VocalsMixingVocal Chain Order
EQ, compression, de-esser, reverb. The order is not taste: each stage changes what the next one hears.
Most harmonizer plugins move the pitch by a fixed interval, which is why the third goes wrong on half the chords. Scale locked harmony does not.
The most common complaint about harmony plugins is short and accurate: they just change the pitch by a fixed amount. Ask for a third above and you get four semitones above, on every note, for the whole song. In a major scale a third is four semitones on some degrees and three on others, so a fixed shift is right where it happens to fit and wrong everywhere else, and the wrong ones do not read as harmony. They read as a mistake on the same word every chorus. Everything below follows from that one interval being variable.
Staying in key means the plugin decides a note rather than an offset. The decision needs two things: a set of notes the song is using, and a rule that every generated voice is snapped into that set before you hear it. Producers usually call that rule a scale gate. A gate applied to the lead alone is only half the job, because the harmony voices are separate decisions and a stack of three can drift out of the scale while the lead sits perfectly inside it.
Every method above is only as good as the key you feed it, and a wrong root makes scale locking confidently wrong instead of randomly wrong. If the song came from a loop or a sample you did not make, work the key out first and set it explicitly rather than trusting the first guess.
A chord track on a sidechain beats a scale, and a scale beats nothing. With a chord locked, a generated line lands on a chord tone 95.5 percent of the time, against 36.3 percent for a fixed note. If the plugin cannot name a key from what it has heard, the correct behaviour is to say so and quantise nothing, because a wrong lock is worse than no lock.
A third below thickens without competing for the melody. Two voices already read as an arrangement choice, three read as a section. In PRISM the three voices are a third down, a third up and an octave down, each on its own switch, and only the third down is on by default for that reason.
Harmony that sits level with the lead turns the phrase into a chord and takes the words with it. Push the lead in front first, then place the stack, then send both to the same space so they read as one performance. Where that sits in the chain matters as much as the interval does.
Scale locking is easy to claim and easy to measure, so here is the measurement. With key and scale set, out of scale decisions in the engine go from 12.1 percent to zero. Across all three harmony voices, out of scale notes measure zero as well, because the voices pass the same gate as the lead rather than a looser one of their own. The lead stays 8 to 17 dB in front of the stack, which is the difference between harmony that supports a line and harmony that argues with it.
That is the half of PRISM people notice second. The first half is the talkbox itself, and both halves share one decision layer, so the note the lead commits to and the notes the harmony commits to come from the same source, in the same order, in the same key. If you want to hear it against your own vocals, the trial runs seven days and takes no card. It opens the plugins, and the sample packs stay closed while it runs.

PRISM Talkbox runs up to three harmony voices through the same scale gate as the lead, so the stack lands in your key. Seven day trial, no card, plugins only.
Guides
You may also like
VocalsMixingEQ, compression, de-esser, reverb. The order is not taste: each stage changes what the next one hears.
Bass designMixingPhones and laptops have no sub driver. If your 808 only exists below 60 Hz, most of your audience hears nothing at all.