AI + Room Correction + Current Equipment = Improved SQ!

[Cool Runnings]

Just a regular non-tech trained hobbyist. I frankly listen to my equipment more than the music even though I do love my music.

ASR has helped me so much to get good value equipment that favors objective acoustic science and neutrality over everything else. Cut out so much fluff, wasted money, time and frustration.

So I've got a baseline set-up of decent ASR recommended systems for my office and bedroom with a legacy home theatre system for my living room. It's a happy mix of Wiims, Ascend Acoustic, Genelec, REL, Kali, Nord with legacy Triangles and Denon AVR3700. I thought it was all good just using the built in Wiim Roomfit and Audyssey software for room correction with low level attention to speaker placement and minor acoustic paneling here and there.

Over the past week on a whim I decided to use Gemini to critique and improve all my set-ups including the stock system in my car. It's been pretty mind blowing what Gemini can do and improve on for non-tech hobbyists like me. It can read pictures and videos of my set-up and sound output. It can read pictures of my app settings esp the Wiim Roomfit settings. And then make detailed recommendations to improve on them with explanation for each recommendation. For eg, I thought my office set-up sounded hollow and thin on the mids and boomy on the bass despite Roomfit. After about 30 mins of chat and info sharing with Gemini, I got a whole new and much better sound to my preference just using the Gemini recommendations which went into detailed parametric EQ setting , cross-overs, gains and pre-gains, speaker placement, acoustic panels placement recommendations etc etc. Pretty amazing and a great experience for non tech people who have always struggled with room correction after we get the equipment baseline sorted somewhat.

I then moved on to do the same for my other set-ups. Same big improvements to SQ. As least to my ears. I also realized how ignorant I was in this hobby about achieving SQ upgrades. I was stuck in a rut of just using in app software where they were available. So much more could be improved without even spending single cent more if we get the right help on understanding basic principles of room acoustics and DSP sound correction. Gemini and I am sure most other good LLMs turned out to be so much more capable then I'd ever imagined.

Do I think this path is for everyone? No for those who are already knowledgeable and proficient on DSP/room correction. Do I think using LLMs will get you to ideal end state for the equipment you have already? Probably not, even though I was mind-blown using Gemini it was clear that there were some gaps in its recommendations and reasoning. It can get a little too positive and sometimes forgetful or mistakes can be made reading some of the pictures. So a little trial and error and to-ing and fro-ing is needed.

I would say I was achieving maybe 5/10 max 6/10 potential of my system and just a first round using Gemini probably got me to 8/10. But I think a lot of hobbyists like me will get similar gains if they try. What have you got to lose other than a few hours to talking to a always positive and enthusiastic AI and tinkering with your equipment and learning so much more about acoustic science in the process.

Highly recommended to all those who have not already tried it.

EDIT: Tried to do same on Claude. Not good at all. Gemini is way better. Even though I think Claude wins Gemini on many other AI tasks. Gemini is much better as of now on audio visual analysis.

 

[Oddball]

Glad it worked for you. If you have any affinity towards tweaking, you might want to continue understanding how it all works and adds up to further improve things. At this day and age, this might be a better way to get somewhere fast, rather than through painful self-learning and reading long and winding posts on the forums that anyway often don't reach consensus.

The best practice is to check the room correction results with REW and UMIK as that is more accurate than our hearing that might also suffer from confirmation bias.

 

[Cool Runnings]

Glad it worked for you. If you have any affinity towards tweaking, you might want to continue understanding how it all works and adds up to further improve things. At this day and age, this might be a better way to get somewhere fast, rather than through painful self-learning and reading long and winding posts on the forums that anyway often don't reach consensus.

The best practice is to check the room correction results with REW and UMIK as that is more accurate than our hearing that might also suffer from confirmation bias.

Seems to have worked a treat so far. Nowhere near as ideal as the expert tweakers can achieve with tools like DIRAC but for non-tech people like me the whole non-AI process of learning about room correction the right way is very daunting, confusing and frustrating and possibly costly without the payoff. Gemini aided RC is like a fun easy tool for people like me to get to maybe 70-80% of the hardware potential at zero cost. Then that gives me some foundation and confidence to move onto more complicated RC tweaks possibly using DIRAC. But yes you are right next step is UMIK and REW. Thanks for that advice!

Also, to your point, it's not even about affinity to room correction. After this positive journey aided by Gemini, I am convinced that ALL audio hobbyists and enthusiasts must put RC at least into top 3 ongoing priority. I now fully realize that the room conditions is the biggest factor in sound quality as long as you don't have very poor quality audio hardware ie anything with no audible distortion is good enough, the rest is about RC.

On a separate note, it's way easier to use Gemini for 2.1 systems. It has been a very difficult using Gemini on my legacy 3.1.2 home theatre system with Denon AVR3700h at the heart. Not that any other AI I know of will make it any easier. Claude sucks at it way more. I think it is more that a 3.1.2 set-up in a bigger living room with the Denon's software for manual set-up is really quite complicated to figure out in any case with or without AI.

 

[Oddball]

Looks like you are on the right way to learn more about room correction that indeed is very important - right after room and speakers.

I would not be afraid of Dirac at all. I spent years and years tweaking Audy and trying to take it to the next level with D&M enhancements like LFE+Main and LFE distribution, as well as Audy's own $200 MultiEQ-X app that is much more flexible and can also import REW filters. This is what is really mind-bending complicated.

Since I switched to full Dirac package in Oct last year (Dirac Live, Bass Control and ART), things actually got much simpler. There is quite a bit to tweak with Dirac ART, but it gets you in most rooms to probably 90% of the endgame just by running a calibration and trying settings that ART suggested. So really not complicated at all. Lots of theory behind but also lots of threads around, including on our own forum, that can help. The consensus is that either compared to Bass control, or probably especially Audy, the end result is superior.

Your 3700H will not support Dirac, and probably not right time now to jump to 3800H until we have more details about 3900H that seems to be just around the corner with some potential enhancements. Audy can be really good, and lots of people around here have enjoyed it for decade or so. But then things move on and new things come...

 

[thcdru2k]

AI has been helpful with Raspberry pi automations for me around my DSP. My raspberry pi is connected to my Flex HTx via USB and Denon (wirelessly). The Pi helps me control and automate volume/input/presets on Flex HTx. It helps me work around HDMI-CEC issues. My LG OLED does not turn off automatically when I turn off my AVR or any connected device so it senses the system state and turns off the LG OLED as needed using a Broadlink transmitter. I actually also use it to control my Sonance amp from a web-based GUI for my outdoor system since it has API control which is nice because otherwise I would have to login to the Sonance GUI to make changes. It is nice to have a single web-based GUI that shows Flex HTx state, Denon state, what is playing on Spotify since my devices are not visible from where I sit down to listen to music or watch movies.

For time/phase alignment I still feel more comfortable verifying real improvements by looking at REW graphs but it is helpful for general starting points for EQ/crossover.

AI is a useful tool but it should be paired with repeatable setup and objective verification.

 

[Cool Runnings]

Looks like you are on the right way to learn more about room correction that indeed is very important - right after room and speakers.

I would not be afraid of Dirac at all. I spent years and years tweaking Audy and trying to take it to the next level with D&M enhancements like LFE+Main and LFE distribution, as well as Audy's own $200 MultiEQ-X app that is much more flexible and can also import REW filters. This is what is really mind-bending complicated.

Since I switched to full Dirac package in Oct last year (Dirac Live, Bass Control and ART), things actually got much simpler. There is quite a bit to tweak with Dirac ART, but it gets you in most rooms to probably 90% of the endgame just by running a calibration and trying settings that ART suggested. So really not complicated at all. Lots of theory behind but also lots of threads around, including on our own forum, that can help. The consensus is that either compared to Bass control, or probably especially Audy, the end result is superior.

Your 3700H will not support Dirac, and probably not right time now to jump to 3800H until we have more details about 3900H that seems to be just around the corner with some potential enhancements. Audy can be really good, and lots of people around here have enjoyed it for decade or so. But then things move on and new things come...

Yup. Next step Umik + REW. I think that's good enough to get a bit further for my 2.1 set-ups. And then when I have the time, patience and spare cash I'll re-think set-up of my home theatre incorporating DIRAC.

 

[Snarfie]

Just a regular non-tech trained hobbyist. I frankly listen to my equipment more than the music even though I do love my music.

ASR has helped me so much to get good value equipment that favors objective acoustic science and neutrality over everything else. Cut out so much fluff, wasted money, time and frustration.

So I've got a baseline set-up of decent ASR recommended systems for my office and bedroom with a legacy home theatre system for my living room. It's a happy mix of Wiims, Ascend Acoustic, Genelec, REL, Kali, Nord with legacy Triangles and Denon AVR3700. I thought it was all good just using the built in Wiim Roomfit and Audyssey software for room correction with low level attention to speaker placement and minor acoustic paneling here and there.

Over the past week on a whim I decided to use Gemini to critique and improve all my set-ups including the stock system in my car. It's been pretty mind blowing what Gemini can do and improve on for non-tech hobbyists like me. It can read pictures and videos of my set-up and sound output. It can read pictures of my app settings esp the Wiim Roomfit settings. And then make detailed recommendations to improve on them with explanation for each recommendation. For eg, I thought my office set-up sounded hollow and thin on the mids and boomy on the bass despite Roomfit. After about 30 mins of chat and info sharing with Gemini, I got a whole new and much better sound to my preference just using the Gemini recommendations which went into detailed parametric EQ setting , cross-overs, gains and pre-gains, speaker placement, acoustic panels placement recommendations etc etc. Pretty amazing and a great experience for non tech people who have always struggled with room correction after we get the equipment baseline sorted somewhat.

I then moved on to do the same for my other set-ups. Same big improvements to SQ. As least to my ears. I also realized how ignorant I was in this hobby about achieving SQ upgrades. I was stuck in a rut of just using in app software where they were available. So much more could be improved without even spending single cent more if we get the right help on understanding basic principles of room acoustics and DSP sound correction. Gemini and I am sure most other good LLMs turned out to be so much more capable then I'd ever imagined.

Do I think this path is for everyone? No for those who are already knowledgeable and proficient on DSP/room correction. Do I think using LLMs will get you to ideal end state for the equipment you have already? Probably not, even though I was mind-blown using Gemini it was clear that there were some gaps in its recommendations and reasoning. It can get a little too positive and sometimes forgetful or mistakes can be made reading some of the pictures. So a little trial and error and to-ing and fro-ing is needed.

I would say I was achieving maybe 5/10 max 6/10 potential of my system and just a first round using Gemini probably got me to 8/10. But I think a lot of hobbyists like me will get similar gains if they try. What have you got to lose other than a few hours to talking to a always positive and enthusiastic AI and tinkering with your equipment and learning so much more about acoustic science in the process.

Highly recommended to all those who have not already tried it.

EDIT: Tried to do same on Claude. Not good at all. Gemini is way better. Even though I think Claude wins Gemini on many other AI tasks. Gemini is much better as of now on audio visual analysis.

My experience is pretty much the same. Speaker placement and DSP were the biggest game changers in my system far more than endlessly swapping gear.

At the same time, you start realizing how much depends on how the music was originally recorded and mastered. DSP and placement can improve a system massively, but some recordings are simply limited (or defined) by the intent and character of the production itself.

Take early Bob Dylan records for example: no matter how perfectly you tune your setup, those forward, sometimes almost abrasive vocals with the instruments sitting further back will never suddenly sound “subjective modern audiophile smooth” without changing what the recording actually is. And that’s perfectly fine the roughness is part of the charm/charakter intent.

A lot of recordings are also made to work well on car stereos, average living rooms containing mediocre acoustics, and everyday consumer gear, not necessarily for perfectly linear high-end playback in an acoustically treated room. Sometimes a ruthlessly neutral setup just exposes the compromises more clearly rather than making the music more enjoyable.

Which is probably the hobby’s most inconvenient truth. Sometimes the biggest upgrade is learning what not to fix.

 

[Cool Runnings]

My experience is pretty much the same. Speaker placement and DSP were the biggest game changers in my system far more than endlessly swapping gear.

At the same time, you start realizing how much depends on how the music was originally recorded and mastered. DSP and placement can improve a system massively, but some recordings are simply limited (or defined) by the intent and character of the production itself.

Take early Bob Dylan records for example: no matter how perfectly you tune your setup, those forward, sometimes almost abrasive vocals with the instruments sitting further back will never suddenly sound “subjective modern audiophile smooth” without changing what the recording actually is. And that’s perfectly fine the roughness is part of the charm/charakter intent.

A lot of recordings are also made to work well on car stereos, average living rooms containing mediocre acoustics, and everyday consumer gear, not necessarily for perfectly linear high-end playback in an acoustically treated room. Sometimes a ruthlessly neutral setup just exposes the compromises more clearly rather than making the music more enjoyable.

Which is probably the hobby’s most inconvenient truth. Sometimes the biggest upgrade is learning what not to fix.

Well said. Agree. It's been many hours of fun tweaking (vs frustration previously) and trying out the settings with different genres and recording quality. That's the difference, AI helps people like me have fun tweaking with our current set-up without needing to buy new stuff. For sure, AI is far from perfect and I have to battle with all manners of AI hallucination and gaslighting but the experience is now fun and I get actual progress especially knowing how to tweak with PEQ and equipment placement.

I first managed to tweaked my SQ to fairly neutral and then moved towards a warmer more fuller bodied and bass heavy sound for my mostly low volume listening. Gemini tells me that I am following the Fletcher Munson curve haha.