Correction, and a reproducer with counters

I need to correct my own post. The central claim: native slider drag jitters, script writes are smooth. That is wrong. I was judging both by eye. The script path only looked clean because the ScriptPanel handle I was watching spreads the 0–1 range over ~676px, while a rotary knob's pointer covers roughly 35px. The same value wobble is unmistakable on one and sub-pixel on the other.

So I replaced the eyeball test with counters. A timer drives the parameter through a strictly monotonic ascending ramp, 400 steps, no mouse involved anywhere. Any decrease is by definition a second writer. The script also counts callbacks against writes, and checks whether the value handed to the callback agrees with getValue() at that instant.

Conditions: 48 plugin parameters, a panel repainting at 60fps to load the message thread.

HISE: 402 paints, 401 callbacks, 0 backward, 0 skew.

Ableton 12.4.3, VST3, same build, three runs: 555 / 562 / 567 callbacks against 401 writes; 153 / 161 / 166 backward; skew 0 / 1 / 0.

The excess callbacks and the backward steps track each other almost exactly — nothing is lost or reordered, something is injecting extra writes. Every backward step is exactly one ramp increment:

cb 0.007500 -> 0.005000
cb 0.012500 -> 0.010000
cb 0.032500 -> 0.030000

The host returns the previous value while the script writes the next. Skew is 0 or 1 across runs, so these are real writes landing in the store and notifying normally, not display sampling artifacts.

Repaints lag the callbacks 132–147 backward paints against 153–166 backward callbacks, because repaints coalesce. Visible jitter is milder than the actual value churn, which is why this was easy to under-report by eye.

Then the same parameter driven by a host automation lane, same build, same host, same session:

Ableton, using automation lane, ramping 0→1: 1000 paints, 1000 callbacks, 0 backward, 0 skew.

That contrast is the useful part. Automation has one writer and nothing to echo. A script write is a second writer, and the host answers it with a stale value. That reads as the a92a701de mechanism directly — a listener can't distinguish an echo of its own write from a real change.

The magnitude varies run to run, which is what a race looks like rather than a logic error. It also explains something I flagged as unexplained earlier: one early Reaper session appeared completely clean and never reproduced. That was the low tail of this distribution, not a mystery.

Consequences for my original post:

The two-path asymmetry isn't real. Both paths are affected. The native drag is worse because the drag base gets perturbed and the error compounds, whereas a panel recomputing position from event.x each event can't accumulate. Difference in degree, not kind.

I pointed at MacroControlledComponents.cpp:735/752 because it was the only differing code path. With the asymmetry gone that nomination is unsupported, and I withdraw it. This looks like shared dispatch, not slider-specific.

Reproducer below as a snippet. Press Sweep, wait ~6.5s for it to stop itself, read the two totals. Results are drawn on screen because Console.print doesn't exist in a compiled plugin.

If you get 0 backward in AU, that's a real result — it would narrow this to the VST3 wrapper, since AU and VST3 don't share the parameter dispatch path. Worth knowing either way.

HiseSnippet 5518.3oc0c80aabkcmzIL+QY2hcAVTTzmtPOrkZMEM4LCojiiSrrrRn2XKqZJa2hfT2QbtTbrGNCwLCEE2tInXet.8khBz8aPeoeB5SEneQJ5Gf7beo8bN26cl6PNTVilDfNDIRl2+btm+cOmem6cn3IgAi3QQAg0pu0oKmwqU+m0X3R+3IGNw10u1ieTs5+oM9stww7vW+b9rv.m4if+4KMp8vkyrih3N0pW+89Jbr0+32uF85G9hGZ6Y6Ohm1TsZuLvcD+ItSciSac1C9ZWOuuz1gep6TsQa8fGOJv+v.uf4.e8dM5Tal8n2ZeN+XabX2pQsA1QSpU+2znaGCGduQV7w1NFF6Op29VbqtNV62qSGqwFcFYae28r26rZ0+fibbiCBGFaGyipU+8eXfyxgSBV3KVfW5F4dlGGeS2ZCgUVz7WF34fhH1ZsCm354bhRgEUCnxIopu2Sn99UMdpqiaR6opweA0AKcF5Jv52JK68dYXut5rWGM1KGVptFK89BV5W1X3nP2Ywo8f7ymz3w9fUcrMXmzYEwXqcq+66z3N2gc+eLesEPP1fGO7H1Lu4m65ylYGBVTfKXugbwX+W+8+SrvD2L1EFrlt9QwgymxAl0Yms9Igkd0fCNkc3fCN9qN5QPCTamNgylExuvMXdD6BdXjafOyN5sbG1xf4r3.1qN3zCGvFGnX91vbriYKriX1rotQw1uk+ov+Tp7s84drI19NdbZAh.pa6DwhgEpytcYg19myYAvRw99960e1krEfwkCDHLH1NbI6s9Am8WDwlE3h1M1HbnQru2r2rKaInX.QrHPkxtv1aNmsH3Lvqg4FwBNSHIfPD3CD02gEM+rcm4dIvUPi3DCfe.RwuctCXaNmH4YKY7knhfLHN.yL11KBlezxofgC3pla6aG6dAm4DZetTSD0hEQBMaQnK3OyrCkx7z.XQ1dm1rGB+FL+wSnNY1iGyGAF31ope2T09gO6EGe5PF5I.pr1rCXwP7hPXIgEVnAS8jhmDFL+7I.qB9Mtih8VxlF3GDG36NhHsczHtuCHgfJe5L1BWfS7CfAMGDLW+KB7t.kT+kK.0AuEpVg2vb3i.CVDoMOaIQIG9XWe2XxyfEwgXVNBIFzhONlAZp.vLM2ONhMx1y6LHHFHtmaiBhT0zBMEDwFMgicCKJZFHgRXDQeFfghEVWEg.5DxAgmX+y4wuDGaycX1wD0hQWQbcr8iSTpOmGM2CXFXGVPHnsQCOXn3b+Vry3irQEvgA9QAd71yBA2LlS.rDfxiwuDbnABxrELavzYfyoibibxJv.CgKJuGC9NHWB+lsKtfTyjzxfvZronJE7Tm4QdUdAKvs6j7BrN4OJIXtuR1YJorbriNKB+U1EGLKwgAdolfU3BGtG3FEtJebUKuhES4CEwkyBhSrPvHIu1UyGf7rHBjt3BgG4rTq4UwCNtQjK.v1wDolFfAIYAiUDcoxp7ziNX3Kd9QOh0jB91ywzp+98FC91ux02IXQDqa2VLq8WKrLnpNadzRnAeJLwNeZpgVDViEsfym0RDUekWVcLXJ6tUmtn0oEqCYZvei5nbo0APLKX6JqqQaq1lrWN7TSVSXaMmyBm6GckZkLu50qG6Nrd8MnetGx.W641smILqt86R+rujqI6ZGrI3+6b8IlIxBxeZsGQrE1gNR0ijNCBfsX1yiClZSwT.LT7TsAQmNc5jnRo2juNkH3QnK.rwEy4mF1vU33pnJoxwPZ1oLEDkclXOBW5Ewx1INd9k1TrULaBEG00GhjfdfeJs.SPgIjGOOzOJyVEhbxzSTBNJmUlzEXK97KiaKz2vp0gAoY6xrGEF.BC5EzRltKRrQESh.gm8Tj.zcTHdo3FAXqDatffYtiWhc4GDNEzJKaQA3fsSy7rEBKEWBGhcHLXPNifDVYMYHO.wLg3Eoplz82xfoDwTw9f3P1dbH2CDI3BAbNEpGP.mB33nn9x.KvhNGjFgdZzDPK1VGYBESyFyhfg+oIj0uAoIo0QzBPFyzdm.3Sd1wGISTITdB8xDTjgzL7QSBvjr5FEgKxviN7YG+H4TEtHI1ZaevWJTXdckLFlCFDZgXjH..MN84O6ICkt8mxgYqhQlMW9beRo.CnM6QH9B23Vnm.36xFRdtDD.n41p8PD.K1qty.JsSj6uS3fEHfWIvewvvisYG+rSWOhmhPBxK4Jgl.B0lfjP3y2gs6myfPmee+18POx3fYQ.yDw8FSJGQh1DZdP3zzrA.ywiE.w.3cwBfBnE3UO9zAO6EmRhFYQTHADPQHscdQdV0CXjM5gRSWAJ4jPLV.g.AMzB1qcdDCz1KPXrYkWIPQh7fYXazaeax8ka6C7J3eKD1ebwnuE.JIFhrzdJ.o9KCg2jT5Ry852oEauNc14dagK7t6B9Py8sgcWQ3aJyqsfcYfq4E1grSN33idxqeEpYtOtZ2as9FH5qqUl9N9EO80mbvyO3oCg9r1+d3f.tLFPz6kW1VHHpWagEWFh.hupQuG9hg+0u9IO93iP5YXz4dB5MEz6P4w6hIIsc.TLvOlA6bTniznvvWczQm75gmdzI.E5ztSGidRhXAoSvn6QHfht8mFACfbrSQLSpXbyiZiDt+pkXSnOVr.x4wKBXQdtNX0ITDxsv0ULgWgrsUaPGk11.Ts0lDkjWvxLNDhAhNy.fFA9iIb2ymDmXlSKLT3zGg0NeCr4Dq3EbNVzCSXEohbtmpim35yiDc7MeK3lgMS50CwsrXyR4gZDSTvVswmXSw3HMtjBiNSNc8AO5L4zy1nb5oTPz7PAJS0XoVoD5uLyfS1WLgbmxM9J6lsuHxygBeeelZKpsiyWCEp1baYWa2hYzA+eXCpro1PfulaGCI3gN2lFT1NeNFqtIv6sPGiVnTr5rgRurgZYHbxamyHD.n6zt2pSzM5DZa2IJk.LavMhu53lkcT3IOoysJcJ.laJBsHm8xSbcbfcEWScJPum4C3oDUF.kiBg5RRttaDrcRyt44Rfqrwe6hvNVR6DcviffuTUtsOBeCn0bnzRpJgPLCyTHQrvujLTQRuZrIWPtvyynoK4PwbYeldDrcYcg1t8s2Yq+tsP2KbJuMOSOP5sY2l4B1E3+.sEN52JzqKbchm.ZxtYaVr2d81kgAgNn8j5cluWRR25d.ZD7Jr+ZiZCV+Sf3pnjAJnaCbpbNJMY6Yyilz7sPqeG4e7Ufux7PwwevhvZyFGFLUGuKd.HfAJfxret9vQHohD+gtXxr1Za5lOaBn0OxG3Pda7PIh4uHhGdBgnP..JrolO5JArwvbMyE6yNrBruePR.8715K6EMn8z19KaNaD.ElssWaLBSrAYhMTV5bnzJQCvLLqMJg+frq7DkAWonL.o69qIJCxUTFr8ZiIwasWp6Z2bnzJhxJx6fTIQzyVt9PkJb1349xrk9RYtIdFMP3de.xLAeTssMMSL0rzG12CAX2VhXnovIdSDev6f3CtNDWKqjX6vtRbDRW0aNZM45k0dRG.KryVzmLsTWSTIp3uLAnjP9Vs6j.URXeq1ODLNXwgpJAQehC8boCcZ.c1t+ZpPFDqAkFQatmf5lmGLGJBf2Toxadtdz1EfHAkoE09bd7qP9roLDDglRqyADWp26EPupDboGvDvA3.n0WbeKM6bY28+R5kb1m2drqm2y4iha9McnP5KfjaeqZtfYT50KNQaUQ5ncc7bOnhbjQk0tDgUbmTUKURFTqb.feUQqHLoEBNFOY5HD72TPugn7nB1N5SIYTUtNiJkQ63HVLg.gpnFeI+LbEQ2UHRZDdRsxyeE4O7H0oCrepajGv6fUARgFNW1MURFUSXhNFs.xMPZZdT0Og8aTd+o8bIBB6BnmlKfzmSVnYPVhc0bB17jcfQ.Ypx0ZfVhCNvzLGqwksXKASAZNRsGqO6L1RGnvtC7bO2mC9BWF2TlFwIXNjk4zfgwgfRn4EsXV6zhg1ajCMrHitQmuE7mgZAhC4NaqVP2wrlR3y6PMH7WSNLXBl6susbzpYbAfsHEZ7tXfN9t2cmjwjRjDBgPiyPGEsRwc+Y2m0emsVsXVEFdQB5sogynD44K8I7UKfZvn1FKyeyC+BbXREL9561JqBfjPvMPLhuaquK+31CE6N2Xn0qPOKqkXckLsQ6yTUP7tzxhpOxUEKKL4ZoeGcF05lUXB14ZpbEm7wlTvRIKIcSFw+o.7p11mE0b8.e3gvQJW1mulRQPWr3JTUnsf4mGSqrgCEW0fJ7eyDi5ZnQVejxNWKa+lF4fURe7T79qRFUR9CHbperxEJETp3PmffcKmwYFf96EOlo434BEFdIqiVNiDWPhfsGgI0vKjUov.Po.W36jfYVtPMMnTFRH14SE1u+2yDMfWh3pN2WYhO8zaRhdIFRUhwgBqtC6Nx3uIp30BEgEOjwA3hjpmyNTxcY8g1MyPWqPzKzXWUmh7eNfjHrLIuWbfzQZ2omW.dS.P1UnRtnjpB0cMS0oz8X9hYWaKiptpuSqhg7NOoazInAzaXLn0ivChGKszVbgRifTrPoPpr3Ms8s8VFgmo7Q+kPmgWv2ocV.cODm1F.zg8oNmg6JAzgscE.5R6NsxSBKndm+zhHKNsHNn+WLCKMeSHx5TDDYYmaeXt86e3gpdoZdvKw49.yeGsSTLkydibnzYA7FwYA7FXSh1XYuAOGf02nN6RX3uA12gKQ1snyVJQIAaIgv8Tv4HW+lwPSls6.M8FpytFHbHw.sLRog.6Blno4L.4C9+RpAuYopTIsTrZVR74KJb8fiJKIYjzhpKiU9n.5NBjcjTVAUEOdNMzs7uHXtmC3dO1CO9Qw8Wfil.OR2tujZN.P24yjW3kxyO4DATWdDF6ExL45XSWFvY3yIwX2ywCFn8JhEFAmDrlc6qs4MveWwk5KOeeV4O9aXIetfVaXGnnW4dPKUQUhVuhcg5CHYenQO09vzt+oYmXoq8YiXsSVXeXgSfG4w8OOdRJWEpuOKTrOKD1m4C+Rs8JeT6JR9Mg.h7uoKVIKtAJD1OzcOfIQL6Xi6g.183ioyrLWbmzgFcnrPmjMDZGuMczxYvfyVoMwoSKSWpAGM4foSwVxVosjC1NyrSNW6jQlnBGAkoEprrotHW8wVH1GrlPlPTndPdXbyNXwMhmfDDOpRJPHphmmDQqjbPMR2NsnQjk24dajrxq.FGqlpMKkSzuuCoStMOm6jTbY92jbz3iLl3ddoijzVBxfvKgOCHwBzx3kXCijBZYKdHrDfUjHuvp7Q5MgtY3P9TvwdjsONB6Qh.hDH.+kvr72UiygrE5I7OK1WbArYh27v4wvvatspWHXgUW0QxoZL6AxIFW1S7CTr4P7zfYJZKO7ZUKqe90Y6IyQXmsq0OE6LC3cjcRdkN2W8Ots1k3oAoV16me+TDpo4lSnQW01rbwoJF1UBVcqzGBJHdJdk2hbPZ8naJjGCed8lA.r5V+bwS.Ov+gxAAyrE9bhLZh5xzSni3gKISAZWs7jirrgRwSV8qpVbYCaPO24J0yo2Uw6PIuR34zNTWX48oRqtWAqCK47IR7BWALgPoh+jCNrqdrNoyJGq+0rdi0ED4stlJ5YiZmfsaqL60Wu9Xk0SCSjc3zLmnG83o.vIV+grf8NCV9Ddbj7wvXsm4hUddKNN8VczdpnBVHdjnltRDO7wCYSw6f9fnG85oEsCZJarNbLazeF5r3dyaz6KMyc9mpbdm+0UMtzqeW2ediNf2PeFoFKWOFnOT2k4wes41O6qYpGRKsGKcrn3kzypNl2V6FRwD5qcqZfMQyTbbPL+YhX6eLvVrU6Z73b6SxyzM6kS2xMgadhM8mO8LLachgmFXs5ue1O3AM17G7.8OWDiDNoZCLv+w9twOaF2eSeZIpI8rqUqtkjqfgFSeDE9SjeDEFRO2H0bcpU+iaHcXpQrr9mkjZWio+AMlY6zoDysaIlqQIlqYIlqUIlauRL29kXt6Uh4teIl6cuYy8CIeiNkYxcKyjMJyjMKyjsJyj6UlI2uLSduxL48KyjKiGlQY7vLJiGlQY7vLJiGlQY7vLJiGlQY7vLJiGlQY7vLJiGlYY7vLKiGlYY7vLKiGlYY7vLKiGlYY7vLKiGlYY7vLKiGlUY7vrJiGlUY7vrJiGlUY7vrJiGl0MzC6m0H8loWkB28AEjBCxPg+1+w+i+yuXiTPTcKQgsZnpnu.xf17+nFhp6xY1u3wOxN1F+fQKK+.JIYFOL1Eq1o9i3W3NhK9XR+wMdDO5swAypUe+CUUJCMeMpP4Rnvo5+JX0Vp9GScc.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.2+A5b9Gc8dRNIN+Oq90hykb4e7eAe8u9EqvVC1.a8CegNa8gB15m2P6KLE4yYo7u7Fq7bV9GzM8+62ZUS++fj2x9WjuZ0+yaro+h7kw99AWI6f+E3XE1Y65WI67GRedVe2qwmzH4S++Jqx+1UuJ+O4rJMtNO2szp7G0s1Bi3G1Pz+0xRsUC0GM3b1SW7coWOgH4g+kVxe8sVSDZz.6s15ewH8KZ7z.m4d1wY+dZB+1rR1A5lq+kiD92cbev8eotS7OZe4MccYweYiSbiGMIed7V4vi0p8SBOJ+Ju5m23H5qumTF78a7k+U+z78aUMweDXN+o1wgtnA+34SGFLObDGVcevOD+LrVG2VHeeGkyzPtuC8l+W3krytpv9XmcUcVaJ9mvxWK+diA+R05inV.dxmxG7wMdJ9dVNeDEfPjtudznrjZsIZbSmn4MchV2zI16lNw92zIt2Mch6+tmH9UvV52EFv.d5IGQe9.pW+H5KM.xas1+G21UoP.

@Lindon

Sorry yes my typo...

Glad you weren't being charged 5% 😂

@shashwatshukla then it's not on your side and pitch module is broken. You're stuck until it's fixed.

@robJ Thank You 🙏

@dannytaurus said in Claude screenshot crashes Hise:

so I don't think the recent commits introduced it, they just make the message thread busier so the 1 s timeout trips more often.

Yes that's it, in fact it was crashing less before for the mere reason that the project I am working on at the moment is way bigger. UI takes long time to recompile, that's the trigger...

So in the end it just needs 2-3000 sec timeout, should be enough for most cases...

I'll test and make a PR when I have a minute 😉

"OK, that's it. Thank you."

@ustk have you attached a lambda? I forgot the exact syntax but you can add a void f(double) lambda that reacts on simple value changes

@LucaL check what's in that array, I don't know without seeing the rest of your project

@Chazrox said in RDGAudio GUI Father GUI Designer for Windows x64bit | Pre Order:

They could've atleast used a different colour scheme than the app I just launched so it didnt look like they took my idea. lol.

Colour Scheme.gif

I really appreciate your app looks stunning , Although You can easily change colour scheme easily on GUI Father too. Not always boring Yellow. Here is how.
My job is to share with HISE people. Actual developer is RDGAudio

@ustk As per @Christoph-Hart 's suggestion, updated the PR to use a noise burst instead of sine burst.

Re-ran tests and it all checks out.

https://github.com/christophhart/HISE/pull/1014

PRs for fixes here:

#1024 Make the "Default" font honour the requested size and measure consistently

#1025 Fix out-of-bounds read in the string width table for non-ASCII characters

@dannycouture Working here, although toggling the module bypass might cause issues if you have notes that need to be killed in a script. If so you'll need to script the action and control the Ignore all events buttons instead.

HiseSnippet 1133.3oc6YE0aaaCDlJNrsNsKacnaXOpG1CoCEC1VIoIXOTmDmjYr5jfnzr8VAC0YahHQpQQkUigAreJ6mR+IreJ8evFojbjThmmsP2FRm8CF5tim32c7t6i.5DofBQQBIxZkyFEBHqGgcGwUC2aHgwQc6frVE2iDo.ocppcGERhh.OjkUsCMJrpuLJ426dwtDeBmB4pPnyELJ7RV.SkqMr82w78Of3AmwBJr50a2kJ36I7EwZ7TC2.ERnWRF.GQLKaIL5aIQCQVeEF11g1ftkGc6MgMZAWP1nIP1.5usyEdWPWeKG51sZ034NHq6suGSIjtJhBhzuzcEdibGJ9Id5FbNKhcgOXDZhb06bpZzdCY9dmLN4DgPV2+j7TUszT0Sv8Xdrq0mmx9jDC14dTLoYszzfTy4.RVEfzxoP5wXWpjEpxsXvyCwc45Sv9D8YSQnjtVj0aw6IzKfq95.xkvARsv0Nr1lMZ7La8eO8a5GyoJlfaK3GITvw70d5J+7J0W4WVw9ll52eh1LaiT36CxIZ1TNHmliqwiCt.jOy9JheLb8B0ge4b58lsbJMMpKrPAuKmoNNDxjOP36YxUlmu8I.JKsoOKpkgP8SpjShUyNI1MVoDbDS+RpiSEZhRfewNGzb3dqwtisR7dNb04Ft9ptcHJhoTJKRzQWHHULSRzpCbkt4MsvpNtCDcoRDlDpAgBtItst+LDruIOP6zdz0B+54skDOl3PoHNLMCGNN210yTzZZh5EqqC0yAj5Q.5mLFd.NsF1D4yAPZU.Hucvb.jGkCjVuWPhSAj7appgDm+Jjb6YB5oQBuXehp7HJyv6LC5drRyELM+7HlZTwRz4XtUioN2ZVg3iwmvTzgSFiKMALpqi+m.iYS6+H7986CTUN.WFevOT0Q6y91uZ51+PrKiCIbyIa9mkHa+8jq.6CANHMIuovN+6u+YmOlpz69YRBOJTDU5E6BAryzMDQEU9pHM2B7imZNLKpeOAQNQSu6EGniwI5iKQEKSpJ1IPDyUkpBpU9XX4YiMnbCQg0OAp2k+O31.Ml8aCTabMS9Hzz6AjKWrW5lIrJekjBDhKMCzRqhYC3BIjopRrheLtO6MtpX5klKdDcyWB5tvDwRmc0GiQWVPnOrO+JvWSJmMRrCzmD6qtVaQP1SvEgCEbFsX7eJnjrAC.YQnOw3YGkRee6bMOo8ofOPJ1Y+kseotijH0oInhoh4+RsS735Kvov01Lkx9tKQVsOrIxNUDqX7A8H5hP8sdvGEG3pIPnfd24bvOxbWzkLSNRkaL9tot.2KQ3Oz+xL1zHakYr4Xiy.Y4mOIxxVKXKWvVNE1xR26uDJVPWd2ft7SuIcYqE7kK3KWvWVM9RmE7kK3KmM9RmE7kePvW5rfubAe4+64K+2XOBHTo30zzARl5zGjnQG27jO9YcbOircyaOAJPO080TZ4W0sbrUUczopNtdUcbip53lU0wmWUG25u2QCo3NwJQPZqIB06j8S9paVV6yI5p7jNBzeFiDq4A

@ustk Haha I am launching HISE with the cli most of the time and there it worked so I never bothered…

hise-cli -hise launch

Just seeing this, this is extremely impressive - also puts into perspective how impressive the FF offering is with all its extra modes and spectral stuff on top. Crazy work mate, well done.

@resonant is it an issue with the Neural node or the global cables? I haven't used the neural node much so I don't have much advice there.

I think the issue is as Christoph noted before: HISE tries to load the network in the hardcoded FX before the first script compilation.

The easiest way to recover your project would be going directly into the .xml and cutting the HardcodedMasterFX out of the project. Then reload the xml and resavee it. Add your Neural node FX back in. Also just comment out all your scripts first so the project can load, then slowly add them back in.

@Morphoice
Ah, no but that is a reason to reset everything!
Under many conditions, changing parameter values all at once can produce monstruous audio. Like, if the DSP effect has feedback, if you don't clear all the buffers, the audio from before the switch, could blow up under the new parameters. There are scenarios that can sound really horrible or blow out your speakers. I mean, let's say the new preset has different reverb parameters, you'll get all the reverb knobs yanked while it's still playing, and get all that washy doppler... Of course that's not a horrible situation, but you can imagine that certain DSP would produce really loud nightmares. Which is why it's quite sane for Hise to send reset() on a preset change, to clear out all the buffers and start from silence. It's the safest thing to do.

But I agree that creatively, there are situations where this is not what you want. But you can understand why Hise went with the blanket solution.

@Christoph-Hart said in HISE with Claude:

just fyi at this point the hise-cli contains the functionality of the mcp server and the lsp server so it's the all in one solution going forward.

so we'd just need to point Claude at the HISE-CLI then? (Asking for a friend)...

@tahinar Yes you need to build it from source. There are many posts on the forum with instructions.