A Serum 2 preset from a vocal chop

A vocal chop is a sample. It plays at the pitch it was sung, for as long as it was sung, and every change you make to it (a new note, a longer tail, a chord) is a stretch or a transposition of that one recording. A preset is different: it is an instrument that plays any note, any length, in chords, with an envelope you can reshape.

Resynthesis turns the first into the second. Give Audial one chop and it builds a Serum 2 patch whose oscillators, filters, envelopes and effects are set to reproduce it. This post covers how to pick the chop, what to expect back, and how to make the result playable.

One thing to be clear about first. Resynthesis rebuilds a tone, not a voice. What comes back is a synth patch with the colour of the vowel you sent, not a sampler that sings your syllable on every key. If you want the word itself, keep the chop in a sampler. If you want a pad, lead or stab that sounds as if a voice is somewhere inside it, read on.

What you need

• A vocal you have the right to use: your own recording, an a cappella that is licensed to you, or the vocal of your own track.
• Xfer Serum 2, version 2.1 or newer. Audial does not sell or bundle it; the preset opens in your copy.
• An Audial account. Every account gets 2 free runs, and a subscription makes runs unlimited.

1. Get the vocal on its own

If the vocal is already a dry stem, skip this. If it is inside a mix, split it out first with the stem splitter and keep the vocals stem. A chop cut from a full mix carries the kick, the bass and the pads with it, and the resynthesis will dutifully try to model all of them.

2. Cut one chop

This is the step that decides the result.

• One syllable, one note. An open vowel held for a moment (an "ahh", an "ooh", an "ay") is the ideal. A phrase of several notes is not one sound, and the Resynth tab warns you when a clip sounds like several notes in a row.
• Short. 1 to 4 seconds is plenty. The limit is 20 seconds, but the useful information is in the attack and the body of the note, not a long tail.
• Trim the front. Cut the silence before the note so the attack sits at the start of the file.
• Dry. Turn off the reverb and delay sends before you bounce, or cut from the dry stem. Anything left in the clip gets modelled into the patch, and you cannot switch it off afterwards.
• Steady pitch. The engine fits one patch to one sound. A chop that slides between notes asks it to model movement as well as tone. If you want the slide, add it back in Serum 2 afterwards with a pitch envelope or glide.

Mono is fine.

3. Run it

Open the Resynth tab on audialmusic.ai, choose Serum 2 above the uploader, and drop the chop in. A run takes about two and a half minutes, plus up to another two and a half if a machine has to start first.

You get back:

• a .SerumPreset named after your file, so ahh_c4.wav becomes ahh_c4.SerumPreset, which is also the name Serum 2 shows;
• a render of the patch;
• two keyboard renders, at C3 and C5;
• a JSON report with a similarity score between the render and your chop, which the page also shows next to the download.

4. Listen before you open it

Play the keyboard renders first. The chop was one note; the renders show how the patch behaves an octave or two away from it. Some vowel colours hold up across the keyboard, others thin out at the top or go muddy at the bottom, and it is quicker to hear that here than to find out in your session.

Then read the score. Voices are harder to rebuild than synth sounds, for a simple reason: a sound that came out of a synth can, in principle, be made again by a synth, while a voice can only be approximated. Breath, noise and the small wobble in a sung note are exactly the parts a patch reproduces least well. Some chops come back close; others come back as a synth that only hints at the vowel. The score tells you which you got. If it is low, try a shorter, steadier slice of the same vowel, or a different vowel from the same take.

5. Make it an instrument

Drag the file onto the Serum 2 window and it loads. To keep it, save it to the user presets folder that Serum 2's preset browser points at.

Now treat it like any other patch, because that is what it is:

• Pad. Lengthen the amp attack and release, and play chords. A chop in a sampler turns a chord into a pitched-up and pitched-down copy of one recording; a patch plays every note of the chord as itself.
• Lead. Shorten the release, add glide, and put vibrato on the mod wheel with an LFO on pitch.
• Movement. Automate the filter cutoff so the vowel colour opens and closes over a phrase.
• Width. Add unison voices and a little detune. A mono vocal chop gets wide without a stereo widener.
• Clean up. The effects chain that came back is a starting point. Strip it and build your own if it fights your mix.

Chop or preset?

	The chop, in a sampler	The preset, in Serum 2
Sounds like	exactly the recording, word and all	the vowel's colour, as a synth
Other notes	transposed, so formants shift and it can sound chipmunky or muddy	played by the synth at every note
Length	stretched, or looped	as long as you hold the key
Chords	stacked copies of one recording	real chords
Editing	start, end, envelope, filter	every oscillator, envelope, modulation and effect

Use both. The chop for the hook where you want the actual word, the preset for the chord bed underneath it that shares its colour.

From code

The same run, in Python:

import audial

audial.config.set_api_key("your_api_key")
audial.config.set_user_id("your_user_id")

result = audial.sound2serum("ahh_c4.wav")
print(result["preset"])  # ...ahh_c4.SerumPreset
print(result["scores"])  # how close the render came to the chop

Over REST, the chop goes inline as multipart form data:

curl -X POST "https://api.audialmusic.ai/api/functions/run/sound2serum" \
  -H "X-API-Key: your_api_key" -H "X-User-ID: your_user_id" \
  -F "userId=your_user_id" \
  -F "original=@/path/to/ahh_c4.wav"

That answers with an execution id at once; poll GET /api/db/{userId}/execution/{exeId} until state is completed. With the Audial MCP server installed, "turn every chop in ~/Samples/vox into a Serum 2 preset and list their scores" is a complete instruction.

Where to go next

Make one on the Serum 2 preset generator. If you do not own Serum 2, the same chop works with the free target: see the Vital preset generator, and Vital vs Serum 2 for resynthesis for how the two compare.