A four-bar loop at 120 BPM runs 8 seconds. The Sonicware LIVEN Texture Lab records samples of up to 6 seconds, so that loop will not fit. Quick granular comparisons break down like this all the time: different source files, different loudness and different settings make the results meaningless.
This exercise fixes all three problems. You use one source file, one written set of starting settings and one level-matching method, then change a single parameter at a time. It works with a plugin, a trial or demo version, or hardware recorded into your DAW.
What You Need Before You Start
The gearnews.com roundup describes granular tools as breaking a sound into tiny grains and reassembling them, with grain position, length, density and tuning as the core parameters. This test isolates those four. It adds jitter because tools such as the Sonicware LIVEN Texture Lab expose it alongside Size, Timing, Density and Position.
You also need a loudness meter in your DAW (LUFS or RMS), a utility or gain plugin, and a notes file or spreadsheet. Pen and paper work too.
The six steps:
- Lock one source sample.
- Write down the baseline settings.
- Capture each tool into its own DAW track.
- Level-match every capture.
- Change one parameter at a time and fill in the checklist.
- Run a blind A/B and make the call.
Step 1: Lock One Source Sample
Pick a 2-4 bar drum loop or a single sustained vocal note. Export it as a 24-bit WAV at your project sample rate, and never edit it again. Every tool gets this exact file.
Run the drum loop first, because transients show how a granular engine smears attacks. Then repeat the exercise with the vocal note. Sustained pitched material exposes grain-boundary artifacts that drums hide.
Check each tool's length limit before you export. The roundup says the LIVEN Texture Lab records up to 6 seconds at 16 or 32 kHz. Two bars at 90 BPM run about 5.3 seconds and fit, while four bars at 120 BPM run 8 seconds and do not. That is arithmetic from the roundup's stated limit. If your tools have different limits, use the shortest loop for all of them.
Step 2: Write Down the Baseline Settings
Write the starting values before you touch a knob. Every granular engine labels things differently (the Waldorf Iridium calls its granular engine a "particle generator"), so note the local name next to each value.
Suggested baseline:
- Grain size: about 75 ms, inside the 50-100 ms range.
- Density: the middle of the control's range. If the tool shows grains per second or overlap, write that number down.
- Jitter or randomness: zero.
- Position: held still at the start of the sample for the first pass.
- Pitch or tuning: original pitch, 0 semitones, 0 cents.
- Grain envelope or window: the default.
- Built-in filter, reverb, delay or chorus: off.
The last item matters. The roundup shows that many of these instruments bundle filters and effects, and the LIVEN Texture Lab, for example, includes a shimmer reverb. A built-in reverb can make one tool sound bigger than another for reasons unrelated to the granular engine. If a tool cannot switch its effects off, record that in your notes.
Step 3: Capture Each Tool Into Its Own Track
For plugins, load the source into each one and play a held note or loop the sample region. For hardware, route the DAW's output into the unit (or play the file from the device) and record its output back at a fixed interface gain. Leave that gain alone for the whole session. For more on this, see related: how to test ai mixing advice in your daw, level-matched.
Record a pass of the same length for every tool, and spend the same amount of time on each parameter move. A sweep that takes 16 bars on one tool and 4 on another is not a fair comparison. Name each clip with the tool, the parameter and the setting, for example "ToolA_Position_sweep".
With jitter at zero, the baseline pass should repeat predictably. If it does not, the tool has built-in randomness. Note that, and capture the pass more than once.
Step 4: Level-Match to Within 0.5 dB
Louder sounds fuller and more detailed, so an unmatched A/B mostly tests volume. Fix that before you listen.
Place a gain or utility plugin after each capture. Play the same section of each, such as the static baseline pass, and read short-term LUFS or RMS over the full section. Adjust the gain until every capture reads within about 0.5 dB of the others. Use the same meter, on the same bus, for all of them.
Apply the matching gain at the track level, not inside the instrument. That keeps every instrument setting identical to your baseline notes.
Granular output can change loudness a lot as density and grain size move. Level-match on the baseline, then watch the peak meter during later sweeps and leave a few dB of headroom so nothing clips. If a sweep ends up much louder on one tool, write "louder at high density" in your notes instead of re-leveling mid-test. That behavior is real information about the tool.
Step 5: Change One Parameter at a Time
Return to the baseline, change one parameter, listen, then reset it. Never move two controls at once. Use the table below as your worksheet. We cover related ground in jazz guitar tone in a daw: do you need a hollowbody? in depth.
| Parameter changed | Test move | What to listen for | Notes | |---|---|---|---| | Grain size | Sweep from about 10 ms up to a few hundred ms | Drum attacks smearing into a wash, flutter or buzz at short sizes, whether the source stays recognizable | | | Density | Sweep from sparse to the maximum | Gaps and stutter at low values, thickening and phasey build-up at high values, level jumps, CPU or voice limits | | | Position | Hold still, then sweep slowly across the whole sample | Clicks at grain edges, whether a still position sounds like a frozen slice, how smoothly the scan moves | | | Jitter | Raise from zero in small steps | Whether it adds natural looseness or just noise, effect on drum timing, change in stereo spread | | | Pitch / tuning | Move in semitones, then in small cents steps | Pitched artifacts and metallic overtones, formant change on vocal, whether pitch tracks cleanly across a range | | | Modulation | Assign one LFO to position or size at a fixed rate | Smooth movement versus stepping, how easily the modulation can be routed | |
Two extra checks apply to every row. First, check stereo width. Does the output stay centered, or do grains scatter left and right? Also listen in mono, because dense overlapping grains can cause phase problems that make a sound collapse.
Second, notice workflow friction. How many menu dives did a grain size change take, and could you do it by ear?
Also read: see finalist 2 pricing: credits vs monthly vs lifetime plan
Step 6: Blind A/B and Decide
Hide the track names or ask a friend to shuffle the order. Listen to the same section of each capture back to back, two or three times, and pick the one you would use. Then reveal which tool it was and compare your picks with your notes.
Score each tool across the parameter rows on a scale you define in advance, for example 1-5 for how usable the result was in your own music. The pattern matters more than the total. A tool might win the position scan and lose at pitch, so decide which of those you will actually use.
If time allows, repeat the exercise with the second source file. A tool that handles drums beautifully can still produce unpleasant artifacts on a sustained vocal note.
Do You Need a Granular Synth, Sampler or Effect?
The roundup calls the difference between a granular synth and a granular sampler marginal. Both break samples into grains. A granular sampler has audio inputs for recording samples on the device, while a granular synth usually needs files imported via USB or another method.
Choose by workflow. If your material lives in your DAW as loops and one-shots, an import-based synth is enough, and your test already mirrored that workflow. If you want to sample a microphone, a guitar or a modular patch straight into the instrument, look for audio inputs. The roundup lists them on the Tasty Chips GR-MEGA (stereo input), the 1010music nanobox lemondrop and the Beetlecrab Tempera, among others.
If you want to mangle live signals, look for a unit that applies the granular engine to its input in real time. The roundup says the Waldorf Iridium, the Sonicware LIVEN Texture Lab and the nanobox lemondrop can do this. For those, repeat Step 3 with a live source and listen for latency or level jumps.
Whether this test works for a particular plugin depends on its trial or demo policy. The roundup covers hardware and does not say what plugin trials exist, so check each developer's page.
What to Expect Next
Your first pass will probably feel slower than watching a demo. In return, your notes should show differences that spec lists never mention. Keep the source file, the baseline sheet and the meter settings. When you consider another tool, or a firmware or plugin update changes behavior, run the same six steps and compare the new notes with the old ones.
