Introducing ReSing - A plug-in and standalone app for professional voice modeling

[jamcat]

Are there any demo videos showing the voice model training process for creating your own custom vocal models?

[Cuauhtli]

I haven't found any. I watched someone do it once with a different program. They fed it audio recordings of the voice they modeled. The more examples they gave it the better the result. They ran it overnight. But that was a while back. Resing may be much faster.

I was just thinking, surely no one would try running recordings of Whitney Houston or Barry White through it...

Edit: Today I've been getting 'Ai engine disconnected, restart Resing' errors. Good excuse as any to reclaim my 10GB of drive space.

[DaveClark]

If and when I ever get around to creating my own models, I'm going to try Replay and whatever else is associated with it to practice a bit with creating models before using up the very miserly allotment of 25 models. Like others, I find ReSing quite disappointing. I've used Replay with one of the thousands of vocal models out there for free in preference to any of the ReSing models for a cover version of Louie Louie. The ReSing instruments are, IMO, a big joke. I think this whole business of making your voice sound like someone else's is misrepresentation of what is going on. These models, again IMO, are acting like nonlinear filters, not "replacing" anything. I've obtained better results with vocal processing chains than most of the ReSing renderings. Now I would agree that the process is made easy. It's really easy to produce something you don't want! Then you can move on more quickly.

[Peter - IK Multimedia]

New ReSing Japanese pack has been added!

ReSing v1.0.6 has been released on IK Product Manager.

What's new:

• Library update with ReSing Voices Japanese Pack
• ReSing Modeler update with Japanese language support
• General usability improvements

[rizetmusic]

Hello for those who encountered the "AI engine unreachable" error,
I found out for some reason Resing failed to launch the AI engine engine and I have to do it myself.
Locate Program Files\IK Multimedia\ReSing\ReSingEngine and run ReSingEngine.exe
then it will prompt:
Insert the desired engine ['ReSing', 'Modeler']:
type ReSing and press enter.
When it says "Starting ReSing AI Engine",it's time to open the Resing Standalone.
In my case the Starting AI engine shall pass successfully.

[Peter - IK Multimedia]

Note also that you can stop any rogue ReSingEngine.exe or similar processes in your process list (and similar processes on macOS in Activity Manager) if you have this issue and that also usually resolves the issue.

Also note that ReSing MAX does qualify for the 50% off MAX titles coupon IKMEMMAX50 we have going this weekend, I've been asked and made sure it was added properly into the materials/graphics for that too. Sorry for any confusion.

[Aloysius]

I tried the free version today. I first used Acon Digital's Remix to isolate a vocal. The result was pretty rough and sloppy, but I dragged it into ReSing and changed the voice anyway.

To my surprise, ReSing did a remarkably good job. Good enough that I went ahead and ordered the Standard version from JRRShop in the sale.

Next, I think I'll try creating a vocal in Dreamtonics Synthesizer V Studio 2 and then changing the voice with ReSing to see how that turns out. Synthesizer V has very prominent resonances in some voices.

I still can't stand the look of the product. The bright red branding and the "rock star" models covered in tattoos just aren't my thing. Every time I see the promo images, I swear I can almost smell hair gel coming out of my monitor. But I can't argue with the results I got today.

One thing to note: it seems to require an internet connection to activate the AI engine. That part was a bit hit-and-miss during my testing.

[Ploki]

nothing is real anymore

[Peter - IK Multimedia]

I tried the free version today. I first used Acon Digital's Remix to isolate a vocal. The result was pretty rough and sloppy, but I dragged it into ReSing and changed the voice anyway.

To my surprise, ReSing did a remarkably good job. Good enough that I went ahead and ordered the Standard version from JRRShop in the sale.

Next, I think I'll try creating a vocal in Dreamtonics Synthesizer V Studio 2 and then changing the voice with ReSing to see how that turns out. Synthesizer V has very prominent resonances in some voices.

I still can't stand the look of the product. The bright red branding and the "rock star" models covered in tattoos just aren't my thing. Every time I see the promo images, I swear I can almost smell hair gel coming out of my monitor. But I can't argue with the results I got today.

One thing to note: it seems to require an internet connection to activate the AI engine. That part was a bit hit-and-miss during my testing.

We borrowed some hair gel from a colleague in the plug-in space But seriously, your comments have been noted thank you.

Also, internet connectivity is strictly for challenge/response on the licensing and no processing of the vocals are taking place online for anybody reading this.

Another note, ReSing (the non-MAX version) is only $/€49.99 for a short time for anybody interested.

[Ploki]

i'd be interested if i could train my own models, there's currently no local tool/plugin that allows that unless it's a very unwieldy opensource.

i'm not too fond of uploading raw vocal tracks to the cloud.

edit:
i'll leave this here because it's on me that i can't f**king read.
can normal version be used to create your own models? i'm only interested in that.

edit2:
Highly optimized through Metal Performance Shaders

I see this is a RVC implementation

The Unlimited Model Generation add-on allows ReSing and ReSing MAX owners to create an unlimited number of voice models using the ReSing Modeler. It’s the perfect upgrade for users who want complete freedom to experiment and build as many custom voices as they wish.

so 150€ to make more models. How many models can you make with the 49€ basic version?

[Peter - IK Multimedia]

i'd be interested if i could train my own models, there's currently no local tool/plugin that allows that unless it's a very unwieldy opensource.

i'm not too fond of uploading raw vocal tracks to the cloud.

edit:
i'll leave this here because it's on me that i can't f**king read.
can normal version be used to create your own models? i'm only interested in that.

edit2:
Highly optimized through Metal Performance Shaders

I see this is a RVC implementation

The Unlimited Model Generation add-on allows ReSing and ReSing MAX owners to create an unlimited number of voice models using the ReSing Modeler. It’s the perfect upgrade for users who want complete freedom to experiment and build as many custom voices as they wish.

so 150€ to make more models. How many models can you make with the 49€ basic version?

Yes you can make your own models with the standard version too. https://www.ikmultimedia.com/products/r ... p=versions

Includes 10 voices, 10 instruments, 10 model generations, and limited imports.

[Ploki]

Thanks, bought

[jamcat]

Can you share the models you make, or are they locked to the account you created them with?

[DaveClark]

Not only did I find ReSing unsatisfactory, but the documentation on such details as sharing models is murky at best. I asked claude.ai which responded:

Short answer: **yes, they can be shared — but the recipient needs ReSing MAX**, and the model only ever works inside ReSing.

The decisive detail is buried in the version comparison table on IK's own product page, not in any of the marketing copy: there are separate rows for "Import RVC models" and "Import ReSing models." Standard ReSing gets 10 RVC imports and a dash (none) for ReSing model imports; ReSing MAX gets unlimited for both. A row for importing ReSing-format models would be meaningless if models were creator-locked, so sharing is clearly a supported path — it's just gated behind the MAX tier on the receiving end.

Why the documentation reads as contradictory: IK says both "Create your own voice models to share or license" and, a few paragraphs later, "Each model is unique and protected from unauthorized use outside of ReSing." Those aren't in conflict — the second is about format lock-in, not user lock-in. The trained model is encrypted/wrapped in IK's proprietary container. You can hand the file to someone else, but you can't convert it back to RVC, ONNX, or anything usable in other software. Note the asymmetry: ReSing imports RVC, but nothing exports out of ReSing. IK's press language reinforces this — models are yours "for personal use or to lease out for profit."

Some of the press coverage garbled this. StudioFest wrote that models "cannot be exported or misused elsewhere", which reads as "can't be shared at all" — that's a journalist collapsing "can't leave the ReSing ecosystem" into "can't leave your machine." Gear Gods got it right: users may share or license their own models, and the Modeler ensures models "remain protected and usable only within ReSing."

Two things I couldn't verify and would check before committing money:

1. **The actual mechanics.** I found no IK documentation describing the export/share workflow — file extension, whether there's an in-app export button or you're copying files out of a models directory, whether shared models carry any account binding beyond the ReSing container. The ReSing Modeler PDF that ships with the product likely covers this; IK support is the reliable answer.
2. **Whether imported models consume anything.** Standard ReSing's RVC import count is 10 and MAX's model *generation* count is 25 — it's unclear whether importing a friend's ReSing model burns a slot from some pool. Worth asking.

If your use case is two people collaborating, the practical read is: you both need MAX, or at minimum the person receiving does.

I followed with:

My next question would have been about the mechanics of sharing. I could not find one word about anywhere.

claude.ai:

I couldn't find it either, and I don't think that's a search failure on your part — I think the documentation doesn't exist publicly. But the search turned up a few architectural clues worth having.

**The feature is newer than almost all the coverage.** Every piece of launch press from Sept–Nov 2025 describes imports as RVC-only: ReSing MAX was announced as "25 voices, 25 instruments, 25 user models and unlimited RVC imports." Sound On Sound's January 2026 review discusses the Modeler at length and never mentions importing another user's ReSing model. The separate "Import ReSing models" row only appears in the current version table. So it looks like a post-launch addition — sometime between January and now — which explains why no reviewer, manual, or press release covers the mechanics. The marketing table got updated; the documentation didn't.

**Models are `.pak` files with an authorization check.** A user on IK's forum reported the app throwing "Model not authorized" while a `.pak` file sat in the correct folder, and IK's fix was to run "restore purchases" in the standalone app. That tells you the container is entitlement-checked against your IK account, not just format-locked. The open question — and it's the one that actually determines whether sharing works the way you'd want — is whether a *user-created* model carries the same account binding as a purchased one, or whether the entitlement check only applies to IK's own catalog. Nothing I found answers that.

**IK's pattern on import questions is to route to support rather than document.** When a user reported that third-party RVC files were greyed out in the import dialog and explicitly asked for "clear instructions or a manual on how this works," IK's reply was to contact support. If the long-established RVC import path has no public documentation, the newer ReSing-model import path almost certainly doesn't either.

One thing worth knowing before you sink time into training: there are forum reports of a ReSing update deleting custom models outright, with the user noting the trainings had each taken over eight hours and the plan counter had also decremented incorrectly. If you're going to invest in this, back up whatever directory the `.pak` files live in.

Practically, your best paths are IK's forum — Peter_IK answers there and is the same person posting the marketing copy, so he can be pinned down — or support directly. Worth asking three specific things rather than "can I share models": whether a Modeler-created model carries account-level authorization, whether importing someone else's model consumes a generation or import slot, and where the models directory lives.