Cloudflare

screenshot-2022-10-05-at-13-29-52

on another note though stream really needs to show errors better it only displayed the error that it couldnt play in the console lol

unless they arent using the stream player

unknown

when i try in Edge without the codec i get this in console

that demo doesnt use the stream player though, yeah

Stream player doesn't support the AV1 beta yet

> To playback video using AV1, you must use your own video player. Read the docs on using your own player, or see our examples of using Stream with the most commonly used video players.

they have in their demo page ```html
        <script>
            const obj = document.createElement('video')
            const canPlay = obj.canPlayType('video/mp4; codecs="av01.0.05M.08"')
            const content = '<div class="center"><p>Your browser does not support AV1</p></div>'
             if (canPlay === ""){
                document.getElementsByTagName('body')[0].innerHTML = content; 
                console.log('dddd');
            }
        </script>
```

"probably" dev was too lazy to actually find out

We're working on supporting this use case, and making it possible to stream without recording. Expect an update on this before the year ends.

Only live streams and recordings of live streams today while in beta, but we're working on adding support for videos uploaded to Stream as well. More soon!

Is it plan to support VP9 or HEVC ?
- VP9 is great on android devices that not support AV1
- HEVC is the only way to stream properly on Safari

You were ahead of us by a day with your comment about this yesterday 😄 

Right now, we're focused on AV1, and are optimistic about future support in the browsers and devices that don't yet support it. We're always considering new codecs and protocols we could support down the road though.

It's amazing how much you cut costs compared to similar services 👌

What do you use to make your AV1 encodes ? SVT-AV1 ?

We have dedicated hardware encoders that are engineered and optimized specifically for AV1 encoding.

SVT-AV1 is a great option for CPU-based encoding — we needed even faster performance, and went the extra mile to install hardware in our own data centers to make sure we can encode in real-time.

I'm really excited about some of the consumer graphics cards that can support AV1 encoding too — the early tests show that Intel's Arc series is particularly efficient

https://www.tomshardware.com/news/intel-arc-av1-encoder-dominates-nvenc

End-to-end streaming using AV1 the whole way is an exciting future to look forward to 😄

thanks for the update. this will be helpful, for the PTZ cam use cases

I'm on the same boat btw 😉  Hope that feature lands very soon!

Testing out AV1 via HLS. Can someone verify how it's working?
https://hls-js.netlify.app/demo?src=https%3A%2F%2Fcustomer-g4cqjmumd7qeeopb.cloudflarestream.com%2Fa458d58997f542ea07c9d87e4d5e07e4%2Fmanifest%2Fvideo.m3u8%3FbetaCodecSuggestion%3Dav1&demoConfig=eyJlbmFibGVTdHJlYW1pbmciOnRydWUsImF1dG9SZWNvdmVyRXJyb3IiOnRydWUsInN0b3BPblN0YWxsIjpmYWxzZSwiZHVtcGZNUDQiOmZhbHNlLCJsZXZlbENhcHBpbmciOi0xLCJsaW1pdE1ldHJpY3MiOi0xfQ==

SRT 19mbps H.264 -> CF Stream -> HLS with AV1 opt-in

From my perspective (lol im streaming on Wifi via M1 mbp YOLO) AV1 + OSX + Chrome seems phenomenal

looks like the live broadcast ended, can I have your consent to view the latest recording on that live input? (Just to be safe, customer content and all that 🙂 )

Okay I'll keep the test stream running for a few days 😉 (hope bills wouldn't kill me)

interesting, for me trying to view it,  regular viewing or with the AV1 opt-in doesn't playback nicely at all. The Apple hardware encoder in the setting may be an issue, could you try swapping to the software x264 encoder and see if that improves it?

Just restarted the stream with x264 CPU encoding. It struggls a bit on my M1 Mac Mini though i think it started working again

ah, I see it working. The performance for the individual segments looks much worse than we aim for, especially for av1. I'll share the specific video with the team so we can look into what's up.

Feel free to delete the video at any point, I'll be grabbing some of the original data so we can replay it through our encoder to test plus a snapshot of the existing metadata.

Thank you very much! However I'm going to stream https://youtu.be/EYVvoLCQ_2g very soon. This won't go through CF Stream but we're aiming for, once it's ready for the prime time ❤️

👍  totally reasonable, excited to get av1 to a place where it works great for y'all 🔥

By the way, I'm deeply interested in providing multilingual **live** captions. Can anybody shed some light about the 'status quo' of the Live WebVTT tech stack? Even Youtube does not provide users a realistic way to inject a multilingual live WebVTT translation/transcription other than extremely limited and difficult protocols like CEA-708. Hmm..

This is something we're interested in providing, and share the sentiment about CEA-708 being tricky to work with (for both us and for customers).

Curious what your workflow would look like for **generating** live captions? Can logistically tricky to generate on-the-fly. Using an automated real-time transcription service/API?

Would love to learn a bit more — there are some use cases (a live conference with people talking) where automated transcription and translation can work quite well, and others like live multilingual captions for a concert where auto-generated captions and translations really don't fit.

In our CURRENT workflow, a translator live-translate & transcribe what is going on the stage just like in a discord message. A message would have a UTC timestamp so that it could be offset by a few seconds ago to mitigate the time it took to type in the translation. It should be able to reach WebVTT generator BEFORE users reach the HLS manifest so we inject segmented VTT files into the .m3u8 file. It technically works, but is VERY clunky and I am not proud of it.
One major issue that prevented us from CEA standards is the fact that we often provide Korean/Japanese/Chinese translations and I don't have any access to such hardware/software to transmit an RTMP/SRT/ whatever... to the cloud.
At the moment I could not find any software that would reliably cover our situation. I am REALLY into this and EAGER to tackle this issue once and forever.

At the moment we just 'overlay' (open-caption) the translations and then render 2+ different RTMP/SRT signals, one per language. This .. well, works, but I think this is very resource intensive and inefficient.

Oh btw our use-case mostly focus on a live concert and a few fan-meetings.

https://youtu.be/EYVvoLCQ_2g This stream will demonstrate the "workaround" we're using. (will go live on 7PM KST)