Opus 1.4 Royalty-Free Audio Codec Released

[phoronix]

Phoronix: Opus 1.4 Royalty-Free Audio Codec Released

Opus 1.4 is available today as the first update in four years to this open-source, royalty-free versatile audio codec...

https://www.phoronix.com/news/Opus-1.4-Released

[Anux]

Nice!
Does anyone know of a blind test in the last few years? Especially under 96 kb/s? I only know of this list: https://wiki.hydrogenaud.io/index.ph...istening_Tests

[anarki2]

It's really sad how Opus got neglected over the years. Everyone's pushing their own crap instead. Not surprising, of course.

[espi]

Opus is amazing. I can't believe it hasn't replaced everything else by now. You would think that streaming services would jump on it? it could halve their bandwidth costs easily.

Youtube picked it up at least, but the audio costs are probably nothing compared to the video costs.

Opus also finally forces everyone to be in the same page when it comes to sampling rate, I'm so tired of resampling by now...

[Quackdoc]

seems like there was some efficiency losses, in exchange for much greater reliability? I have yet to test it much, but it seems like encoding music may be better done on 1.3. that being said, bluetooth might be nicer now.

its also worth noting, as far as I can tell, FEC is still only availible on silk, the ultra low bandwidth setting, not CELT, which is used for medium to high fidelity encoding

Nice!
Does anyone know of a blind test in the last few years? Especially under 96 kb/s? I only know of this list: https://wiki.hydrogenaud.io/index.ph...istening_Tests

I always recommend doing your own and deciding for yourself, I think foobar2k has a plugin for this

Opus is amazing. I can't believe it hasn't replaced everything else by now. You would think that streaming services would jump on it? it could halve their bandwidth costs easily.

Youtube picked it up at least, but the audio costs are probably nothing compared to the video costs.

Opus also finally forces everyone to be in the same page when it comes to sampling rate, I'm so tired of resampling by now...

most streaming services do use opus, they also have support for AAC which is about as good as opus at the medium to high fidelity range. for the ultra medium to low range you have xhe-aac, but you still have opus there too.

[microcode]

It's really sad how Opus got neglected over the years. Everyone's pushing their own crap instead. Not surprising, of course.

What do you mean by "their own crap"? What other codecs have been deployed in any major application since Opus 1.3?

[hajj_3]

Opus is amazing. I can't believe it hasn't replaced everything else by now. You would think that streaming services would jump on it? it could halve their bandwidth costs easily.

many streaming services like netflix use xhe-aac as it is superior to opus.

[Mar2ck]

What do you mean by "their own crap"? What other codecs have been deployed in any major application since Opus 1.3?

In the Bluetooth audio world we have SBC, AAC, aptX, LDAC and now LC3, as far as I'm aware they all require paying royalties to use. Opus is high quality, low latency and royalty-free so you'd think it would be perfect but it's yet to be adopted into the a2dp standard. The Bluetooth Special Interest Group has actually tried to justify Opus's exclusion by claiming that LC3 performs better in tests, even though those tests were done with voice samples and they only tested against Opus-Celt and not the part of the codec specifically designed for speech: Opus-Silk.

[microcode]

In the Bluetooth audio world we have SBC, AAC, aptX, LDAC and now LC3, as far as I'm aware they all require paying royalties to use. Opus is high quality, low latency and royalty-free so you'd think it would be perfect but it's yet to be adopted into the a2dp standard. The Bluetooth Special Interest Group has actually tried to justify Opus's exclusion by claiming that LC3 performs better in tests, even though those tests were done with voice samples and they only tested against Opus-Celt and not the part of the codec specifically designed for speech: Opus-Silk.

Ah, I mean; there may be other reasons: SBC, SBC-HQ, and aptX are all rather unconventional to begin with, because of perceived benefits of these approaches on realtime and resource-constrained devices. I was looking at Opus as an A2DP codec years ago (never quite got it working on a device), but now as far as I'm aware it's actually at least specified (even if it is a ‘vendor’ mode, which doesn't really matter).

I agree that there's not any actual practical barrier to Opus on A2DP, and I share the disappointment that LC3 was adopted by Bluetooth SIG (unfortunately the subcommittees that decide this stuff are not visible to Bluetooth SIG members by default, so I have NFI what their internal discussions around this were like).

RE Silk vs. CELT, the Silk mode is not applied to fullband audio regardless, so if it was fullband speech then the Silk mode would not be applicable either way. In hybrid mode, I'm not really certain it would win in those evaluations, everything is up for improvement.... not that it really matters, since A2DP and even HFP are not generally bandwidth limited to anywhere that it would matter which was more efficient.

[Quackdoc]

many streaming services like netflix use xhe-aac as it is superior to opus.

I wouldn't say it is IMO, it is better at ultra compressed, but for higher fidelity streams they either use aac-lc or opus.

In the Bluetooth audio world we have SBC, AAC, aptX, LDAC and now LC3, as far as I'm aware they all require paying royalties to use. Opus is high quality, low latency and royalty-free so you'd think it would be perfect but it's yet to be adopted into the a2dp standard. The Bluetooth Special Interest Group has actually tried to justify Opus's exclusion by claiming that LC3 performs better in tests, even though those tests were done with voice samples and they only tested against Opus-Celt and not the part of the codec specifically designed for speech: Opus-Silk.

both pipewire and google have their own implementation of opus bluetooth, and to be clear, I do think that opus probably isnt what most people want, since it does seem to chew through battery life pretty fast, LC3plus gets pretty close to being "HI-FI" audio when at typical BT bitrates, and it can be hard to distinguish between it and opus. that being said opus does win out on audio quality, another win for opus by a significantly large margin is stream durability, with opus' apparently superior PLC, opus via bluetooth is significantly more resilient then LC3P. it beats every single other codec I have tested by a significant margin, including AAC-LC. a loss for opus on the otherhand is latency, while opus and LC3Plus latencies are well within lipsync, for me LC3Plus latency was consistently lower, that being said opus would be fine for non competitive such as racing games, RPGs etc. LC3Plus in my testing was fine for "non tryhard" competitive games, I was able to comfortably play tarkov and RS6 using LC3Plus in a casual setting (non ranked, not trying hard).

I dont possess the tools to do meaningful power consumption testing, but generally LC3Plus afaik is a fairly less complex codec, and is also fairly low bandwidth, meaning it should be quite a degree more efficient then opus is. My testing was all done on linux via pipewire using multiple different machines and bluetooth chips, some high quality (intel AX200 wifi+BT) and some low end USB dongles from amazon I bought a few years ago.