I just stumbled upon this (from HN (https://news.ycombinator.com/item?id=33158475)):
https://www.vectis.com/media/vectis-ip-announces-call-for-patents-essential-to-the-opus-codec (https://www.vectis.com/media/vectis-ip-announces-call-for-patents-essential-to-the-opus-codec)
https://www.opuspool.com/ (https://www.opuspool.com/)
This was dug up after Oracle replaced Opus with Vorbis (https://www.virtualbox.org/wiki/Changelog-7.0) for the new VirtualBox release.
They do this now?
I hope they fail miserably.
Pardon me, did they took Opus and are selling licencing rights? Can they do that?
Pardon me, did they took Opus and are selling licencing rights? Can they do that?
They say Dolby and Fraunhofer have patents they deem applicable to Opus.
This is once again a prime example of a policy of creating Fear, Uncertainty and Doubt (FUD) (https://en.wikipedia.org/wiki/Fear,_uncertainty,_and_doubt). They say they have applicable patents, but they do not publish a list. In fact the page says
Patent essentiality evaluation by independent experts is currently ongoing and a list of patents licensable under the Opus Pool will be published soon.
I read that as: we don't even know whether there are patents that are applicable to Opus. We just like to scare people, so they use someone elses product (which are products by Dolby or Fraunhofer in certain use cases).
It might be this 'soon' will stay there forever. Just give people the idea there might be applicable patents. Can't sue them over patents when it is not known which are meant. Once again, prime example of FUD.
Oracle replaced Opus with Vorbis (https://www.virtualbox.org/wiki/Changelog-7.0) for the new VirtualBox release.
What a silly decision.
Anyway, the patent pool thing has about the same vibe as that AV1 patent pool thing (https://en.wikipedia.org/wiki/AV1#Patent_claims), i.e. it's BS.
Would there be legal ground for suing these scammers for their FUD, then?
Just wondering.
This is so wrong.
MS did the same with ANS.
Same thing will be done with Opus.
2 questions: Which software could be affected by the vague wording? And, do players use native opus player or add own code to support opus on 2018 and 2019 phones?
One of my favorite aps on android app plays opus. The app is largely free, but the paid version offers better file navigation.
So, by the wording, they claim only open source sw won't pay royalties, which is a very fuzzy statement: a commercial sw using the open source decoder might be exempt, or might not; free software is not necessarily open source; lots of free software using opus get paid for value added services.
Plus, I am now clueless if the audio player I use is likely using phone built in opus decoding libraries or included them into the apk.
I don't think the player could play xHE-AAC so I am not considering using it. If opus support is taken out, I am going back to mp3, which is a shame because I need the sub 24 kps range for my thousands of hours of voice recordings.
It doesn't make sense that the press release tries to calm down fears of software fees, if they were only trying to scare people with fud.
Plus, I am now clueless if the audio player I use is likely using phone built in opus decoding libraries or included them into the apk.
Then you probably shouldn't think on it too much. As you say, there are too many unknowns for you to figure this out. Just wait this out, I suspect the chances are quite high opus support will not be ripped out and nothing bad will happen.
It doesn't make sense that the press release tries to calm down fears of software fees, if they were only trying to scare people with fud.
Of course it does. The open-source community, and specifically the part backing opus, might be very much able to fight back. With this statement they might be pre-emptively trying to make lawsuits by the open-source community against them 'inadmissible' on the ground of them being exempt from this patent pool. Just a guess, I'm not a lawyer.
Hardware manufacturers on the other hand, might not care much for legal proceedings and might not fight back. Especially smaller ones. In the case of hardware devices, there is simply too large a risk of the devices having to be recalled and/or scrapped. The chance of a patent 'enduring' might be small, but if it happens, the losses are far too great.
Anyway, three months have passed since the announcement, but no list of patents has been published yet, while the announcement says it is about 'hundreds of patents'. So much for it being published 'soon'.
Has company/project bitten on this FUD-bait and stopped using Opus as a result?
Oracle has, as noted in the first post of this topic.
A quick search turned up this: http://lists.digium.com/pipermail/asterisk-dev/2013-May/060419.html But that was years ago, so it predates this patent pool.
I'd be shocked if they ever actually list any patents publicly since the open source community and even more shocked if they try to go to court over it.
If there are any patents, it's likely that they're questionably valid, questionably applicable to Opus, or both.
Anyway, I think these people benefit more from FUD, say, leading people to go with AAC rather than Opus and license the AAC patents, than actually enforcing this.
License terms have now been published
https://www.opuspool.com/licence-terms
Standard License Fee
License Fee: € 0.15 per unit
Additional Terms for Compliant Licensees(1)
The following discounts are available to Compliant Licensees:
- Annual License Fee Cap: € 15,000,000
- Past Practice: not payable for any Licensed Products sold in the period prior to January 1, 2023
(1) A Compliant Licensee is a willing licensee acting in good faith and in complete and continued compliance with the terms and conditions of Opus PLA (Patent License Agreement).
Terms for Early Licensees(2)
The following additional discounts are available to Early Licensees:
- License Fee: € 0.10 per unit
- Annual License Fee Cap: € 10,000,000
- Past Practice: not payable for any Licensed Products sold in the period prior to the Starting Date(3)
(2) An Early Licensee is a Compliant Licensee that executes the Opus PLA by the Starting Date.
(3) The Starting Date is the later of (i) October 1, 2023, or (ii) within six (6) months from the date Licensee first offers a Licensed Product for sale.
There's also a list of patents covered by that license published.
I have strong feelings regarding this development and I hope some sort of defense strategy can be implemented.
MOD Edit: Paragraph was edited per OP's request.
It seems to me some patents on this list don't even exist?
FR3605534, GB3605534, NL3605534 and IL3605534 don't seem to exist, but EP3605534 does.
Same goes for NL2589046 and IT2589046, they don't exist, but EP2589046 does. GB2589046 is something completely different, not a patent of Dolby, and was only applied for in 2018. FR2589046 wasn't applied for by Dolby either, is titled "DEVICE FOR PREPARING INDUSTRIAL FOOD" (translated from DISPOSITIF POUR LA PREPARATION D'UN ALIMENT INDUSTRIEL) and has long expired.
So from a quick glance this list seems 'bloated' already. I'll have to dive into this a little more.
So, first of all, I'm just some dummy somewhere on the internet. Don't take my word, this is all based on information available to the public.
I've looked at the European Patents.
- EP2207170, "System for audio decoding with filling of spectral holes". Expires halfway 2023. It seems this would be applicable to both encoding and decoding. I have no clue whether this applies to Opus. Even if it does, the patent expires soon.
- EP1873754, "Audio encoder, audio decoder and audio processor having a dynamically variable warping characteristic". Expires 2026. I have no clue whether this applies.
- EP2144171, "Audio encoder and decoder for encoding and decoding frames of a sampled audio signal". Expires 2028. I have no clue whether this applies
- EP2352147 and EP2304723, expire 2029. Apply to encoding and decoding, one patent for each. This patent specifically states using bandwidth extension. As far as I know, Opus doesn't do that. The Opus specification says:
The output of the Opus decode is the sum of the outputs from the SILK and CELT decoders with proper sample rate conversion and delay compensation on the SILK side, and optional decimation (when decoding to sample rates less than 48 kHz) on the CELT side.
So there is no clever combining like the patents claim, simply added together.
- EP3364414, "Audio bandwidth extension decoder, corresponding method and computer program". Expires 2029. As far as I know Opus doesn't use bandwidth extension.
- EP3605534, EP2589046, EP3079153, EP3422346, EP3079152 and EP2757560, "Audio decoding with selective post-filtering ". Expires 2031. Applies to decoding. This first of these patents was filed at a point where the Opus specification wasn't final, but the draft was already 160 pages long, excluding the appendices.
- EP2625688, "Apparatus and method for processing an audio signal and for providing a higher temporal granularity for a combined unified speech and audio codec (usac)". Expires 2031, filed even later than the previous entry on this list. It seems to apply at least to decoding.
As noted before, the numbers of these European patents have been 'copied' as being French, British, Italian, Dutch patents, but these patent numbers do not seem to exist or are completely unrelated.
A couple of the Dolby patents also seem to resolve around voice activity detection (VAD). While the official Opus library certainly includes VAD in the encoder to automatically switch encoder modes, this is not an "essential" part of the codec (ffmpeg has a CELT-only Opus encoder and no VAD as far as I can tell). Also, libopus uses a very advanced neural-net VAD method (https://jmvalin.ca/opus/opus-1.3/) that appears to be very different to the simplistic approaches (moving average over subband energy) Dolby has claimed. In addition, this should not apply whatsoever to the decoder.
I think this patent list may have been assembled more by keyword search than actual critical technical review.
It's almost like they're going for quantity over quality for maximum FUD
It's almost like they're going for quantity over quality for maximum FUD
Their "Early Licensees" tier also smells a bit like "don't look too closely!" in my opinion. I assume they quickly want to aggregate a high number of licensees to construct an appearance of acceptance and validity.
I'm still skeptical they'll actually sue since it sounds like their patent pool is questionably applicable
I just realised how clever the timing of this is. Currently the biggest hurdle to a wider Opus adoption is support in devices such as car stereos, portable media players, etc. It's the perfect time to either try to FUD that away (if that's your goal), or to at least scam some money off of it.
opusfile hasn't been updated in over 2 years (problems with OpenSSL)
seems like they partnered with Dolby and Fraunhofer
legal issues aside, why isn't Opus more widely adopted? is this something to do with VR audio?
and i didn't know what these were until i searched
https://en.wikipedia.org/wiki/Patent_pool
It seems to me some patents on this list don't even exist?
FR3605534, GB3605534, NL3605534 and IL3605534 don't seem to exist, but EP3605534 does.
Same goes for NL2589046 and IT2589046, they don't exist, but EP2589046 does. GB2589046 is something completely different, not a patent of Dolby, and was only applied for in 2018. FR2589046 wasn't applied for by Dolby either, is titled "DEVICE FOR PREPARING INDUSTRIAL FOOD" (translated from DISPOSITIF POUR LA PREPARATION D'UN ALIMENT INDUSTRIEL) and has long expired.
So from a quick glance this list seems 'bloated' already. I'll have to dive into this a little more.
Yeah, the FR, GB, NL, IL, and IT "patents" with the same numbers as the EP filing only mean that the EP filing listed those countries as EP countries relevant to the EP filing. This large list, in general, includes a smaller number of base patents as well as their patent family member equivalents in other jurisdictions.
So, I took a look at the pool, and that whole list of patents distills down to the following 16 base patents, FYI:
US7447631
US7516064
US7873511
US8195454
US8296159
US8457975
US8892449
US9082395
US9100768
US9111530
US9224403
US9589571
US10026408
US10229696
US11081177
US20130226570
Everything else is a patent family member of the above 16 patent filings (15 granted US patents, 1 pending US patent application).
patent list was updated in October: https://www.opuspool.com/s/Opus-Patent-Pool-Licensed-Patents-01-Oct-2023.pdf
Because I mentioned Oracle/VirtualBox in the OP, I just want to say that I didn't actually see any evidence of causality there and apparently they removed Opus (https://github.com/mdaniel/virtualbox-org-svn-vbox-trunk/commit/10179927ec38ebcbcc3a606aa5d7caf67b810687) almost two months before the patent pool announcement. So it's possible that it's unrelated after all. But I also can't find any other explanation (besides the mysterious "bugref:10275"), so we can keep speculating...
There are currently 1 or more patent lawsuits going on about these 3 patents:
EP2304723
EP2352147
EP2144171
I don't know whether the patent owners are suing a company for violating their patents or whether someone is trying to invalidate these patents in court. 1 patent is owned by voiceage, the other 2 by fraunhofer.
i looked up the US patents and 2 of them are currently involved a lawsuit. I have also added their expiry dates from google patents.
US7447631 - 2024-10-07
US7516064 - 2026-01-30 lawsuit
US8195454 - 2029-03-28
US8296159 - 2029-06-23
US7873511 - 2029-11-18
US10229696 - 2030-08-23
US9082395 - 2031-07-23 lawsuit
US9224403 - 2032-02-04
US8892449 - 2032-02-10
US9111530 - 2032-02-29
US9100768 - 2032-04-04
US8457975 - 2032-04-05
US20130226570 - 2033-06-10
US9589571 - 2033-07-19
US11081117 - 2033-07-19
US10026408 - 2034-05-23