|🛠||This subject is work in progress. What is being worked on: Upcoming software
Please bear with us while improvements are being made and assume good faith until the edits are complete.
For information on how to help, see the guidelines. More subjects categorized here.
Maghni AI is a upcoming singing synthesizer created by Misbah Studios in collaboration with VocaTone. It will also feature deep learning technology. It was announced on December 25, 2020 and will be a crowdfunded project.
On December 11, 2020, VocaTone's twitter account started tweeting after years of inactivity. From that point throughout the month of December, VocaTone's twitter account kept teasing a big announcement coming soon, aswell as constantly retweeting fanart of their previously established characters, and other English VOCALOIDs. Maghni AI was finally announced on December 25, 2020. In a Facebook post, VocaTone revealed that the software had already been in development for three years at that point. At the time of the announcement, there was a prototype that was currently in pre-alpha. They also hope to have the editor available for free with voicebanks available at a low cost. It was set to be a Kickstarter that would go live on January 7, 2021, but was postponed due to community concerns about the quality of the synth due to not having any sort of software demonstration at the time.
On December 27, 2020 there was another Q&A, this time including VocaTone on the VocaVerse Discord. In it it was stated that the Kickstarter would of had a goal of $35,000. It was also mentioned that Aurum and Audine would not be the only vocals for the software, and are interested in distributing a paid software development kit. They also mentioned that they will have third-party support but are not at liberty to say who. The software's default phoneme system would be X-SAMPA, but would also have support for ARPABET and IPA among others that weren't named. The software will be both a VST and standalone with both Windows and Mac support. There are currently no plans for Linux, but would consider it if there is a demand. There is also interest in making the software available for educational purposes. The software's name comes from the Arabic word for singer "maghni" (مغني), looking to avoid the letter "V" which is commonly used in the names of other vocal synthesizers. When asked about speech synthesis support, they mentioned that singing was the primary focus, but would not rule it out. They were also considering adding accessibility features for hard of hearing and dyslexic people. AI implementation would be used for pitch and other parameters and was mentioned to be similar to Synthesizer V's AI. It was also mention that pitch drawing would be similar to DeepVocal. The rendering system was expected to be multi-threaded. When asked about the two other vocals mentioned around Aurum's VOCALOID development, it was revealed that there were actually five vocals in development. The second being Audine, the third being a girl called Era, and another female named Nyx. Audine and the others were never announced due to VocaTone losing the VOCALOID license.
On January 25, 2021, VocaTone announced they dropped YOHIO from their list of vocal providers after some dissatisfactory political comments from the singer were made. Until this point, YOHIOloid was considered for an update on Maghni AI, as mentioned in the Christmas Q&A with the VocaVerse Discord. It was noted that this meant they no longer had to make a Swedish phonetic list for the software, hinting this was a language they would have covered with the update.
On January 31, 2021, a progress report was released detailing what was accomplished throughout that month on the Maghni AI's development. On the 22nd, work on the synthesis engine had begun and the next day rules for AI implementation had been established. By the end of the month French, Japanese, Malay, Italian, Mandarin Chinese, Swahili, and Korean had been fully implemented aswell as German, Spanish, and Cantonese were reported to be halfway done. By the 31st English and Japanese maps were also created and a test Japanese voicebank has been recorded.
On September 9th, 2021, an audio demo of Maghni AI was posted on Misbah Studios' Twitter. This demonstration showed comparisons of Audine and Aurum's voice providers compared to how they sounded in Maghni AI.