New updates to the engine and voice libraries make it effectively impossible to produce the same output as the prior versions. This means that users who do not have backups of their software installers may find that a finished project needs unexpected adjustment. In an even worse scenario, a user trying to provide additional renders to a client may be unable to produce files that match those previously provided.
Edit: I’d also like to note that currently while update installers for the editor are saved on the file system, updates for voice libraries themselves are not persisted and so downgrading can be completely impossible, even if you do back up your installers.
Allow users to easily select which version of the editor/engine and voice library to use, including the ability to downgrade to any past version. Ideally we would be able to install multiple versions of the same voice library simultaneously, though I acknowledge that poses additional challenges.
That may be a bit impossible. Mainly, when a voicebank gets updated for clarity or improvments for the software and ai patches. The voicebanks would no longer work with older builds of SynthV . The best solution is to keep the builds synth v downloads an back them up so you could do a downgrade. But, by doing so there will be compatibility issues. An some voices will not sound right . Resulting in more crashes than, you think.
So, unless you back up the voices, and engine. It’s unlikly a downgrade feature will happen. Since, older builds get deleted from the Synth V servers. Mainly to help cut back the cost of file storage . Since, Synth V is used by a nich community -.- Which yeah it sucks to hear the creator of Synth V even say that .
But, unless you have the pro version to mod the engine with lua scripts. It may not be worth downgrading synth v in the long term unless, a major bug happens .
Uh, 70mb wouldnt even cut the memory space needed . For the engine your looking at 41mb , with the base voicebanks . Your looking at 534 mb for the free voices , So assuming everytime synth gets updated with all the voices . That file size increases . So, you add 595 mb for every time a new build comes out . The file size your assuming to back up every synth v engine , and free voices would start costing the devs money .
Realisticly , it’s not within the Synth V teams interest to back up or allow downgrades . If synth v basic started to charge people you would see a decline in users and thats not always the best thing for a niche program as it it .
The ai changes may be not what you want. But, when its liked by most people. There likes vs dislikes of the ai changes doesn’t make it worth backing up. When, the latest updates add more good than, it sounding bad.
The best you can do at this point is google old builds . But, that would take you to the og synth v engine which the voicebanks has been removed. So, the best you can do as of now is the conract the synth v team .
As you can see, past installers only take up 70-75MB each. Each engine update also only results in 1-3 voice library updates, and only for AI voices (15-25MB each). You could fit every update installer released to date in the free Google Drive storage provided with a new Google account.
Maybe, its because of some other reason prior download can’t be done. Say, if someone just bought Synth V pro and the email you get (if you get one ) sends you you zipped exe and a link just to get the latest pro / past builds. It would be easy for someone to leak the pro builds and people would pirate synth v pro like what people did with vocaloid .
I would have to assume that would be a partical reason why past builds or a down grade option is unlikly . Never hurt to use the synth v contact forum just incase : Contact | Dreamtonics株式会社
I am not talking about the original SynthV editor from years ago. This thread is exclusively about SynthV Studio and the changes to AI voices that occur when updates cause them to behave significantly differently.
Please stop trying to provide support and advice. I am not asking for help with anything. I am merely suggesting that the way product versions are currently handled is flawed, and proposing a potential solution.
I’d like to thank Claire for bringing up this issue. I agree it’s an important point.
I think that Synth V is an excellent, worthwhile project, and it would be unfortunate if it limited it’s own scope by not taking issues of professional usability seriously – and I do think it tries to! I do think many of us here have vision and inspiration for the project (including the team who makes it), and do see it with a bright future with potential for being much more than a “niche” toy software.
I second Claire’s proposal to support maintaining the integrity of older projects, so that Synthesizer V may be considered a viable production software for artists and other professionals. It would be throwing away a lot of brilliant work and potentential if we didn’t consider these kinds of issues seriously.
Additional evidence that the current version control method should be revised:
Users that cannot upgrade the editor due to instability issues lose access to intermediate version numbers of voices, even if those voices are compatible with the version of the editor being used.
In this case the user has version 1.5.2. Their voice installer is for Saki AI 110, however Saki AI 118 is the compatible version for SynthV Studio 1.5.x and enables cross-lingual synthesis. They do not have an installer for version 118, and only have the option to update to version 119, which is not usable for the user due to crashing issues in SynthV Studio 1.6.x.
Either the updater should check for “latest compatible version” for voices, or the user should have some way of installing intermediate versions manually.
With the recent issues introduced by 1.7.0, Dreamtonics has made previous version installers available. This does not include all past versions, just recent ones. For example, the available versions for Saki AI are 119 and 123.
Good news! In 1.7.1 users will be able to install multiple versions of the same voice, and freely swap between installed versions. This will allow users to continue using previous voice versions, and will prevent situations in which a project is disrupted due to unexpected changes to the AI model.
Using the pre-release version 1.7.1b1, you can use these features before 1.7.1 is fully complete. Please remember this is a beta release and may introduce new bugs, but should also resolve some of the major bugs with 1.7.0.
Big thanks to Dreamtonics for implementing this feature, it really is appreciated!