the box) or even POTS in an emergency.

Your asterisk solution could include a webpage and webrtc, which would actually be an easier client setup than Skype. No app, just hit this URL and start talking.

The quality of your call is totally unrelated to SIP. SIP isn't the transport, it has no CODECs, you're most likely thinking of (s)rtp. SIP will negotiate the CODEC to use, but that's all.

For audio only nothing beats mumble/murmur. If you want video stuffs, jitsi.

I refuse to use Skype, discord, or any other proprietary black box system for something so important.

No echo cancellation on Mac OS is kind of a deal breaker unless everyone with a Mac can use a headset.

This is no longer the case. Murmur / Mumble 1.3.0 have multiple forms of echo cancellation. It works great.

Oh really? There hasn't been any activity on the bug I was following, but if it has been fixed nevertheless that's awesome! goes point Mac OS friends at the development snapshots

Wish Google would do a web version of Duo. My family has moved to using it for our family video calls at Sunday dinner and works well even with low bandwidth.

If discord doesn't like your browser it asks for your phone number before it will let you use it, even as a guest login. It is not acceptable for you to require your business partners and potential employees to share their personal data with a third party company you have no formal partnership with in order to contact you. Please do not use discord for professional purposes.

Either use something that asks nothing of your guest, such as appear.in, or host your own service where at least you control the data. Jitsi or Lync might be palatable to your IT department.

Does Discord have screen sharing yet?

Yes but having tried it recently it provides very poor quality. The entire time the screen was blurry. Couldn't make out a single piece of text.

From what I understand before they rolled it out to everyone the system worked perfectly so it's just an issue of them scaling their backend.

No idea why they don't just use WebRTC for video&audio though.

I think that's what roll20 uses for it's video and it leads to constant "Oh I can't see you but I can see everyone else" and just constant fiddling to reconnect people.

How recently? We’ve made strides in screen share quality over the past few weeks.

Yesterday. We were unable to look at some code together. We had to sit and wait 2 or 3 minutes before I could start to make out text.

What kind of codec are you using? Are you using Amazon's streaming service or did you roll your own?

I've never had any huge issues with the screen sharing. Being limited to 720p kind of blows, but they have a product to sell. My only complaint is that sometimes when you stop the screen share, it leaks, a process consuming a huge amount of CPU time is left running, and Discord has to be started for it to go away. I'm guessing this will be promptly fixed, but we'll see.

It's only in DMs as well. We record while letting our audience listen and because of that we can't use video.

Reply on news.ycombinator.com