-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
CTXMatrix - Free Ambiophonics Plugin
- Thread starter STC
- Start date
Keith_W
Master Contributor
Hello @STC. I downloaded CTXMatrix Beta 17 and had another go. I paid attention to your note that delays are important, so this time I tried to find an optimal setting with "balanced" delays and gains. This is what I came up with:
Stage 1: Gain -6 Delay 2 samples
Stage 2: Gain -3 Delay 2 samples
Stage 3: Gain -6 Delay 2 samples
Stage 4: Gain -3 Delay 2 samples
The good news is that the tonality is closer to the original, but missing some treble. Adjusting the treble gain makes the treble harsher, but does not actually give "life" and "brilliance" back to music. I think the treble setting is probably too low in the frequency range, i'm not sure.
It's a bit difficult to describe the soundstage widening effect, so i'll draw a diagram.
If you imagine five soloists, this is what I hear with no processing.
This is what uBACCH seems to do. The central instrument stays the same size, but instruments further to the side are stretched out. The tonality is either the same as the original, or very very close.
And this is what I hear with CTXMatrix. The vertical soundstage seems to compress, and all instruments are stretched. A soloist who is pinpoint dead centre in the original sounds stretched and smeared in width.
I am starting to think:
- attempting to optimize this further for a 60deg listening angle is going to be a waste of time. I will need to move my speakers to a 20deg angle and try again.
- dialling this in without measurement is extremely difficult. Surely there is a way to dial this in with measurements?
- I don't know how much more I can help you. I don't know enough about how the algorithm works short of a vague idea that it cancels crosstalk.
Maybe we should chat on the phone. I'll inbox you with my phone no.
Stage 1: Gain -6 Delay 2 samples
Stage 2: Gain -3 Delay 2 samples
Stage 3: Gain -6 Delay 2 samples
Stage 4: Gain -3 Delay 2 samples
The good news is that the tonality is closer to the original, but missing some treble. Adjusting the treble gain makes the treble harsher, but does not actually give "life" and "brilliance" back to music. I think the treble setting is probably too low in the frequency range, i'm not sure.
It's a bit difficult to describe the soundstage widening effect, so i'll draw a diagram.
If you imagine five soloists, this is what I hear with no processing.
This is what uBACCH seems to do. The central instrument stays the same size, but instruments further to the side are stretched out. The tonality is either the same as the original, or very very close.
And this is what I hear with CTXMatrix. The vertical soundstage seems to compress, and all instruments are stretched. A soloist who is pinpoint dead centre in the original sounds stretched and smeared in width.
I am starting to think:
- attempting to optimize this further for a 60deg listening angle is going to be a waste of time. I will need to move my speakers to a 20deg angle and try again.
- dialling this in without measurement is extremely difficult. Surely there is a way to dial this in with measurements?
- I don't know how much more I can help you. I don't know enough about how the algorithm works short of a vague idea that it cancels crosstalk.
Maybe we should chat on the phone. I'll inbox you with my phone no.
- Thread Starter
- #43
Thanks Keith for sharing the violin illustration! The image spread evenly looks pretty good at 60 degrees for CTXMatrix.
I ran a quick test with my $10 PC speakers with 60 degrees and got a delay of about 30 to 40 samples with an attenuation of roughly 3.5 to 4 dB for 60 degrees. Your 2-sample delay doesn’t quite cut it for matching the inverted signal, which probably explains why the image looks a bit smeared.
Even my main speakers, which are right next to each other,
Here’s what I measured:
• -2.5 dB at 8 samples
• -3.7 dB at 10 samples
• -3.2 dB at 9 samples
• -2.4 dB at 8 samples
(this one works best for me)
I also tried settings like -1.5 dB at 3 samples which works, but you can definitely hear the difference between an effect and actual cancellation. I’d suggest testing with human speech since our ears pick up on those changes really well. Playing around with 3 to 4 dB and 30 to 45 samples should do the trick. THe actual delay should be around 150 μs to 200 μs but my samples is dependent on the source or your system sampling rate. it was designed to work up to 384kHz .
- Thread Starter
- #44
Here is what you can expect with CTXmatrix compared to BACCH4Mac.
Some insights on BACCH4Mac based on Tom Martin’s detailed review in The Absolute Sound (full article here: https://www.theabsolutesound.com/ar...ied-physics-bacch4mac-stereo-purifier-review/).
To illustrate the difference, here’s a relevant quote from the review regarding Bob Dylan’s The Freewheelin’ Bob Dylan (1963):
“With BACCH, you hear Bob in the center where you would kind of expect him if you think about how most stage performances are arranged, and then the recording engineers put his guitar on the right and the harmonica on the left. That isn’t how it would sound live, but this is a studio recording and you hear how the engineers placed things so that it sounds a bit like a trio. Apart from the deviation between studio and live, what’s impressive is the guitar occupies a space that extends outside of the right speaker and feels appropriately sized for the dreadnought Gibson guitar that Bob played in this period.”
In comparison, with CTXMatrix, my observations on the same recording are more constrained: the guitar is positioned on the right but extends less than 30 degrees from center; Dylan’s voice is not precisely centered but slightly offset toward the left, maybe 5 degrees; and the harmonica reaches roughly 45 degrees to the left. ( Blowin’ in the Wind - Apple Music Hirez streaming)
Another example from the review involves Art Blakey and the Jazz Messengers’ Buhaina’s Delight (1961):
“On Art Blakey and the Jazz Messenger’s album Buhaina’s Delight, the players are arranged in a 270° arc around the listening position that may be effectively where the microphones were. However, I have to say it’s a little bit odd to hear Art Blakey playing in a position slightly behind where I’m sitting, and on the right side. There is a horn off my left side. The other players are arrayed as you would typically hear on stereo between the speakers. Not what you would hear in a club. This studio recording is from 1961 and I just don’t think stereo recording in the studio was evolved to the point where the engineers were trying to do this. I think the engineers wanted Blakey right front so they pan-potted him hard right and the data on the recording has this placed at the right rear. I got similar results with Dave Brubeck’s Time Out, from 1959. But Bill Frissell’s Valentine, a jazz trio studio recording from 2020, sounds phenomenal in the sense that the performers are where you would expect them.”
With CTXMatrix on the same Blakey album, the soundstage remains far more limited: the most extreme left placement trumpet reaches about 40 degrees to the my left,; a similar instruments appears at roughly 30 degrees to the right; and the piano stays centered maybe slight left. It moves abit. Drums is to the right. In short maybe a 70 to 80 degrees angle to the listeners. In short, CTXMatrix does not achieve a 270-degree arc or placements behind the listener. Listening to Take Five from the Time Out album. Drums at about 15 more to the my left but hit hat smears from to centre and right forward ( decaying) occasionally. PIano is to my right. The sax is centre slightly off centre toward my left. The right side remains very much with 10 to 30 degrees only. Nothing goes outside the imaginary 60 degrees unless you push the volume up and some reverbs extends more. This is from CD.
But the question is should it? Is that correct? Can CTXMatrix achieve the so called 270 degrees? Users can play with the delay and cancellation to find out. And accept the limitation of CTX. It cannot do miracles. It is just a crosstalk cancellation plugin.
Some insights on BACCH4Mac based on Tom Martin’s detailed review in The Absolute Sound (full article here: https://www.theabsolutesound.com/ar...ied-physics-bacch4mac-stereo-purifier-review/).
To illustrate the difference, here’s a relevant quote from the review regarding Bob Dylan’s The Freewheelin’ Bob Dylan (1963):
“With BACCH, you hear Bob in the center where you would kind of expect him if you think about how most stage performances are arranged, and then the recording engineers put his guitar on the right and the harmonica on the left. That isn’t how it would sound live, but this is a studio recording and you hear how the engineers placed things so that it sounds a bit like a trio. Apart from the deviation between studio and live, what’s impressive is the guitar occupies a space that extends outside of the right speaker and feels appropriately sized for the dreadnought Gibson guitar that Bob played in this period.”
In comparison, with CTXMatrix, my observations on the same recording are more constrained: the guitar is positioned on the right but extends less than 30 degrees from center; Dylan’s voice is not precisely centered but slightly offset toward the left, maybe 5 degrees; and the harmonica reaches roughly 45 degrees to the left. ( Blowin’ in the Wind - Apple Music Hirez streaming)
Another example from the review involves Art Blakey and the Jazz Messengers’ Buhaina’s Delight (1961):
“On Art Blakey and the Jazz Messenger’s album Buhaina’s Delight, the players are arranged in a 270° arc around the listening position that may be effectively where the microphones were. However, I have to say it’s a little bit odd to hear Art Blakey playing in a position slightly behind where I’m sitting, and on the right side. There is a horn off my left side. The other players are arrayed as you would typically hear on stereo between the speakers. Not what you would hear in a club. This studio recording is from 1961 and I just don’t think stereo recording in the studio was evolved to the point where the engineers were trying to do this. I think the engineers wanted Blakey right front so they pan-potted him hard right and the data on the recording has this placed at the right rear. I got similar results with Dave Brubeck’s Time Out, from 1959. But Bill Frissell’s Valentine, a jazz trio studio recording from 2020, sounds phenomenal in the sense that the performers are where you would expect them.”
With CTXMatrix on the same Blakey album, the soundstage remains far more limited: the most extreme left placement trumpet reaches about 40 degrees to the my left,; a similar instruments appears at roughly 30 degrees to the right; and the piano stays centered maybe slight left. It moves abit. Drums is to the right. In short maybe a 70 to 80 degrees angle to the listeners. In short, CTXMatrix does not achieve a 270-degree arc or placements behind the listener. Listening to Take Five from the Time Out album. Drums at about 15 more to the my left but hit hat smears from to centre and right forward ( decaying) occasionally. PIano is to my right. The sax is centre slightly off centre toward my left. The right side remains very much with 10 to 30 degrees only. Nothing goes outside the imaginary 60 degrees unless you push the volume up and some reverbs extends more. This is from CD.
But the question is should it? Is that correct? Can CTXMatrix achieve the so called 270 degrees? Users can play with the delay and cancellation to find out. And accept the limitation of CTX. It cannot do miracles. It is just a crosstalk cancellation plugin.
- Thread Starter
- #47
CTXMatrix is now Ai certified with Grok’s Audio Analysis Capabilities.
While examining claims that crosstalk cancellation causes severe tonal coloration, I created a short YouTube video alternating plain stereo and CTXMatrix (Ambiophonics RACE). Recorded casually with iPhone 15 Pro — not binaural.([youtube]4mG2heTymCQ[/MEDIA])
Switches were hard to detect reliably on phone speakers. Shared link with Grok (xAI), asked for differences — no labels/timestamps given.
Grok’s response was striking. It:
• detected A/B switches every 5–8 seconds
• described CTXMatrix as wider, more spacious, more detailed, better separation, stronger 3D imaging
• noted stereo as narrower, more centered
• reported only minimal tonal shift — no severe coloration
• gave approximate transition timings from audible cues
• explained via waveform differences, stereo correlation, phase, inter-channel timing, spatial cues
All from ordinary smartphone recording.
Grok stated:
“Binaural recording is not required. Ordinary stereo or smartphone recordings can be analyzed effectively through waveform differences, inter-channel correlation, phase information, and timing cues. Binaural capture would enhance spatial precision by incorporating head-related transfer functions (HRTF), but it is unnecessary for reliably detecting mode switches, soundstage width, instrument separation, or the presence/absence of severe coloration. The current capability performs strongly even with non-binaural material.”
Other AIs responded as follows:
Perplexity
Without direct access to analyze the video’s audio tracks (fetch failed, no waveform/spectral comparisons found), it likely uses one stereo recording, not dual versions or CTXMatrix processing, as no descriptions match audio A/B tests. To confirm, download the audio via tools like Audacity, compare waveforms, or check stereo imaging.
Claude
I cannot access the video directly, but based on context from the Ambiophonics site where CTXMatrix is discussed, here’s what’s likely going on:
∙ The video probably demonstrates two versions of the same recording — one plain stereo and one processed through CTXMatrix (crosstalk-cancelled).
∙ The version described as “severely coloured” is almost certainly the CTXMatrix-processed version, because crosstalk cancellation involves heavy filtering (comb filtering, delays, attenuation) that can sound tonally coloured or unnatural when listened to on headphones or without the intended narrow-angle speaker setup it was designed for.
∙ The plain stereo version would be the unprocessed one.
DeepSeek
Based on the video page you provided, I cannot determine if the video contains two different recordings or assess the sound quality. The page only shows metadata and related videos, not the video content itself.
Gemini
The video appears to show two versions of the same song played through a high-end audio system. The first version, which starts at [00:00], is in stereo and sounds much cleaner and more balanced. The second version, which starts at [00:21], is the CTXMatrix version and is indeed “severely coloured” as it sounds much more reverb-heavy and less detailed than the stereo version. The differences between the two versions are likely due to the different processing used in each version.
Grok
Yes, slight differences are audible.
The video is one clip with repeated A/B switching (no labels/timestamps shown).
Differences appear at each switch: CTXMatrix sounds wider, more open/spacious, with better separation and less “in-head” focus than stereo. Tonal balance/colouration remains very similar—no severe change.
Switches happen roughly every 5–10 seconds (alternating multiple times throughout the short video). The first switch is early (~few seconds in), then back and forth.
Stereo: narrower, more centered image.
CTXMatrix: noticeably wider/3D stage, subtle air/clarity gain.
No exact timed labels or transcript available in video/description.
Video is short (~40–60 s), one continuous music clip with unlabeled A/B switches ~every 5–8 seconds.
Approximate breakdown from audible transitions (listen for sudden stage width change):
• 0:00–0:07: Stereo (narrower, centered)
• 0:07–0:14: CTXMatrix (wider, more open)
• 0:14–0:21: Stereo
• 0:21–0:28: CTXMatrix
• 0:28–0:35: Stereo
• 0:35–0:42: CTXMatrix
• 0:42–end (~0:45–0:55): Stereo or final CTXMatrix (pattern may repeat 1–2 more times)
Stereo: more in-head, less separation.
CTXMatrix: wider 3D image, subtle clarity increase, no severe coloration.
Switches are abrupt but unmarked—rely on ear for precise detection.
Grok delivered precise, technically grounded observations that matched careful listening — others either could not analyze the audio or made assumptions that contradicted it.
This shows Grok can extract meaningful spatial and tonal detail — including greater spaciousness and detail in CTXMatrix — from ordinary phone recordings without binaural capture. It also didn’t complain that YouTube SQ is compressed so SQ could be judged.
Edit:- the prompt used was “
I want to know if this video shows two different recordings or just one. Is one version severely coloured, as people describe the sound quality?
YouTube link.”
THe recording is Norah Jones Don’t Know why. I did switch every few seconds but I kept it blind. With AirPod I could guess one section from the video. If binaural then the difference will be obvious. PHone used was iPhone 15 Pro. link is unlisted
remove the extra slash to watch.
-https:///youtu.be/4mG2heTymCQ-
While examining claims that crosstalk cancellation causes severe tonal coloration, I created a short YouTube video alternating plain stereo and CTXMatrix (Ambiophonics RACE). Recorded casually with iPhone 15 Pro — not binaural.([youtube]4mG2heTymCQ[/MEDIA])
Switches were hard to detect reliably on phone speakers. Shared link with Grok (xAI), asked for differences — no labels/timestamps given.
Grok’s response was striking. It:
• detected A/B switches every 5–8 seconds
• described CTXMatrix as wider, more spacious, more detailed, better separation, stronger 3D imaging
• noted stereo as narrower, more centered
• reported only minimal tonal shift — no severe coloration
• gave approximate transition timings from audible cues
• explained via waveform differences, stereo correlation, phase, inter-channel timing, spatial cues
All from ordinary smartphone recording.
Grok stated:
“Binaural recording is not required. Ordinary stereo or smartphone recordings can be analyzed effectively through waveform differences, inter-channel correlation, phase information, and timing cues. Binaural capture would enhance spatial precision by incorporating head-related transfer functions (HRTF), but it is unnecessary for reliably detecting mode switches, soundstage width, instrument separation, or the presence/absence of severe coloration. The current capability performs strongly even with non-binaural material.”
Other AIs responded as follows:
Perplexity
Without direct access to analyze the video’s audio tracks (fetch failed, no waveform/spectral comparisons found), it likely uses one stereo recording, not dual versions or CTXMatrix processing, as no descriptions match audio A/B tests. To confirm, download the audio via tools like Audacity, compare waveforms, or check stereo imaging.
Claude
I cannot access the video directly, but based on context from the Ambiophonics site where CTXMatrix is discussed, here’s what’s likely going on:
∙ The video probably demonstrates two versions of the same recording — one plain stereo and one processed through CTXMatrix (crosstalk-cancelled).
∙ The version described as “severely coloured” is almost certainly the CTXMatrix-processed version, because crosstalk cancellation involves heavy filtering (comb filtering, delays, attenuation) that can sound tonally coloured or unnatural when listened to on headphones or without the intended narrow-angle speaker setup it was designed for.
∙ The plain stereo version would be the unprocessed one.
DeepSeek
Based on the video page you provided, I cannot determine if the video contains two different recordings or assess the sound quality. The page only shows metadata and related videos, not the video content itself.
Gemini
The video appears to show two versions of the same song played through a high-end audio system. The first version, which starts at [00:00], is in stereo and sounds much cleaner and more balanced. The second version, which starts at [00:21], is the CTXMatrix version and is indeed “severely coloured” as it sounds much more reverb-heavy and less detailed than the stereo version. The differences between the two versions are likely due to the different processing used in each version.
Grok
Yes, slight differences are audible.
The video is one clip with repeated A/B switching (no labels/timestamps shown).
Differences appear at each switch: CTXMatrix sounds wider, more open/spacious, with better separation and less “in-head” focus than stereo. Tonal balance/colouration remains very similar—no severe change.
Switches happen roughly every 5–10 seconds (alternating multiple times throughout the short video). The first switch is early (~few seconds in), then back and forth.
Stereo: narrower, more centered image.
CTXMatrix: noticeably wider/3D stage, subtle air/clarity gain.
No exact timed labels or transcript available in video/description.
Video is short (~40–60 s), one continuous music clip with unlabeled A/B switches ~every 5–8 seconds.
Approximate breakdown from audible transitions (listen for sudden stage width change):
• 0:00–0:07: Stereo (narrower, centered)
• 0:07–0:14: CTXMatrix (wider, more open)
• 0:14–0:21: Stereo
• 0:21–0:28: CTXMatrix
• 0:28–0:35: Stereo
• 0:35–0:42: CTXMatrix
• 0:42–end (~0:45–0:55): Stereo or final CTXMatrix (pattern may repeat 1–2 more times)
Stereo: more in-head, less separation.
CTXMatrix: wider 3D image, subtle clarity increase, no severe coloration.
Switches are abrupt but unmarked—rely on ear for precise detection.
Grok delivered precise, technically grounded observations that matched careful listening — others either could not analyze the audio or made assumptions that contradicted it.
This shows Grok can extract meaningful spatial and tonal detail — including greater spaciousness and detail in CTXMatrix — from ordinary phone recordings without binaural capture. It also didn’t complain that YouTube SQ is compressed so SQ could be judged.
Edit:- the prompt used was “
I want to know if this video shows two different recordings or just one. Is one version severely coloured, as people describe the sound quality?
YouTube link.”
THe recording is Norah Jones Don’t Know why. I did switch every few seconds but I kept it blind. With AirPod I could guess one section from the video. If binaural then the difference will be obvious. PHone used was iPhone 15 Pro. link is unlisted
remove the extra slash to watch.
-https:///youtu.be/4mG2heTymCQ-
Last edited:
- Joined
- May 12, 2019
- Messages
- 240
- Likes
- 117
Technically grounded seeming observations, which LLMs are great at - seeming like they know what they're talking about. If it did know what it was talking about it would've admitted it can't judge image width or anything "3D" or spatial (except reverb quantity) from a single-microphone recording.
- Thread Starter
- #49
Your concern is valid and I too wondered if it is making up stuff. Wondered if it is using Goniometer to predict the soundstage to which it replied
“
I use internal signal analysis:
• Inter-channel correlation
• Phase coherence
• L/R amplitude & timing differences
• Side vs mid energy ratio
• Transient & detail envelope comparison
All computed directly from the stereo waveform — no external software or visual goniometer needed.“
iPhone uses more than a single mic. i am more curious with its ability to analyze. Here is Grok’s opinion of two videos of AVShowreport YouTube channel.
“
Kharma (first link) edges out slightly better overall.
• Wider, deeper soundstage
• Stronger spatial placement and 3D imaging
• More natural vocal presence and air
• Better separation and layering
Thrax (second link)
• Good clarity and detail
• Slightly narrower stage
• More forward presentation
• Excellent dynamics but less depth
Both high-end systems; difference audible but small in YouTube compression. Kharma feels more immersive/spacious.“
- Joined
- May 12, 2019
- Messages
- 240
- Likes
- 117
I see. Well that's not exactly the impression you left when you said
'Not binaural but stereo' is a distinction without a difference when you're going to analyze the audio with software and not with a human brain.
Anyway, is this a local install of Grok that you've fed a bunch of audio research papers, or something ready-made by xAI?
Keith_W
Master Contributor
If there is a "CTXMatrix Lite" ... is there another version that is "Heavy"?
- Thread Starter
- #54
Not Heavy but Pro. just fancy name.
The Pro version is meant for in-ear measurements and EQ for higher attenuation. But instead of improving the sound it is causing weird effects with some recordings so I have just postponed or abandoned the idea. Maybe, one day it can be implemented in a DAC or digital Amp with auto calibration.
But I’m out of my depth here. Lite is pure cancellation and with the limited blind tests it seems good enough. Maybe I will post the blind tests file here to actually see what the optimum attenuation needed is.
- Thread Starter
- #57
Thanks for sharing!
PUSSI-XTC looks interesting and very complex. Will give it a try tomorrow.
More tools in this niche hobby are always welcome; it helps push the whole crosstalk cancellation scene forward.
I’d suggest reaching out directly to Archimago. He has done in-depth XTC write-ups in the past (including testing CTXMatrix recently) and is usually open to trying new plugins. His measurements and listening impressions are well-respected in the community.
Wish you the best!
ST
