VA Mixing
Overview
Voice Acting Mixers, Mastering Engineers, Dialogue Editors... No matter what we're called, our job is to make Skywind's voices sound great.
Voice actor recordings arrive to us as mono, 44.1 kHz, 16-bit files. The audio is typically raw and unaltered (though there are exceptions). Our "filecutter" volunteers have already reviewed these recordings and picked out the best takes for each line. Our job is to make them sound as good as possible.
Our Workflow
ClickUp is the website where we track who is working on what. On ClickUp, you'll see a list of the tasks that are ready for mixing. You will claim one of these "cards" and move it along towards completion. Here is a general guide of how to get started with ClickUp.
1. Ready for Mixing
There are currently three columns from which you can grab mixing/mastering tasks:
Ready/Batch: Batch processing these cards is recommended, as they are for minor characters/townsfolk where it's not as important to clean up the audio file-by-file.
Ready/Creature: These cards are for unusual characters which will require some special effects treatment. Not recommended for newcomers to the team.
Ready/Loudness: These cards were mastered in the early days of Skywind, before we standardized our mixing process. They could use a quick check for making sure the loudness levels match our current targets.
Ready for Mixing: Major characters where it's important for the audio to be fully mastered. Cards at the top are most important, cards at the bottom are less important.
2. Peer Review
Once you've chosen a card that you want to work on, you should assign it to yourself and drag it into the "Peer Review" column.
These cards often have hundreds of voice lines. It's a lot of work. Before you start working through all those files, we should talk it over!
Choose a few audio clips from your card that represent the audio well (some loud lines, some quiet lines, some retakes, etc.). Mix these clips how you intend to mix the entire card, and post your audio on Discord for the other mixers to hear. Together we can talk about your EQ, compression, etc. and iron out any issues before you get going. Anyone with the "VA Mix Approver" role can approve you to move on to the next step.
3. Mix In Progress
The other mixers have heard your sample audio and approved it -- you are good to go! Drag your card into this column and start working on the entire card.
Your final target is:
WAV
Same filenames as what you received
Similar audio length as what you received*
16 bit
44100 Hz
Mono
True peaks should not exceed -1 DBFS
Loudness of approximately -20 LUFS. Shoot for a range between -22 LUFS and -15 LUFS.
About Audio Length
The audio files that you received should already be trimmed to have an appropriate amount of silence before and after the line. You should generally leave the amount of silence alone and you won't have any issues. If you're curious to learn more, the amount of silence needed is driven by two factors:
The game engine. If an audio file has too little silence at the beginning, the lip sync animations will get confused and look strange for the duration of the line.
Pacing. If an audio file's silence is too long, the dialogue feels awkwardly slow and robotic.
For these reasons, the silence surrounding a typical line should be about 0.25" long, but can be as short as 0.1" for combat lines where perfect lip sync isn't as important.
Regarding AI
Do NOT train AI models on our voice actors' performances. We have an explicit agreement with all Skywind VAs that we will not clone their voices. If a voice line is missing or has been flubbed, we can easily find a human to re-record the line for us. They will love the opportunity! Skywind is great because it is covered in the thumbprints of hundreds of people all over the world. Let's continue that tradition.
"What about my new denoise plug-in? Technically, it uses machine learning algorithms to analyze...-" If it isn't creating a voice, it's probably fine. When in doubt, ask a lead on Discord. The distinction we care about is not how a plug-in works under the hood, but if it breaks the trust that our VAs have given us.
4. Final Review and Delivery
Nice job cleaning all that audio! Place all of your cleaned audio files into one flat folder (no subfolders). If there are any filename conflicts, prefer the more recently recorded file (E.G. audio from "Retakes 3" should replace audio from "Retakes 2"). Zip your folder, and name your zip like so...
_Mixer[MyName]_[Gender][VA][Race]
for example ...
_MixerGanz_FemaleJoyHDunmer.7z
_MixerBenBurtt_MaleJeffBakerAshlander.zip
Upload your zip to the appropriate folder on the OneDrive.
Lastly, ping an audio lead saying you're done, and drag your Trello card from "Mixing In Progress" to "Mixing Final Review".
A lead will check that your files are in the correct format and they'll listen to a few of your lines. The lead may give you some feedback for next time, but it's unlikely you'll have to redo any work. Congrats on finishing a card!