Capo vs Transpose: what is different

Browser pitch shifters look alike from the outside. Both attach to a page, both give you a dial, both change a key without changing a tempo.

The engineering underneath differs in ways you can hear at extreme settings. This article compares Capo and Transpose on eight technical points.

It is written by the Capo team, so treat the framing accordingly. Every figure is sourced, and the points we could not verify are listed at the end rather than guessed at.

Key takeaways

• Both extensions process audio on-device through the Web Audio API. Neither uploads anything.
• Capo exposes three named time-stretch engines; Transpose ships one pipeline it describes as high-quality and low-latency.
• Pitch range: Transpose ±12 semitones; Capo ±12, widening to ±48.
• Speed: Transpose 25–400%; Capo 25–300%, widening to 800%, in 1% steps.
• Both offer formant control and centre-channel vocal reduction on their paid tiers.
• Only Capo ships a parametric equalizer. Only Transpose ships cloud sync.

Prerequisites

To follow the listening tests in this article you will need:

• Either extension installed on Chrome. Both have a free tier with no account.
• A track you know well, ideally with a sustained vocal and audible cymbals.
• Headphones. Laptop speakers hide most of what is discussed below.

1. How each extension attaches to the page

Both work the same way at this layer. A content script locates the HTML5 media element on the page and routes it through the Web Audio API.

In both cases the sound is intercepted between the page and the speakers and rebuilt in real time on your own machine. Nothing is uploaded and no file is produced.

The practical consequence is identical for both: no file is produced, nothing is uploaded, and neither can reach audio outside a browser tab. Desktop applications are out of scope for both.

2. The time-stretch engine

This is the largest technical difference between the two.

What Capo does

Capo ships three engines and exposes the choice as a setting:

Changing the setting swaps the engine live, on the audio already playing — no tab reload needed.

What Transpose does

Transpose does not name its algorithm, but does characterise it: Pro runs a different audio engine from the free tier, described as higher-accuracy with frame-based timing, and as reducing the clicks and noise the free engine produces.

Rather than exposing engine choice, it offers a Low Latency mode under Settings › Audio engine options, which its own documentation says trades audio quality for less delay. Capo ships the same switch — its own low latency mode sits beside the Audio quality setting — so this one is shared ground rather than a differentiator.

Why the difference matters

Every time-stretch algorithm trades quality against latency and CPU. More analysis before committing to a grain boundary gives cleaner joins, at the cost of delay and processing.

Exposing the choice means you can take the latency hit only when you need it. Hiding it means one compromise applies everywhere, chosen by the vendor.

Neither approach is correct in general. Above roughly 0.8× speed the engines are hard to tell apart, so the choice only earns its complexity at the extremes.

3. Pitch range and resolution

Twelve semitones is one octave, which covers standard transposition. Capo widens to four octaves in Settings for sound-design use.

The wider range is opt-in because it costs precision. The same dial travel then covers four times the ground, making small moves harder to land.

Capo’s fine control spans ±50 cents, which is exactly the gap whole semitones leave. Transpose offers fine tuning but does not publish its range.

4. Speed range and step

Transpose reaches 400% by default where Capo reaches 300%. Capo widens further, to 800%, behind a setting.

Both floor at 25%. That floor is a property of the algorithms rather than a product decision: below it, artefacts dominate the signal on any engine.

5. Formant preservation

Both extensions offer formant control on their paid tiers, and both are solving the same problem.

A voice carries resonances set by the size of the throat and mouth. Naive pitch shifting moves those resonances with the note, so the listener hears a body that changed size.

Capo’s formant control spans −1 to +1, where full scale equals one octave of formant shift. Everyday corrections sit near ±0.2.

It requires the Signalsmith or Rubber Band engine. On the lightest engine the control is disabled, because the implementation does not exist there.

Transpose describes the same feature as changing timbre or voice colour without moving the pitch, and names the same engine constraint Capo documents for itself: low latency mode is unavailable while formant shifting is on, in both products. Different mechanism, same trade — formant correction costs something in the engine either way. Neither the scale nor the range is published.

6. Vocal reduction

Both offer vocal reduction on their paid tiers. Neither is source separation; both reduce what is common to the two stereo channels.

Capo exposes four parameters rather than a single slider:

The default range deliberately excludes frequencies below 120 Hz, so the bass and kick survive a process that would otherwise remove them.

Transpose states the mechanism plainly — it lowers centred vocals in stereo mixes — which is the same centre-channel approach, with the same consequence: a voice mixed off-centre, or a near-mono recording, gives it little to work with. Its individual parameters are not published.

7. Filtering and equalization

Both have one, and they are closer than they look: Capo’s six-band layout and its preset curves are matched to Transpose’s on purpose, so moving between them feels the same. What Capo adds is the filter type — each band can be set to high-pass, low-pass, peaking, low shelf, high shelf or notch, so it can be a bell, a shelf or a cut rather than only a gain slider.

The difference we can point to is scope: Capo’s curve is global rather than per-site, so it compensates for your headphones once instead of per tab. Transpose does not state whether its curve is global or per-track.

8. Platform coverage

Transpose lists YouTube, Spotify, Apple Music, SoundCloud, Deezer, Vimeo, Tidal and local files.

Capo lists YouTube, the Spotify Web Player, Apple Music, Tidal, Deezer, Amazon Music, SoundCloud, Bandcamp and podcast players, with a dedicated per-track resolver for each. Local files play through Play local file in Settings, which opens Capo’s own web player.

Apple Music, Tidal and Deezer work in both, and Capo’s per-track memory works there too, so platform coverage no longer decides between them.

Capo runs on Chrome, Edge (published on Edge Add-ons) and Firefox, where the Pro workspace opens from the browser sidebar rather than a side panel. Transpose lists Chrome and Edge.

What could not be verified

Two things, down from five once we read Transpose’s own help pages properly:

• Which algorithm the Pro pipeline uses. It is characterised — higher accuracy, frame-based timing, fewer clicks than the free engine — but not named, where Capo names all three of its own.
• Numeric ranges. Neither the fine-tune range in cents, the speed step, the formant scale nor the equalizer band count is published, so the two cannot be compared digit for digit.

Everything else in this article comes from what each product documents about itself. Where Transpose states a mechanism rather than a number — centred vocals, timbre without pitch, quality traded for latency — that is what is compared.

Comparative listening tests are absent on purpose. Any test built from our own material would flatter the engine that happens to suit it.

Which pitch shifter should you use?

The two extensions overlap heavily on core function and diverge on control surface.

Transpose syncs across machines, which Capo does not. Streaming coverage no longer separates them: both work on Apple Music, Tidal and Deezer, and Capo’s per-track memory follows you there too.

Capo exposes more of the signal chain: engine choice, a four-octave pitch range, an 800% speed ceiling, four vocal-reduction parameters and a parametric equalizer.

Both have free tiers with no account. Test either on a track you know well before paying for anything.