Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Opus vs FDK-AAC in 2025 (Read 1828 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Opus vs FDK-AAC in 2025

Hello everyone. In light of the latest updates, which is better in 2025.

Re: Opus vs FDK-AAC in 2025

Reply #1 –
Based on multiple listening tests here, OPUS is likely superior to Apple AAC, and Apple AAC is probably better than FDK. There haven't been any significant improvements recently, and to my knowledge, no new listening tests have been conducted to compare these two encoders/format.
You can expect OPUS to be more robust and efficient than FDK-AAC, especially at low bitrate (< 96 kbps).
Wavpack Hybrid -c4hx6

Re: Opus vs FDK-AAC in 2025

Reply #2 –
Indeed, Opus is more promising candidate. Opus is significantly better than AAC-LC on 96kbps. The main exception will be on very low bitrate, probably below 40kbps. If the user absolutely have to encode music (not speech) on very low rate, say 32kbps, probably due to  very narrow Internet connection or restrictive data caps, probably using HE-AAC or HE-AACv2 on FDK-AAC will be a better choice.

Re: Opus vs FDK-AAC in 2025

Reply #3 –
Thank you for your answers. I wanted to ask more for encoding multichannel sounds >=256k. Or does what is good at low bitrates also apply to high bitrates?

Re: Opus vs FDK-AAC in 2025

Reply #4 –
Quote from: Kamedo2 on
Indeed, Opus is more promising candidate. Opus is significantly better than AAC-LC on 96kbps. The main exception will be on very low bitrate, probably below 40kbps. If the user absolutely have to encode music (not speech) on very low rate, say 32kbps, probably due to  very narrow Internet connection or restrictive data caps, probably using HE-AAC or HE-AACv2 on FDK-AAC will be a better choice.

Thank you for your answers.

for Multichannel Sounds (5.1 or 7.1) >= 256k (fdk-he-aac)
OR stereo = 128k (fdk-he-aac)

Opus or FDK-AAC Which one should I use?

If I use FDK, do I lose much compared to opus with same bitrate?

Re: Opus vs FDK-AAC in 2025

Reply #5 –
Quote from: Kamedo2 on
Indeed, Opus is more promising candidate. Opus is significantly better than AAC-LC on 96kbps. The main exception will be on very low bitrate, probably below 40kbps. If the user absolutely have to encode music (not speech) on very low rate, say 32kbps, probably due to  very narrow Internet connection or restrictive data caps, probably using HE-AAC or HE-AACv2 on FDK-AAC will be a better choice.
Exactly. If they have a dial-up modem that's not the 56k one, HE-AACv2 at 24 kbps or lower is a great choice.
If they have a very slow dial-up modem or even a 2G GSM network (the one that's used by a lot of phones for calling people or sending SMS), using xHE-AAC at 12 kbps or 6 kbps in mono will be a lot better than Opus.
Opus is very glitchy and xHE-AAC will sound a lot better.

Re: Opus vs FDK-AAC in 2025

Reply #6 –
Quote from: MihaiPopa12346 on
Exactly. If they have a dial-up modem that's not the 56k one, HE-AACv2 at 24 kbps or lower is a great choice.
If they have a very slow dial-up modem or even a 2G GSM network (the one that's used by a lot of phones for calling people or sending SMS), using xHE-AAC at 12 kbps or 6 kbps in mono will be a lot better than Opus.
Opus is very glitchy and xHE-AAC will sound a lot better.

Yeah, Opus codec lacks Spectral Band Replication(SBR), the patent which would help in going very stingy on bitrate, so Opus understandably and predictably struggles at 32kbps or lower. On the upside, Opus is simple and computationally cheaper than HE-AAC.

Re: Opus vs FDK-AAC in 2025

Reply #7 –
@Kamedo2 I read somewhere that Opus has spectral folding, and when I look at spectrograms at low bitrate Opus I see it has some content spectrally copied and pasted above which verifies some SBRish stuff is going on; but what is the difference of that with SBR?

Re: Opus vs FDK-AAC in 2025

Reply #8 –
Quote from: Kamedo2 on
Yeah, Opus codec lacks Spectral Band Replication(SBR), the patent which would help in going very stingy on bitrate, so Opus understandably and predictably struggles at 32kbps or lower. On the upside, Opus is simple and computationally cheaper than HE-AAC.
I seen someone on a YouTube comment that says along the lines of "Opus has SBR, which helps at low bitrates" or the likes: :D
Quote from: photoniccannon2117
For anyone who is wondering why Opus sounds so good at ludicrously low bitrates, the answer is a technique called SBR. It’s used by Opus, as well as HE-AAC (not to be confused with regular AAC), and allows harmonic higher frequency content to be algorithmically reproduced by recreating it from the lower frequencies rather than simply encoding all of them separately. It’s a technique that is not without is tradeoffs (it does result in some artifacts), but it greatly increases the encoder’s efficiency at very low bitrates, and is the reason that codecs like HE-AAC and Opus sound reasonably good at such ludicrously low bitrates like 24kbps.

SBR is rarely used for higher bitrates. Opus only does this for very low bitrates, and HE-AAC is rarely used for higher bitrates (where regular AAC is almost always used instead). There are tradeoffs, but SBR is a game changer in terms of getting better audio quality out of some extremely low bitrates that might be required for streaming in environments where saving data is critical.
I mean that Opus sounds good at even 16 and 24 kbps is shocking, as I checked from a few bitrate comparison YouTube videos.

At 6 kbps it sounds EXACTLY like music on a GSM phone call. Specifically, the music on calls that get you while you are on hold.
Quote from: Xavier_Astraeus
Low bitrate for OPUS is great for achieving an authentic crappy phone call from that one company putting you on hold while you angrily wait for them to deal with your package that was supposed to come yesterday.
Yes, I mean this kind of thing. :D :D

Re: Opus vs FDK-AAC in 2025

Reply #9 –
I don't think we have evidence enough to conclude that using FDK-AAC to encode HE-AAC, whether v1 (SBR) or v2 (SBR+PS), will sound clearly better than Opus at any bitrate.

There's no direct comparisons I know of involving FDK at very low bitrate. When FDK has been compared to the closed source FhG encoder at higher bitrates, it had similar performance in normal situations and worse performance on problem samples.

While Opus development post 1.0 has rarely had large impacts on its performance on stereo music at usual recommended bitrates (80kbps+), it has seen more changes that improved lower bitrate music in each of 1.1, 1.2, and 1.3. Some old comparisons do not reflect recent performance.

Here's much of what evidence we have available that could be relevant:


That's not enough to make a solid conclusion for low bitrates.

It's understandable that these bitrates have received less careful listening test attention for stereo music. Neither encoder will excel at 24kbps and use cases demanding that low a rate are not that common today.

Klymins and Mihai, yes, Opus's spectral folding, along with the fact that Opus preserves the spectral energy envelope for each band, accomplish some of the same purpose as SBR. Jean-Marc said in the AES paper that Opus's spectral folding in itself "is far less advanced than spectral band replication (SBR) from HE-AAC and mp3PRO, but is computationally inexpensive, requires no extra delay, and the decision to apply it can change frame-by-frame." (SBR does add delay; HE-AAC in practice has considerably higher delay than LC-AAC.) I don't think one can conclude much a priori about the actual quality impact resulting from the difference between these two approaches.

Of course, the FhG xHE-AAC encoder does substantially better than Opus at low bitrates. If you really must encode music at 32kbps or below, and if your application doesn't need low delay and works just fine with needing to pay for a closed source encoder and patent license, and if you don't need it to be played on platforms that don't support xHE-AAC, the closed source FhG xHE-AAC encoder is pretty clearly the way to go.

Re: Opus vs FDK-AAC in 2025

Reply #10 –
Oops, I'd missed this 40kbps comparison between Opus 1.3 and FDK HE-AACv2 done by Kamedo2, which showed them tied. Again, doesn't directly lend itself to saying FDK-AAC outperforms Opus at low bitrate.

Re: Opus vs FDK-AAC in 2025

Reply #11 –
Quote from: Kamedo2 on
Yeah, Opus codec lacks Spectral Band Replication(SBR), the patent which would help in going very stingy on bitrate, so Opus understandably and predictably struggles at 32kbps or lower. On the upside, Opus is simple and computationally cheaper than HE-AAC.
Opus lacks SBR as it is was patented. Correction, patents related to SBR expired. While I am not a lawyer, all SBR-related patents mentioned online appear to have expired. As a general rule, after 22 years from a format's approval, defending the existence of related patents becomes difficult. Anyways, 20+ years later all interest in low rates, SBR and such things are all long gone.
Instead of SBR, Opus uses band folding. Jean-Marc was quite modest and didn't state that band folding is superior. Unlike SBR, which resamples audio to 22kHz from the original 44.1kHz, limiting its use to lower bitrates, band folding maintains higher quality and remains effective even at bitrates of 128 kbps and above. 

Opus bandwidth extension is not as simple as it appears. As @jensed said band folding in conjunction with spectral energy envelop outperform SBR.

Quote from: Kamedo2 on
... Opus understandably and predictably struggles at 32kbps ...
Opus low performance at low rates is due to frequency leakage (low delay), not due to lack of / or weak bandwidth extension.

Re: Opus vs FDK-AAC in 2025

Reply #12 –
One last thing about the above before getting back to the OP's main question: I should have added that the one private listener who tested and rated HE-AACv2 over Opus was using CBR.

Quote from: Neptune1984 on
for Multichannel Sounds (5.1 or 7.1) >= 256k (fdk-he-aac)
OR stereo = 128k (fdk-he-aac)

Opus or FDK-AAC Which one should I use?

If I use FDK, do I lose much compared to opus with same bitrate?
For 128kbps stereo, FDK-AAC will use LC-AAC rather than HE-AAC, as it should. No reason to use HE for stereo at >80kbps. The difference beween Opus and FDK won't be significant for most listeners at 128kbps stereo. Opus is likely to be a little better, especially on genres other than classical.

I'm not aware of any blind listening tests that compare the codecs for surround. For Opus, here's some bitrate recommendations from Xiph for 5.1 or 7.1 surround, and the one Opus surround sound test I could find, which used Opus 1.1 and thus misses some later improvements. Opus has some surround channel VBR allocation smarts, and should be good at 256kbps for 5.1; I don't know what FDK does.

I hope someone who's actually used these codecs for surround can add to the conversation.

Re: Opus vs FDK-AAC in 2025

Reply #13 –
Quote from: jensend on
No reason to use HE for stereo at >80kbps
mp3PRO can do up to 96 kbps, HE-AAC can do up to 264 kbps and HE-AACv2 can do up to 132 kbps.
But of course, HE-AACv2 sounds a lot worse than AAC LC for the same bitrate.
Quote from: Xiph.org's wiki
Codec 2 handles ultra low bitrate speech at 0.7 - 3.2 Kb/s.
TWELP from DSP Innovations can do at 0.3 - 3.6 kbps.
Who needs this of a bitrate to store or stream high quality speech?
Of course, if you have a 1.44 MB floppy disk, you can record for hours and store with this codec. :D
 

Re: Opus vs FDK-AAC in 2025

Reply #14 –
Quote from: btc on
Quote from: Kamedo2 on
Yeah, Opus codec lacks Spectral Band Replication(SBR), the patent which would help in going very stingy on bitrate, so Opus understandably and predictably struggles at 32kbps or lower. On the upside, Opus is simple and computationally cheaper than HE-AAC.
Opus lacks SBR as it is was patented. Correction, patents related to SBR expired. While I am not a lawyer, all SBR-related patents mentioned online appear to have expired. As a general rule, after 22 years from a format's approval, defending the existence of related patents becomes difficult. Anyways, 20+ years later all interest in low rates, SBR and such things are all long gone.
Instead of SBR, Opus uses band folding. Jean-Marc was quite modest and didn't state that band folding is superior. Unlike SBR, which resamples audio to 22kHz from the original 44.1kHz, limiting its use to lower bitrates, band folding maintains higher quality and remains effective even at bitrates of 128 kbps and above. 

Opus bandwidth extension is not as simple as it appears. As @jensed said band folding in conjunction with spectral energy envelop outperform SBR.

Quote from: Kamedo2 on
... Opus understandably and predictably struggles at 32kbps ...
Opus low performance at low rates is due to frequency leakage (low delay), not due to lack of / or weak bandwidth extension.

There's a thing called AAC-ELD, it's basically a low-delay HE-AAC, built over AAC-LD (which already starts to sound too muffled below ~64kbps, so adding SBR was needed to keep a good bandwidth). According to this graph its algorithmic delay is ~20 ms like Opus. There's also AAC-ELDv2, which adds some low-delay form of Parametric Stereo. I'm not using ELDv2 as I don't have any way to encode it... it seems that it could be a little better than Opus for practically most low-bitrate cases, though having a slightly higher delay. (AAC-ELD only with SBR is sometimes better, many times worse or equal to Opus at the 32-48 kbps edge; above 48kbps Opus is definitely better. I suppose that below 32kbps AAC-ELDv2 would be the best option... afaik xHE-AAC isn't designed for low-delay)
 :>
I attach a 7zip file with six lossless audio samples, and AAC-ELD (m4a) / Opus audio files at 32kbps. Just in case there are decoded (flac) versions of the m4a files ("fdk-aac packet decoder" plugin is needed for playback of AAC-ELD on foobar2000).