Audio Rendering

Wikipedia documents the first song sung using computer speech synthesis: Link. The voicing module itself is an area of study. Here is a paper that discusses MIDI-to-singing.

Potential software packages

We need software that:

  • Sings lyrics with an associated melody
  • Renders harmony (in any format, including notes)
  • Takes a file as input
  • Ideally can be run from the commandline

s://www.reddit.com/r/Vocaloid/comments/2yjhhs/know_of_any_good_alternatives_to_vocaloid/

CeVIO

://cevio.jp/downloads/

For windows only.

ChipSpeech

s://www.plogue.com/products/chipspeech/

This is from the same company as AlterEgo, but it comes recommended on a Reddit page. Might give it a second look. Looks like it has intended musical usage. I actually don't think that this will work. It's like AlterEgo in that it only does one voice.

The demo gives you four minutes and then shuts off, so you can't really figure out how to use it anyway.

VirtualSinger

VirtualSinger from Myriad.

  • Sings melody with lyrics and plays accompanying harmony
  • Renders harmony
  • Loads from file
  • Can't be run from the commandline?
  • Has options for better “real” voices

Got it to load Let_it_be.mid and had to manually associate lyrics with a staff, then just hit play. Easy to use. This could work, but it would need to be manually rendered each time.

From correspondence with Myriad, I confirmed that there is no commandline option, but you can write a script to be run within HarmonyAssistant, which may work out alright.

October 6, 2016

K, we're going to give a legit go at this. Harmony Assistant is more powerful than MelodyAssistant so we're using that. Importing a .kar file and playing it takes all of three clicks (“open” and then “associate this track with the melody” and then “play”). I need to figure out how to adjust volume. The voice is discernable, but very tinny. Probably give it like a 6 on a scale of 1 to 10. I think you can sub in other instruments, though, as wanted.

They have a GOLD sound base to upgrade the MIDI voices. They have different audio voices and even the ability to record your own voice. HarmonyAssistant costs 85 bucks and VirtualSinger module costs 25 bucks. Probably worth it.

Yamaha CyberDiva

://www.synthtopia.com/content/2015/02/08/yamaha-intros-english-speaking-cyber-diva/#more-61924

://www.vocaloid.com/us/

Sounds pretty good. No trial version to try it out. It's mostly used for techno. Costs $120 ish. Tricky because the version that runs on mac requires cubase. Then you have to buy voices. So probably cost a lot more.

From correspondence with Yamaha, they referred me to Hatsune Miku's concert as an example of what she's capable of. Doesn't really answer any of my questions. No way to test it out. Probably can test out some voices, though.

Sinsy

Not bad. Funny because the english is clearly a Japanese-English voice. But it does a pretty good job and is discernable. Only has web interface and only produces a single monotone voice. You give it a musical xml file. This might not actually be a bad approach. We could probably figure out how to automate it and the musical xml file has a fairly straight forward. We'd have to figure out how to combine it with the harmony. Might be the most promising so far in terms of automatability. ://www.sinsy.jp/

AlterEgo

Site

  • Sings lyrics with an associated melody, but lyrics must be typed manually
  • Doesn't render harmony (in any format, including notes)
  • Takes a file as input for melody, but not for lyrics
  • Can't be run from the commandline

eCantorix

s://github.com/divVerent/ecantorix

Looks promising as a simple commandline tool. Licensing is convenient. Can use it for whatever I want. Tried it on a midaoke file and it croaked, but I think it's because there were other voices in the midi file. 'Course if this is any suggestion of what we can expect, it's not promising.

UTAU

://utau-synth.com/download.html

Notice, Japanese for “sing”. Desktop application totally works, but is in Japanese. So I don't understand anything of what's going on there. Looks like you can import a midi file.

Lyricos

Site

Got Festival to work. Now trying to figure out lyricos.

Couldn't get it to install correctly. With more work, it might work. But all in all seems old, lots of broken links, and I think it's main focus is speech, not singing or music.

Flinger

Site

This, like Lyricos, is an extension to Festival. Download available here

Cutting my losses. It looks like I might eventually get it to work if I really worked at figuring it out (right now it doesn't seem to recognize Flinger), but even if I do, getting anyone else to be able to use it seems like a stretch. I need something easy to use. Not to mention Flinger and Lyricos and Festival are all sorely outdated.

I tried again to get Flinger to work. No dice. Just doesn't recognize it. Documentation is so outdated and piecing things together everywhere. Got Festival to speak, but that's it. Would have been nice because it's all commandline.

Singing Computer

Site

Github code

Hasn't been touched sing Feb 2008. Also looks to be designed for use with Festival. Could probably get it to work, but looks messy. Not sure I like Festival.

Candencii

Japanese product downloaded here.

Hasn't been touched since 2011. Disk image for download is damaged. Can't install it.

Tried an older version. Still corrupted. Didn't work.

Check this out

StackOverflow

://stackoverflow.com/questions/2695865/computer-generated-vocals

Most of these are terrible suggestions, like “use text to speech and then auto-tune it”. It references vocaloid.

Cantor 2

://www.crypton.co.jp/mp/do/prod?id=25701 Discontinued. Supposedly inferior quality to vocaloid.

mind/audio-rendering.txt · Last modified: 2016/10/06 20:55 by norkish
Back to top
CC Attribution-Share Alike 4.0 International
chimeric.de = chi`s home Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0