# Vibe-coded Dorico development? **URL:** \ **Category:** Dorico\ **Created:** [October 14, 2026, 2:00am UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550 "2026-10-14T02:00:13Z")\ **Posts on this page:** 20\ **Page:** 2
**Author:** ![phase\_Shift](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/phase_shift/32/207296_2.png) [@phase\_Shift](https://forums.steinberg.net/u/phase_Shift)\ **Post date:** [October 15, 2026, 2:14pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/21 "2026-10-15T14:14:10Z")
> [@judddanby](#): > > I struggle to understand people’s positions at times. One important reason is that “AI” (as a sobriquet for a number of technologies, some of which are already in use by more familiar names) will at much-hyped hyperscale become Jevon’s Paradox raised to an order of magnitude. There is not an infinite amount of energy available to build all these deafening data centers, not even the ones already planned. For a reality check, look at your electric and gas bills. Homes in some places with data centers have been reduced to a tickle of brown tapwater. And we aren’t close to what is sketched up on silicon valley napkins. The second fundamental problem I see is that language is not knowledge, and an assemblage of probabilistic words will yield an immeasurably small amount of new thought, nor even an accurate rendering of what has been ingested from such sources as Britannica and Reddit. You might not even get a reader’s digest level of feedback. (Note: coding is different - a limited language set, clearly delineated syntax, and libraries of well proven solutions - tailor made for this use case.) If Cantai could read and syllibate words the way (in my case) a straight-tone choir does, and include young treble voices, and If the Dorico playback engine could render audio and tempo in a realistic, rubatic way, then…maybe? Right now a daw and several existing vst’s give me an audio demo that is just short of the real thing. The Dorico _score_ is my product - my audio comps aren’t for consumption, they are only proof for music directors. AI may not be an insurmountable problem for me; I struggle with the limitations of playback details in a notation program. The scores and parts I render from Dorico are realized in live performance by real musicians, singers, and instruments. I want the Dorico _scores_ to be perfect, not its audio. ---
**Author:** ![benwiggy](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/benwiggy/32/1715_2.png) [@benwiggy](https://forums.steinberg.net/u/benwiggy)\ **Post date:** [October 15, 2026, 2:21pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/22 "2026-10-15T14:21:57Z")
> [@Shoobidoo](#): > > if musicians (not to mention Dorico developers) are starting to be replaced by AI, we can choose not to support such a development. I’m not convinced that any musicians are being “replaced” by Cantai; or losing revenue because of it – in the same way that the BBC Symphony Orchestra isn’t losing money from me because I’m using Spitfire’s sample library. I might be able to instruct AI to make me a picture of a moody bint in front of the Arno valley, but that’s not going to depreciate da Vinci’s talent. ---
**Author:** ![judddanby](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/judddanby/32/183340_2.png) [@judddanby](https://forums.steinberg.net/u/judddanby)\ **Post date:** [October 15, 2026, 3:08pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/23 "2026-10-15T15:08:35Z")
> [@dan\_kreider](#): > > He said a few years ago this would have taken a team of people probably nine months to do. The flip side to your friend’s anecdote might come down the road. Since code can be difficult to parse after the fact, I’m reading more and more that companies who went the assisted-coding path are now backing away because, while it saves initial development time, it net-costs time when later anyone tries to go in to grapple with it/expand it/understand it. Has your ~~fired~~ (corrected:) friend experienced anything like this, do you know? ---
**Author:** ![phase\_Shift](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/phase_shift/32/207296_2.png) [@phase\_Shift](https://forums.steinberg.net/u/phase_Shift)\ **Post date:** [October 15, 2026, 3:10pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/24 "2026-10-15T15:10:36Z")
> [@judddanby](#): > > Has your fired [sic] experienced anything like this I have a coder friend - a release engineer! this is his job - whose workload has tripled, and resulted in the hiring of human assistants. ---
**Author:** ![asherber](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/asherber/32/277014_2.png) [@asherber](https://forums.steinberg.net/u/asherber)\ **Post date:** [October 15, 2026, 3:13pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/25 "2026-10-15T15:13:21Z")
> [@judddanby](#): > > while it saves initial development time, it net-costs time when later anyone tries to go in to grapple with it/expand it/understand it. Not even later – it takes more time to code review it in the moment. I agree that this is a drawback, and I’ve recently read articles describing studies that concur. At the same time, I have some friends who say that this is temporary, that the ultimate goal is code which is both written and reviewed by AI, no humans necessary. If you think about it, we currently trust our compilers to accurately turn our code into an executable without review; it’s plausible that we might eventually similarly trust AI to accurately turn our instructions into code. But I’m not sold yet. ---
**Author:** ![Shoobidoo](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/shoobidoo/32/344813_2.png) [@Shoobidoo](https://forums.steinberg.net/u/Shoobidoo)\ **Post date:** [October 15, 2026, 3:18pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/26 "2026-10-15T15:18:23Z")
> [@benwiggy](#): > > I’m not convinced that any musicians are being “replaced” by Cantai; or losing revenue because of it – in the same way that the BBC Symphony Orchestra isn’t losing money from me because I’m using Spitfire’s sample library. > > I might be able to instruct AI to make me a picture of a moody bint in front of the Arno valley, but that’s not going to depreciate da Vinci’s talent. Revenues for film-scoring musicians have dropped substantially in recent decades, especially in LA. Whether this is due to sample libraries, I don’t know. However, given the enormous growth in the number of films, TV shows and ads produced in recent decades, it seems plausible that it has contributed. However, sample libraries don’t simply replace musicians, as in order to create them, musicians are needed. But as for Cantai and similar, in 2024, [a study commissioned by CISAC](https://www.cisac.org/Newsroom/news-releases/global-economic-study-shows-human-creators-future-risk-generative-ai) projected that generative AI could put 24% of musicians and producers revenue at risk by 2028. That’s just around the corner. ---
**Author:** ![asherber](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/asherber/32/277014_2.png) [@asherber](https://forums.steinberg.net/u/asherber)\ **Post date:** [October 15, 2026, 3:49pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/27 "2026-10-15T15:49:49Z")
> [@Shoobidoo](#): > > But as for Cantai and similar, Cantai uses sampled voices. They pay their singers, and the say that the singers are also compensated on an ongoing basis depending on how many times each voice gets used. AI is used to make the sampled voices sing words. ---
**Author:** ![phase\_Shift](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/phase_shift/32/207296_2.png) [@phase\_Shift](https://forums.steinberg.net/u/phase_Shift)\ **Post date:** [October 15, 2026, 5:05pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/28 "2026-10-15T17:05:24Z")
> [@asherber](#): > > the sampled voices sing words They sing phonemes - vowels and consonants. ---
**Author:** ![asherber](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/asherber/32/277014_2.png) [@asherber](https://forums.steinberg.net/u/asherber)\ **Post date:** [October 15, 2026, 5:10pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/29 "2026-10-15T17:10:12Z")
I’m not sure what you’re saying. They may have been recorded singing phonemes – I don’t know what their process is – but the AI is only responsible for assembling those phonemes into words. That is, the voices are not AI-generated. ---
**Author:** ![Shoobidoo](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/shoobidoo/32/344813_2.png) [@Shoobidoo](https://forums.steinberg.net/u/Shoobidoo)\ **Post date:** [October 15, 2026, 6:52pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/30 "2026-10-15T18:52:27Z")
> [@asherber](#): > > Cantai uses sampled voices. They pay their singers, and the say that the singers are also compensated on an ongoing basis depending on how many times each voice gets used. Yes, and that is commendable. But they are a tiny minority, and my point isn’t to scrutinize who makes an effort to act responsibly. It’s about the principles behind the generative AI industry, and most importantly the consequences for the people whose work and livelihoods are being undermined. The vast majority of generative AI have been developed using other people’s work, their music, singing, visual art or writing, often without their consent or even their knowledge. Paying the singers addresses one ethical concern, but doesn’t resolve my much broader objections to the principles behind the generative AI industry, and their consequences. ---
**Author:** ![Shoobidoo](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/shoobidoo/32/344813_2.png) [@Shoobidoo](https://forums.steinberg.net/u/Shoobidoo)\ **Post date:** [October 15, 2026, 7:01pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/31 "2026-10-15T19:01:40Z")
> [@asherber](#): > > They may have been recorded singing phonemes – I don’t know what their process is – but the AI is only responsible for assembling those phonemes into words. That is, the voices are not AI-generated. I’m afraid they are. Cantai generates synthetic singing based on real singer’s recorded voices, so the resulting vocal performances are AI-generated. That’s also why it sounds “uncanny valley” when listening to them. It’s very, very close to the real thing, but something is ever so slightly off. ---
**Author:** ![Phillipus](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/phillipus/32/146731_2.png) [@Phillipus](https://forums.steinberg.net/u/Phillipus)\ **Post date:** [October 15, 2026, 7:11pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/32 "2026-10-15T19:11:21Z")
This post was flagged by the community and is temporarily hidden. ---
**Author:** ![asherber](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/asherber/32/277014_2.png) [@asherber](https://forums.steinberg.net/u/asherber)\ **Post date:** [October 15, 2026, 7:23pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/33 "2026-10-15T19:23:39Z")
> [@Shoobidoo](#): > > I’m afraid they are. Cantai generates synthetic singing based on real singer’s recorded voices, so the resulting vocal performances are AI-generated. Yes, but isn’t this the same as sampled orchestral instruments? With an instrumental work in Dorico, individual notes are sampled and then assembled into longer lines. The only difference here is that Cantai needs to decide how a word is pronounced, and then manipulate the phonemes to accomplish that. But the underlying sound itself is not generated. (This is my understanding of how Cantai works, based on various posts I’ve read. I couldn’t find a technical explanation online, so I could be wrong.) ---
**Author:** ![Shoobidoo](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/shoobidoo/32/344813_2.png) [@Shoobidoo](https://forums.steinberg.net/u/Shoobidoo)\ **Post date:** [October 15, 2026, 8:06pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/34 "2026-10-15T20:06:00Z")
> [@asherber](#): > > Yes, but isn’t this the same as sampled orchestral instruments? With an instrumental work in Dorico, individual notes are sampled and then assembled into longer lines. The only difference here is that Cantai needs to decide how a word is pronounced, and then manipulate the phonemes to accomplish that. But the underlying sound itself is not generated. > > (This is my understanding of how Cantai works, based on various posts I’ve read. I couldn’t find a technical explanation online, so I could be wrong.) It’s not the same. Usually choir libraries are fundamentally designed as you describe: Dorico or a regular sequencer determines which samples are needed to assembly a phrase, and the playback engine, such as Kontakt, play back a sequence and/or layers of those samples. Cantai works differently: its output is synthesized by a machine learning model trained on recorded and analysed vocal data. It’s not simply playin back selected and assembled recordings of sung notes or syllables; it synthesises the vocal sound itself. You can read about it in [Cantai’s manual](https://www.musehub.com/sv-se/learn/cantai-sings-your-score): “Most choir libraries are recordings. Singers were recorded holding vowels or a set list of syllables, and the library plays those recordings back when you press play. Cantai uses machine learning to synthesize the voice instead, so it can pronounce the text you type rather than pick the closest recording. That is also why the whole library is an 11 MB download, while the free Muse Choir takes about 0.64 GB. Cantai includes four solo voices, soprano, alto, tenor and bass, as well as choir ensembles. You can write an aria for one voice or a four-part anthem in the same library. Because the voice is synthesized, Cantai has to render your score before it can play it.” ---
**Author:** ![asherber](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/asherber/32/277014_2.png) [@asherber](https://forums.steinberg.net/u/asherber)\ **Post date:** [October 15, 2026, 8:49pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/35 "2026-10-15T20:49:32Z")
Thanks for the info—it seems my understanding was incomplete. ---
**Author:** ![SugarFree](https://avatars.discourse-cdn.com/v4/letter/s/57b2e6/32.png) [@SugarFree](https://forums.steinberg.net/u/SugarFree)\ **Post date:** [October 15, 2026, 9:14pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/36 "2026-10-15T21:14:26Z")
You really hijacked this thread. The topic was vibe-coding, which is different than the type of generative AI you are referencing. ---
**Author:** ![arco](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/arco/32/222268_2.png) [@arco](https://forums.steinberg.net/u/arco)\ **Post date:** [October 15, 2026, 9:31pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/37 "2026-10-15T21:31:04Z")
Perhaps a moderator will split it off. ---
**Author:** ![Pete\_the\_Musician](https://avatars.discourse-cdn.com/v4/letter/p/439d5e/32.png) [@Pete\_the\_Musician](https://forums.steinberg.net/u/Pete_the_Musician)\ **Post date:** [October 15, 2026, 9:40pm UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/38 "2026-10-15T21:40:29Z")
Vibe coding sucks. AI will fix a bug by adding 50 lines where simply moving a single line to elsewhere in the function will do. This happened to me a few days ago. Adding 50 lines to fix a bug that requires moving a single line is a coding error. Even if it “works” as a short-term fix it’s still an error. Using AI to code something on the scale of Dorico is like building a skyscraper on extremely weak foundations. It might stay up for a short while, but it will come crashing down sooner or later. ---
**Author:** ![dan\_kreider](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/dan_kreider/32/254871_2.png) [@dan\_kreider](https://forums.steinberg.net/u/dan_kreider)\ **Post date:** [October 16, 2026, 1:20am UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/39 "2026-10-16T01:20:52Z")
> [@Pete\_the\_Musician](#): > > Vibe coding sucks. Sometimes. Except when it doesn’t. ---
**Author:** ![Shoobidoo](https://dub1.discourse-cdn.com/steinberg/user_avatar/forums.steinberg.net/shoobidoo/32/344813_2.png) [@Shoobidoo](https://forums.steinberg.net/u/Shoobidoo)\ **Post date:** [October 16, 2026, 1:55am UTC](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550/40 "2026-10-16T01:55:01Z")
> [@SugarFree](#): > > You really hijacked this thread. The topic was vibe-coding, which is different than the type of generative AI you are referencing. I did not. You brought Cantai into the discussion by telling another poster that he was denying reality if he wasn’t on board with AI cloning while still using Cantai. I responded to your post, and then to the replies I got. Conversations naturally evolve, and if you tell people they’re denying reality, it’s to be expected that they’ll respond and explain their position. But let’s just bring it back to the topic of vibe-coding, mate. No need to point fingers. [Previous page](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550.md?page=1) [Next page](https://forums.steinberg.net/t/vibe-coded-dorico-development/1043550.md?page=3)