Infinite benchmarks

At a glance

We measured Infinite against the tools you already use, on one ordinary laptop, with the same tests for everyone, and we publish the numbers that are good for us and the ones that are not.

Where Infinite is behind

• Idle CPU. With nothing loaded Infinite sits at 12.3 % of one core. Logic, Blender and Vital sit under 4 %. We are working on it.
• Some built-in effects cost more CPU than the plugin equivalents: EQ, Limiter, Drive and Chorus (Section 4).
• Instancing. 10,000 instanced objects cost 3.1 ms a frame in Infinite and 1.3 ms in Blender.
• The heaviest projector scene (four 4K videos, three output windows, a camera and Syphon at once) drops frames. Normal and 1080p scenes do not.
• Default levels. Reverb, Delay, Chorus and Flanger come out louder than they go in at their default settings.

How we tested

Everything here was measured on one machine: a MacBook Air with an Apple M2 and 8 GB of memory, running macOS. That is the base model, on purpose. If it is fast here it is fast on yours. Nothing was measured on Windows or Linux.

• Black box. Notes or audio go in, the result comes out and is measured. We never open or read the other product.
• Same test for everyone. A competitor plugin runs inside Infinite's own plugin host, receives the same notes or the same test signal, and is measured by the same code as Infinite's own nodes.
• Default settings. Every instrument and effect is tested as it loads, so presets are not hand-picked to flatter anyone. Where a default does nothing (an EQ with no bands) only its CPU cost is meaningful, and the report says so.
• CPU is the render cost. Milliseconds of processing per second of audio, with the machine otherwise idle. Lower is better, and 10 ms/s is 1 % of one core. Differences under about 1 ms/s are noise.
• Versions. Infinite 0.6 (a development build matching the 0.6.0 release); Ableton Live 12.4.6, Bitwig Studio 6.0.3, Logic Pro 12.3.1, TouchDesigner 2025.32820, VCV Rack 2.6.6, Blender 5.1.0, BespokeSynth 1.3.0, Vital 1.5.5, Serum 2.0.16, Diva 1.4.5, FabFilter Pro-R 2 2.04, Pro-C 2 2.20, Saturn 2 2.11, plus the installed versions of Valhalla and Kilohearts plugins.
• Reproduce it. The tools are in the repository under tools/re-survey. Section 8 lists the commands.

Dates: instruments, effects and apps were measured on 17 October 2026; the internal tests and 3D/render comparisons on 12 to 16 October 2026.

Instruments

Eight notes from MIDI 24 to 108 are played into each synth, and what comes out is analysed. Pitch error is the worst note, in cents (a hundredth of a semitone). Aliasing is the energy that is not part of the note, in decibels; more negative is cleaner, and the high notes at 96 and 108 are where cheap synths fold back and sound harsh. CPU per voice is the extra render time for each added note.

What this means

• Infinite's Wavetable and Analog synths are as accurate and as clean as Vital and Serum 2: 0.2 cents, and about -81 to -85 dB off-note energy even at the top of the keyboard.
• Wavetable is also light: 1.45 ms/s per voice against 2.82 for Vital and 3.23 for Serum 2. Analog is the heavier of the two at 9.05 ms/s.
• Diva is measured as it loads and sits an octave down with a modelled, imperfect oscillator. Its figures describe an analog-style design, not a flaw.
• Infinite's synths respond to how hard you play across a 24 dB range. Vital and Serum 2 ignore velocity until you route it yourself, so that column compares defaults, not capability.
• Not comparable: Kilohearts Phase Plant and FabFilter Twin 3 came up silent or unusable with no patch loaded, so they are left out rather than guessed at.

Effects

A test tone or noise goes through each effect at its default settings. CPU is the render cost in ms per second of audio (lower is better). Latency is the delay the effect adds after the host compensates for it.

What this means

• Infinite is lighter on Reverb, Delay and the compressor, and about level on Flanger and Phaser.
• The plugins are lighter on EQ, Limiter, Drive and Chorus. The EQ gap is the clearest: 14.5 ms/s for Infinite against 3.0 for Pro-Q 4. These are on our list to fix.
• Infinite effects add no delay once compensated (0 ms). ValhallaRoom reports 11.4 ms.
• At default settings Infinite's Reverb, Delay, Chorus and Flanger are louder than the input and can clip. FabFilter Pro-R 2 and ValhallaDelay stay within about 1 dB. We list this as a known issue.
• An EQ with no bands, a limiter that is not limiting and a compressor with no gain reduction are close to a wire. Their distortion figures match the test rig, so only their CPU cost is a fair comparison.

3D, video and export

Blender runs headless through its public Python interface; Infinite runs the matching patch. Cloth and subdivision are one simulation step or one deform per frame. Movie export times the whole render and encode, measured as the slope between a 1-second and a 3-second render so start-up cancels out. Blender here is a general-purpose renderer, so this compares export throughput on matched scenes, not image style.

What this means

• A 10-minute film at 30 fps is about 18,000 frames: roughly two minutes to export from Infinite at the 720p rate above, about three hours from Blender. Audio and picture stay locked: drift over ten minutes is zero.
• Instancing is the one 3D test Infinite loses, by a factor of 2.4, and is flagged below target.
• A rendered arrangement runs at 2.25× real time on the reference scene, with no audio/video drift.

Launch, memory and download size

Each app is opened with nothing loaded and left alone. Window is the time from opening it until the first window appears (the first launch, then a warm launch). Idle CPU is the share of one core used while doing nothing, over the whole app including its helper processes. Memory is what Activity Monitor shows.

What this means

• Infinite is a small install: 45 MB, against 874 MB to 8 GB for the DAWs and TouchDesigner. Only VCV Rack and Vital are as small.
• Memory is lower than Ableton, Blender, Bitwig, TouchDesigner and BespokeSynth. Only Logic sits lower.
• Infinite opens its first window in under a second here, the quickest first launch of the apps tested. Infinite, Blender and Vital were re-measured on a quiet machine; the rest come from the first pass.
• Idle CPU is mid-pack: higher than Logic, Blender and Vital (all under 4 %), lower than Bitwig, BespokeSynth, Ableton and TouchDesigner. VCV Rack's high figure is its audio engine running continuously by design.

How fast Infinite is on its own

These are Infinite's own targets, tested on the same base laptop. They matter for live use: if the audio buffer is missed you hear a click, and if a frame is missed you see a stutter.

Correctness checks

The full list at the end of this report covers every further check, all passing or listed as known exceptions: pitch within one cent at 44.1, 48 and 96 kHz; no clicks when notes start or stop; no stuck notes after release; the same patch renders bit-for-bit the same twice; no hum or DC offset from effects; blend modes within 0.125 of 255 against the W3C formulas; no broken meshes. The three known exceptions are the insert-level jump on some effects, a small level error on stretched drum hits, and some winding-order flips in the Screw geometry node.

Not yet measured

Four long or heavy tests are defined but have not been run: a 4,000-node canvas under pan, select and delete; an edit storm while audio plays; a multi-hour soak with memory growth and temperature; and a real-time audio-reactive scene of 10,000 instances. Until they are run we do not claim results for them.

What we did not test, and how to reproduce this

• No head-to-head against other DAWs on audio load or latency. Matched track-count tests in Ableton, Bitwig and Logic need a hand-built project in each, and round-trip latency needs a loopback device.
• No live visual comparison with TouchDesigner or Resolume yet. The harness exists; it needs a manual step inside TouchDesigner.
• One machine, one operating system. Numbers on faster hardware will be better, and Windows and Linux are unmeasured.
• Defaults, not best case. A plugin with a tuned preset would measure differently.
• Our numbers only. Every figure here was measured by us on this machine. No vendor claims or published specs are used.

Run it yourself

git clone https://github.com/n1m21n/Infinite && cd Infinite
python3 tools/re-survey/synth_probe.py run Vital "Serum 2" node:Wavetable
python3 tools/re-survey/plugin_probe.py run dry node:Reverb ValhallaRoom
python3 tools/re-survey/app_bench.py run Logic Blender /Applications/Infinite.app --runs 2
python3 tools/re-survey/render_bench.py infinite

Corrections and re-runs are welcome: open an issue at github.com/n1m21n/Infinite or tell us on Discord.

Full results

Every one of the 97 metrics we track, with its target and the worst result across all the nodes or scenes tested. "Reported" means there is no pass/fail target (a design choice is compared, not judged). The competitor comparisons are in Sections 3 to 6.

AV integration

Note nodes

Project files

Stability

Audio I/O

Built-in instruments

Effects

Plugin hosting

Timeline audio

Project files

3D engine

Compositing

Geometry

Bench: audio

Bench: canvas

Bench: fundamentals

Bench: live edit

Bench: media I/O

Bench: memory

Bench: offline render

Bench: projector

Bench: soak

Built-in instruments

Live scene

Questions people ask

Is Infinite faster than Vital or Serum 2?

On sound quality they are level: all three measure about 0.2 cents of pitch error and about -81 to -85 dB of off-note energy. Infinite's Wavetable synth uses 1.45 ms of CPU per second of audio per voice, against 2.82 for Vital and 3.23 for Serum 2.

How fast does Infinite export video compared with Blender?

On matched test scenes on a base MacBook Air M2, Infinite exported ProRes 4444 at 170 fps at 720p and 44 fps at 4K, against 1.6 fps and 0.5 fps for Blender. These are different renderers, so this compares throughput, not image style.

How much memory and disk does Infinite use?

244 MB of memory with the app open and idle, and 45 MB on disk. That is less memory than Ableton Live, Blender, Bitwig, TouchDesigner and BespokeSynth, and a fraction of the size of the DAWs tested (Ableton Live is 8 GB on disk).

Where is Infinite slower than other tools?

Idle CPU (12.3 % of a core with nothing loaded), the built-in EQ, Limiter, Drive and Chorus effects, 10,000-object instancing, and the heaviest multi-4K projector scene. All are listed as below target.

Which computer were these benchmarks run on?

A base MacBook Air with an Apple M2 and 8 GB of memory, running macOS, on 12 to 17 October 2026. Windows and Linux were not measured.