Anyone with personal experience with both, for comparison?
I have not tried the Sonarworks product, so my experience is limited to the free version of ReSing, which provides only two voices (one male, one female) plus a couple of instruments (alto sax and bass guitar, if I remember correctly). From the articles you linked, it sounds like they are quite similar overall, at least now that there is a potential of running SoundID locally, rather than having to pay for and use tokens to generate the transformations.
On the pricing front, it would seem SoundID VoiceAI could be a somewhat better value in terms of pricing at each package level versus the number of voices/instruments included.
I don’t know enough about the Soundworks product’s flexibility in fine-tuning the results, but ReSing does provide a fair bit there, with some tuning dials for character (e.g. energetic, mellow, etc.), accent (based on the source vocal you’re replacing, the voice model singer’s, or a blend of the two), and dynamics (just following those of the source vocal or “flat”), as well as some included effects (which I don’t use since I want to do that stuff in Cubase with the plugins I usually use for that sort of thing). ReSing will also let you blend multiple voice models to get a hybrid result, though I haven’t tried that.
I have used both of the voices in the free model, both for a few quick experiments and, in real productions, to double my own background vocals to get an effect that is closer to a group vocal with multiple singers (versus just one person overdubbing all the harmonies). In that latter case, which is my only use case at the moment, I double my own BGVs, then copy those and process each with ReSing. So, for example, if I’m doing two parts of harmony (beside the lead vocal), I’ve got four tracks of my own BGVs plus four additional tracks run through ReSing.
One thing worth mentioning on the local processing (which ReSing also does) is that, when you first use ReSing, it wants to run a process that checks your computer and tunes the processing strategy for your system. I suspect that is similar to the tuning that is possible in SpectraLayers for running the AI components, where, if you have a powerful enough video card with enough VRAM, then some portion of the processing happens there instead of on the CPU, but, if not, it will happen (much more slowly) on the CPU. I know going with the video card makes a pretty huge improvement in my stem separation processing with SpectraLayers, and doing the ReSing transformations is quite quick on my system. Note that any time you make significant tuning changes in the ReSing settings, it does have to render the audio again, so this could be a significant impact if you don’t have a capable enough system.
In my early experiments, I just tried translating a lead vocal on a song I’d recorded to the female voice, the male voice, and the sax. The sax was pretty laughable. Of course, the female voice was in a different range than my voice, but there is a transpose feature, so I just transposed the accompaniment part, too, and it seemed reasonable. My source vocal on that song (nothing I will release) was pretty all over the place in areas like pitchiness, so that did compromise the results. (I think the effects in ReSing include tuning, in which case that might help with fine pitch issues.) I didn’t much care for the male singer’s voice – it’s pretty raspy, and could work for some things, but not the specific song I was trying with it. However, both voices work well enough for my BGV blending purposes, as long as I set the parameters however makes the most sense for the specific song.
FWIW, I am running ReSing via ARA. (I think you can also run it as a VST3 plugin.) One UI thing on that is that the ReSing UI seems pretty small, and I haven’t been able to figure out how to make it get better on that front. The weird thing is that I could have sworn the first time I tried it that that was not the case – i.e. I could expand it to a larger size. However, there have been both ReSing and Cubase version updates in between, so maybe one or both of those were responsible for the change. It’s a somewhat minor annoyance, but might be more of one for users with vision problems (and I wonder if it would be even worse on a 4K screen – mine is just HD).
Beyond the basics, IK also provides some add-on packages (even beyond MAX) to add some additional voices, and they also have a new program where you can essentially rent a voice for a month. Thos have a range of price points, probably partly depending on the success level and/or recognition of the singer (e.g. the one who is most expensive, at over $40 for a month’s use, sang on the road with Pink Floyd and has some other significant credits). ReSing also lets you model your own voice (or another voice you have rights to model, for example via a work-for-hire agreement, or if a client needs you to model their voice for some reason). I think even the free version lets you do that with one voice, and the higher end packages more model creations, with an unlimited modeling package also available for an extra cost.
Bottom line for me is that I’ll probably purchase one of the packages, preferably MAX, when budget allows and I have the sorts of needs that demand different voice styles (e.g. I think if I were trying to simulate a gospel choir, the two voices included in the free version would not be very helpful). I may experiment more with the instruments at some point, but my only application at the moment is for layering with my own background vocals.
I don’t know if there is a free version of the Sonarworks package, but, if so, or if there is a time-limited demo, you might consider trying both ReSing Free and the Sonarworks equivalent, using the same source vocals for each and seeing which gives you better results.
1 Like
SoundID VoiceAI does have a free tier as well. You get 8 models with it (4 voices and 4 instruments), so the side by side comparison you mentioned at the end is totally doable. Also, you can start a free trial anytime to unlock the whole catalog of 50+ models.
2 Likes
You ever seen one of these? It’s pretty good tech.
I heard it’s really efficient at getting vocal recordings.
A timely note for anyone considering a paid version of ReSing: I got an email this morning from IK advertising 60% off all software – it said for 3 days only, but the actual page on the site says it’s only through May 17th. I ended up going for ReSing MAX – $80 instead of the usual $199.99.
Here’s a link to the sale page:
1 Like
Thanks for that heads up, @rickpaul .
I’ve got a dinosaur computer, “and no GPU”. I don’t have time before the 18th to run their free or almost free trial, so I’ll have to decide whether to risk the $80-90 on their discounted Resing Max product.
I have been a fool for lesser things. And if it doesn’t work, I guess I could resell it.
The ReSing Free is permanently free, so you can keep using it indefinitely. Had the 60% discount not been available, I’d likely be continuing to do that, for the limited use I was making of it, until such time as some sale provided a significantly lower price than the $200 list or I had a strong enough need for it (e.g. another recording that needed to make a choir out of my background vocals, thus wanting multiple additional voices in addition to mind – I wish it had been available, and I’d had it a couple years ago for my “Brand New Mirror” recording, where I was trying to do exactly that, specifically trying for a gospel choir sound, and had to resort to various creative uses of processing more copies of my own vocals to try and get as close as I could, which was nowhere near as close as I’d have liked).
I looked at your computer specs in your profile, then at the ReSing system requirements:
Windows® (64-bits)
Minimal: Core i5 Processor or equivalent, 8 GB of RAM, Windows 10 (64 bit) or newer.
9 GB of hard drive space.
Requires an ASIO compatible sound card.
Requires an OpenGL 2 compatible graphics adapter.
Supported Plug-in formats (64-bit): VST 3, AAX. Requires an OpenGL 2 compatible graphics adapter.
It doesn’t mention ARA in this, and I’ve only tried it in Cubase 15, but I would expect it should work in Cubase 14, too. If not, it could be used standalone or as a VST3 plugin.
It would at least seem your computers might meet their minimum specs (assuming the OpenGL 2 part anyway).
That said, when I first tried using it to render audio prior to having run the optimize procedure, it was EXTREMELY slow (and I gave up and ran the optimization), and my computer is significantly newer and faster than yours would be. So, it might be a bit painful to use if you need to do a lot of experimentation with settings (which I tend to do in my uses to try and see what settings work best in the specific context of the track).
If I understood correctly (which I may not have), the SoundID VoiceAI does at least have the option of doing its rendering on their server, which might be less painful than trying to do it on a limited computer, albeit at the expense of pay-per-render (i.e. purchasing and using tokens), and I have no clue how quickly that might add up to the $80 for ReSing Max (or likely $52 for the standard ReSing).
1 Like