# Resing Modeler AI error **URL:** \ **Category:** Recording\ **Tags:** resing, support\ **Created:** [October 6, 2026, 9:39am UTC](https://forum.ikmultimedia.com/t/resing-modeler-ai-error/1300 "2026-10-06T09:39:48Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![arfran](https://avatars.discourse-cdn.com/v4/letter/a/7c8e57/32.png) [@arfran](https://forum.ikmultimedia.com/u/arfran)\ **Post date:** [October 6, 2026, 9:39am UTC](https://forum.ikmultimedia.com/t/resing-modeler-ai-error/1300/1 "2026-10-06T09:39:48Z")
I started experimenting a bit with ReSing v1.0.3 last year, but only recently began seriously using it, especially the Modeler. I first updated to the latest version (1.1.3). I gave Modeler a clean set of ~30 wav files (~14 minutes total audio length), all mono, no clipping, no artifacts. I set the appropriate parameters, including the exact vocal type and range. Trying to train on the files using quality “high” or “highest” always resulted in an “AI error” after 1-1.5 hours training. I checked the wav files again and again and made sure each file had at least 3 seconds of audio. I had the Windows 11 Task Manager open next to it to check system resources: no peaks, no overflows, all dedicated memories (RAM/VRAM) were well within limits during the training. No other software was running. I finally discovered that closing and restarting the Modeler after an error reloads the same training session with a “resume” button, showing the red progress bar where it left off. So I resumed the training after each error. After 3-4 error-resume attempts and some 8 hours of training, the model was finally trained and finished successfully. This is of course a bad work-around. A much better approach would be to close the AI error dialog, then presented with a resume button directly. Also, the training progress indicator (red bar) doesn’t show percentage completed, nor does it show which file is being processed. Showing the file being processed might help isolating potential bad/difficult audio files for the AI engine. ---
**Author:** ![tombuur](https://avatars.discourse-cdn.com/v4/letter/t/e480ec/32.png) [@tombuur](https://forum.ikmultimedia.com/u/tombuur)\ **Post date:** [October 7, 2026, 11:43am UTC](https://forum.ikmultimedia.com/t/resing-modeler-ai-error/1300/2 "2026-10-07T11:43:56Z")
I bought Resing when it was launched primarily to make a model of my own voice. I am on Mac, but bought a second hand NVidia card to run the modeler on my older PC. It took several attempts to make it actually produce a model. Usually it hanged somewhere in the process. The final model based on 30 minutes of songs didn’t sound like me at all. Haven’t tried again for a long time. Has anyone had any success actually producing a model that sounds like the material it is trained on? I used the same 30 minutes of my voice with ACE Studio and got a better sounding model, but here it had an East Asian accent. ACE Studio, however, doesn’t run locally on my computers, so there I need to pay for an annual subscription … BTW I don’t like IK has set a limit for the number of models I can make. If it required their online computer resources it could be justified. But they have set a limit on an app that only uses my local resources. At least for the Max version I own they ought to remove the limit. ---
**Author:** ![arfran](https://avatars.discourse-cdn.com/v4/letter/a/7c8e57/32.png) [@arfran](https://forum.ikmultimedia.com/u/arfran)\ **Post date:** [October 7, 2026, 12:03pm UTC](https://forum.ikmultimedia.com/t/resing-modeler-ai-error/1300/3 "2026-10-07T12:03:27Z")
I have the unlimited version, but I read about users wasting slots while trying to create models. IK should certainly free slots when a model is deleted. The model I created based on ~14 minutes audio material actually sounds pretty decent, but I found that when replacing a voice, ReSing can struggle to translate certain notes. In one verse the a note can be translated perfectly, while the same note in the second verse sounds like a muted synthesizer. So the result can be inconsistent.