We built Welder for ourselves as we are a small product team that was forced to go remote and wanted to produce content.
We used Zoom, but the quality was.. Well, S*t.
It’s a super easy way to record full quality recordings remotely as each participant is doing local recording which is progressively uploaded to our servers so as soon as the call ends, it’s ready for editing.
We’ve validated the product in 2 weeks by doing LP and running YT ads. Built MVP in 1 month and found out no one used it as we did a mistake by not validating if Electron app is ok for people to download.
Lessons learned, Welder is now browsers based. We decided to stay bootstrap and right now we are trying to find a pivot from a tool that “Allows you to record an interview in HQ remotely” to a tool for a team like us to produce meaningful content with ease from start to end.
I would love to hear your feedback either about the tool or about your content creation process. As we are doing extensive research on the topic of how startups and product teams produce content.
P.S. We have now quite a lot of experience with running online video service in production for some time, so if you have any questions regarding this topic, let us know we will be more than happy to answer.
It confuses me that you used a video with that much high-pitched background noise in the audio to showcase your tool - in the end, both version seem a bit low-quality because of the noisy audio.
It's true! This was done with an early prototype of the Welder so the quality is not the best. We should've reshot it already, but we love the vibe of it, haha.
> I was wondering where this product was, and/or why some of the podcasts I listen to don't use something like this.
These are called "multi-enders", and lots of podcasts do them. There are many[1] tools for this which are focused specifically on podcasters.
Additionally, it's straightforward to do it manually as well. (Tip: Have participants on all ends do a "1-2-3 <CLAP>" to make it easy to align the audio during production. Do this at the beginning and end to account for digital clock drift.)
Most of the higher production value podcasts I listen to with remote hosts/guests do this, albeit manually (ie record locally with something like Audacity, sync on a click/clap, recorded live call as a backup). This feels like a nice improvement to streamline an otherwise cumbersome process.
We store unuploaded chunks of the video in IndexDB in a browser. Btw if plan to do something similar beware of the fact that storing a lot of blob references in IndexDB can cause some wired issues in Chrome a some blobs can get corrupted. Our workaround is to store it as byte arrays.
Video is saved as a binary blob in some kind of storage. In our case that would be IndexedDB. But since the data is uploaded in parallel with the call there is only few hundred MB used at any give moment.
Generally browsers will let you store a few GB of data (depending on the user free space) if your really wanted to but it's not recommended since they can delete it any time.
Congrats on the release and the excellent product! I work on a Remote Desktop app and one of my customers used Remote Desktop + your tool to collaborate live with his partner.
Would you be able to talk a little more about your Webrtc stack? Do you use SFUs in your stack?
Thank you! We use Twilio for our WebRTC infrastructure with VP8 simulcast setup so we do use SFU but only through Twilio. We were considering how to get rid of Twilio and quite extensively researching what would it mean to run our own WebRTC infra as Twilio costs are pretty high (sometimes up to 1/4 of our revenue). But operating multiple servers around the word to act as SFUs or creating them on demand in e.g. GCP and paying the cloud tax in form of extremely expensive egress costs doesn’t seem to be the right solution either. Other service providers have very similar pricing, that's why we will stay with Twilio for now.
Thank you for the info! Incase it helps, we’ve been running our own WebRTC infrastructure (TURN servers only at the moment) for the last 5 years. It’s globally distributed, fault tolerant and allows us to mitigate the high costs. We’re doing 10s of TB of traffic at a fraction of the cost. Would be happy to discuss our infra incase it helps.
That would be super helpful to us. Can you please hit me at [email protected]? We could schedule a call or just text chat about it if that's ok for you.
Audio is recorded in HQ too, thanks for the feedback, seems like it's not clear enough.
About the phone on linux - it will work if there is any solution out there that allows you to use your phone as a webcam on linux, Welder will pick that up.
It's actually quite simple. In a addition to a web call we record a separate high quality tracks for both audio and video locally.
The challenge was doing all of this in browser since the APIs are not really consistent + there are a lot of weird bugs and issues with both MediaRecorder API and IndexedDB.
We spend quite a bit of time setting up logging for client app. Since most of these weird issues are very hard to reproduce. (For example Chrome randomly purging binary blobs after storing them in IndexedDB)
I think in your product descriptions you should consider “conversations” as a more generic, friendly, and open ended alternative to the word “interviews.” Many of these podcasts are presenting the experience as a conversation, not an interview, even if the traditional old fashioned word for what they are doing is interview.
It would also remove the negative reactions (my own included) arising from confusion where people worry that your product might be intended for job interviews.
I use Riverside nearly weekly for my podcast, designbyus. I do like it, but I think the biggest opportunity for solutions in space is to reduce 'jank', and make the recording process as transparent as possible.
We once recorded an interview in Riverside and had mistakenly added the guest as a 'Producer', and not as a 'guest', meaning it did not record one stream of the audio. Anything the UI can do to essentially scream in my face 'this is recording, and everything is ok' is what will build trust in products like these.
Thanks for this valuable point of view!
This is exactly our focus right now. The next two weeks we are doing a sprint called "Improving reliability by UX" where we are going to do exactly what you said.
Good question! We are quite similar. The main difference is, I believe in the attitude. We focus a lot on the ease-of-use and reliability and not so much on adding robust features that are requested by B2Bs like podcasting studios or media.
I also believe we are heading in different directions as we want to shift from being just a tool that does HQ recording remotely to start-to-end solution to help startups and product teams produce purposeful content with ease.
We believe there is a lot of space for an improvement for businesses in term of showing who is behind the product, building empathy and educating their customers. The process of putting out meaningful content and building brand love is cumbersome right now and takes a lot of energy.
Yeah my first reaction reading the title was job interviews, but then the site made it clearer. Maybe worth updating it to say “podcast interviews” or “webcast” (not sure if it’s a thing)? EDIT: videocast!
So, that's another take-away lesson then: how can you make this point clearer to your audience? I, too, thought you were targeting job interviews - if it's a common misunderstanding, it might really hurt you.
We built an (internal) product back in March, based around vmix and webrtc, for doing interviews for braoadcast purposes.
Guest(s) get a link, interviewer gets a link, producer can watch the whole thing and provide comms to the right people at the right time, combined and ISO feed recorded separately and dropped into package, fairly reliable although occasionally get issues with some guests being unable to click "Allow camera" or whatever in chrome.
Wouldn't have even considered using it for interviews, but I guess that's because our minds tend to work on a "see problem, build solution" rather than "build solution, expand market"
Content creators in general and right now podcasters, videocasters, media, podcasting studio and others who want to pre-record content and distribute later in best quality possible.
We found out that the quality of content is not only nicer to watch or listen to, but it actually makes the content easier to consume and digest.
We built Welder for ourselves as we are a small product team that was forced to go remote and wanted to produce content.
We used Zoom, but the quality was.. Well, S*t.
It’s a super easy way to record full quality recordings remotely as each participant is doing local recording which is progressively uploaded to our servers so as soon as the call ends, it’s ready for editing.
We’ve validated the product in 2 weeks by doing LP and running YT ads. Built MVP in 1 month and found out no one used it as we did a mistake by not validating if Electron app is ok for people to download.
Lessons learned, Welder is now browsers based. We decided to stay bootstrap and right now we are trying to find a pivot from a tool that “Allows you to record an interview in HQ remotely” to a tool for a team like us to produce meaningful content with ease from start to end.
I would love to hear your feedback either about the tool or about your content creation process. As we are doing extensive research on the topic of how startups and product teams produce content.
P.S. We have now quite a lot of experience with running online video service in production for some time, so if you have any questions regarding this topic, let us know we will be more than happy to answer.