I asked AI to design a bass cab

[Admiral Akbar]

When AI can harness the knowledge, EXPERIENCE, and EARS of people like agedhorse and Jim Bergantino, maybe it would be worth asking AI to ‘design’ a custom cabinet for a specific sized room and bass and bass playing style.

Until then, asking AI to “design” a cabinet is just a waste of carbon footprint.

 

[waveman]

I remember the last time I remember someone doing this. They built whatever AI told them. It worked, Was pretty. Terrible efficiency for a bass cab.

 

[IanStephenson]

given the large number of published designs this should be one of the easier tasks for AI to do. If you googled for appropriate cab designs for a specific driver you'd quickly find the sample designs from the manufacturer or other designers. It's safe to say that those designs are in the training data, so when asked, the model should be able to come up with the answer its already been shown. Of course it has no understanding of what those designs mean, or if they're good or bad - just that it has seen an answer to that question.

 

[micguy]

There is a 3 way tradeoff in the low frequency performance of loudspeakers:

Size

Low frequency extension

Efficiency.

Cabinets that will do justice to the fundamental of the B on a 5 or 6 string do exist. They're not horribly efficient, but with a few kilowatts behind each one (tens of kilowatts behind a modest sized array of them), they do low and loud very well. Those are called subwoofers - not the kind you can buy a GC, the kind that tours use. They are ginormous - They don't fit in your car.

The kind of woofers that are used in bass guitar rigs give up low frequency extension to get decent efficiency (loud with only a few hundred watts), and small. They will not reproduce the fundamentals of your low B string. But...they fit in some cars, Not easily in some cases, but in a van, you can get a few of them.

If you want a woofer that fits in the overhead, you will have to give up even more. Doesn't matter if it's a human or an AI doing the designing, the laws of Physics dictate the tradeoff.

This is not an area where some unknown optimization is there to be had by using AI. Predicting the stock market is a better place to direct that kind of intelligence.

 

[DrummerwStrings]

An 8 1/2" deep cab with that height and width is going to be pretty unstable and very likely too constrained on internal volume, IMO. I don't think a high performance 112 made to fit in an overhead bin and work well for a 5 string bass is probably actually realistic but I'd be happy to be proven wrong.

A friend of mine is an aspiring commercial cab builder, and has been sending me AI design suggestions regularly for me to shoot down. I've yet to see one I'd want to build, but if enough of you guys keep training the models I expect they will get there in a few years or so. The trick, as always, is knowing the right questions to ask.

I’m more interested in what questions were asked to which AI, & what data sets were used by the AI. Could you include your questions that lead to your designs?

 

[DrummerwStrings]

There is a 3 way tradeoff in the low frequency performance of loudspeakers:

Size

Low frequency extension

Efficiency.

Cabinets that will do justice to the fundamental of the B on a 5 or 6 string do exist. They're not horribly efficient, but with a few kilowatts behind each one (tens of kilowatts behind a modest sized array of them), they do low and loud very well. Those are called subwoofers - not the kind you can buy a GC, the kind that tours use. They are ginormous - They don't fit in your car.

The kind of woofers that are used in bass guitar rigs give up low frequency extension to get decent efficiency (loud with only a few hundred watts), and small. They will not reproduce the fundamentals of your low B string. But...they fit in some cars, Not easily in some cases, but in a van, you can get a few of them.

If you want a woofer that fits in the overhead, you will have to give up even more. Doesn't matter if it's a human or an AI doing the designing, the laws of Physics dictate the tradeoff.

This is not an area where some unknown optimization is there to be had by using AI. Predicting the stock market is a better place to direct that kind of intelligence.

…Or folding proteins

 

[Passinwind]

I’m more interested in what questions were asked to which AI, & what data sets were used by the AL. Could you include your questions that lead to your designs?

I think you’ll need to ask the OP that, I’ve never used AI for any of my DIY cab designs, unless you consider Boxsim and various other software to be AI.

 

[JazznFunk]

Predicting the stock market is a better place to direct that kind of intelligence.

And I wouldn't trust it to do that. You cannot time the market. AI might be able to identify and codify trends over years/decades, but as the fine print says 'Past performance is no indicator of future results.'

 

[gmarcus]

The second link asks about a more optimum cab if it is bigger. It suggested 2.5 cu ft. Then I asked where the sweet spot is if I want the smallest cab that can handle a 5 string it came up with 1.5 cuft in line with a barefaced mini and super compact.

I was impressed with the speaker choices too and the fact that it suggested a bump from 80 to 120hz for punchy bass.

I'm pretty sure if you follow those links you can also ask follow up questions. And it is free for the user. Vented speaker design is pretty well understood so the AI can crank through a lot of options which gets you into the ball park.

 

[gmarcus]

I think you’ll need to ask the OP that, I’ve never used AI for any of my DIY cab designs, unless you consider Boxsim and various other software to be AI.

I used Perplexity. If you follow the links I posted you can see all of my questions and the models answers.

 

[DrummerwStrings]

I think you’ll need to ask the OP that, I’ve never used AI for any of my DIY cab designs, unless you consider Boxsim and various other software to be AI.

Oops, was intended for OP & now see that’s not clear - proofread, ppl!

OP, please post your AL questions

 

[DrummerwStrings]

I think you’ll need to ask the OP that, I’ve never used AI for any of my DIY cab designs, unless you consider Boxsim and various other software to be AI.

Oops, was intended for OP & now see that’s not clear - proofread, ppl!

OP, please post your AL question

 

[Johnny Fingers]

Should also fit underneath the seat as well.

 

[JazznFunk]

I'm pretty sure if you follow those links you can also ask follow up questions. And it is free for the user. Vented speaker design is pretty well understood so the AI can crank through a lot of options which gets you into the ball park.

Just post the collective text here so people aren't baited into feeding the AI machine. Nothing's free with AI... and everything is stolen.

 

[fdeck]

Take it out for a spin.

"For entertainment purposes only."

[This invalid link has been removed]

 

[agedhorse]

given the large number of published designs this should be one of the easier tasks for AI to do. If you googled for appropriate cab designs for a specific driver you'd quickly find the sample designs from the manufacturer or other designers. It's safe to say that those designs are in the training data, so when asked, the model should be able to come up with the answer its already been shown. Of course it has no understanding of what those designs mean, or if they're good or bad - just that it has seen an answer to that question.

There are a LOT of terrible designs for AI to train on, more bad than good IMO. Therefore, it's more likely that you end up with a pretty poor design when everything is factored in.

 

[agedhorse]

Out of (morbid) curiosity, I modeled the recommendation of 2CF tuned to 50Hz just to see how close the math and the word salad agreed... this isn't exact math but plenty close enough

Average sensitivity is ~99.2dB/1W/1M
F3=80Hz
F6=50Hz
F10=42Hz
Mechanical power handling is about 200 watts RMS above 50Hz

So you can quickly see that the AI predictions are way off. Probably because it trained on poor, unrealistic designs that builders who are clueless have bragged about enough that it becomes training material. It's really pretty junky.

 

[fdeck]

Based on the model that I posted, it suffers from what I call "small box crisis" which is that the port has to be too big to fit in the box. This isn't a solvable problem without compromises. Maybe the second design is workable.

 

[No such thing]

My requirements were light, loud and small enough to fit in an airline overhead. I asked for a 6o to 3.5k hz frequency response.

Invalid Link Removed

The design looks pretty good to me.

Comments?

For fun, I asked Claude (4.5 Sonnett) your question:

Final Design Summary

Parameter	Spec
Drivers	2× Eminence CB3010N-8
Wiring	Parallel = 4Ω
Power handling	900W program
Box volume	~26.6L gross internal
Tuning	~66Hz
Ports	Two 2" diameter × 3" long
-3dB low	~62Hz
High end	3kHz+ (per driver spec, usable to 3.4kHz)
Sensitivity	~97dB/1W/1m
Max SPL	~120dB real world
Cabinet size	17" W × 14" H × 11" D
Est. weight	~28-30 lbs
Construction	1/2" Baltic Birch, internal bracing

---

A 212 Neo cab, 28-30 lbs, hitting 120dB, tuned to 62Hz. Fits your amp, fits your five-string, fits in the back seat of most cars. The CB3010N-8 at 94dB sensitivity is actually slightly more efficient than most comparable drivers in this class, and two of them in a well-tuned small vented box is a very strong answer to your requirements.
The one caveat — at the box volume per driver I've calculated (~13L each), you're running on the lean side of optimal for this driver. WinISD would let you fine-tune whether a slightly larger box (say 30L total) gives meaningfully better low-end extension or whether the current numbers are the right tradeoff for the size constraint. Worth the 20 minutes to model it before cutting wood.

 

[DrummerwStrings]

The second link asks about a more optimum cab if it is bigger. It suggested 2.5 cu ft. Then I asked where the sweet spot is if I want the smallest cab that can handle a 5 string it came up with 1.5 cuft in line with a barefaced mini and super compact.

I was impressed with the speaker choices too and the fact that it suggested a bump from 80 to 120hz for punchy bass.

I'm pretty sure if you follow those links you can also ask follow up questions. And it is free for the user. Vented speaker design is pretty well understood so the AI can crank through a lot of options which gets you into the ball park.

I’m not going to ask the AI any more cab questions, once is enough. The energy & resources used isn’t worth the sketchy cab answers the AI generates; the questions asked the AI are definitely more valuable to learn from.