Honest comparison · Updated 2026-08
Audacity vs Reverie for time stretching
Audacity ships Paulstretch built in, and it is the real algorithm, free, on every platform. So the useful question is not which one stretches audio. It is what happens after the stretch, and whether the result stays alive for thirty minutes. Here is an honest breakdown, including the cases where Audacity is simply the right answer.
Which one should you use
Most people asking this question are really asking one of two different things. Here they are separated.
Use Audacity if
- ✓You want to hear what extreme time stretch sounds like and you have not tried it yet
- ✓The stretch is short and the frozen, static quality is the effect you want
- ✓You need to edit, record, repair or convert as well
- ✓You want to spend nothing
- ✓You want the reference algorithm with no interpretation layered on it
Use Reverie if
- ✓You have tried Paulstretch on a long file and found it goes static
- ✓You need a finished bed at an exact length, not raw material
- ✓You want the stretch inside a designed chain rather than as a single effect
- ✓You need to reproduce a result on another machine, months later
- ✓You are producing this regularly and the manual chain has become the bottleneck
The short version
| Audacity | Reverie | |
|---|---|---|
| What it is | A free, open-source audio editor | A soundscape generator built on a 37-module DSP engine |
| Paulstretch | Built in, two parameters, free | Dreamtime, an evolution of the same algorithm, with six engineering differences |
| What you get out | A stretched file you still have to process | A mastered soundscape at the length you asked for |
| Long renders | Tends toward static after a minute or two | Built to keep moving across a thirty minute render |
| Price | Free, GPL | Free tier, or 29 EUR once for Pro |
| Editing | A full multi-track editor | No editing. It generates, it does not edit |
You already have Paulstretch
This is the single most common misunderstanding, so let us clear it first: Paulstretch is not a plugin you need to hunt down and install. It has been a built-in effect in Audacity for years. If Audacity is installed, Paulstretch is installed.
It exposes exactly two parameters, and the Audacity manual is unusually blunt about what they do. Stretch Factor sets how much longer the result will be: set it to 10 and one minute becomes roughly ten. Time Resolution is the size of the analysis window in seconds, and it is the real control. Small values give good time resolution and poor frequency resolution. Large values give the opposite. The manual recommends 0.25 seconds for most music, and that is a genuinely good starting point.
The manual also says something worth quoting, because it is the honest limit of the effect: use Paulstretch only for an extreme time-stretch or "stasis" effect. Stasis is exactly the right word. That is what the 2006 algorithm does, and it is why a two minute stretch sounds gorgeous and a thirty minute one sounds like a held chord that forgot to end.
- 01Select the region you want to stretch. Paulstretch will refuse to run if the selection is shorter than the Time Resolution value.
- 02Effect menu, then Pitch and Tempo, then Paulstretch.
- 03Set Stretch Factor. Start at 8 to 10 to hear what the source has in it.
- 04Set Time Resolution to 0.25 s for music, higher for pads and drones, lower if you want the transients to survive as texture.
- 05Apply, then wait. Long stretches are slow, and Audacity renders the whole thing before you hear any of it.
One gotcha worth knowing: applying Paulstretch to a clip with an active amplitude envelope discards the envelope and returns the audio to full amplitude. Flatten your levels before you stretch, not after.
Read the full explainer: what Paulstretch is and how the algorithm works
What each one is actually for
Audacity is an editor
It is a free, open-source, cross-platform audio editor with a thirty year lineage and an enormous effect list. Recording, cutting, noise reduction, spectral repair, batch processing, format conversion. Paulstretch is one effect among a hundred, and it is there because someone contributed it, not because Audacity is a soundscape tool. If your job is to edit audio, Audacity is one of the best free tools that has ever existed and nothing here argues otherwise.
Reverie is a generator
It does one job: take a source recording and produce a finished long-form ambient piece from it. There is no timeline, no cut tool, no track list. You pick a source, pick a character, set a length, and it renders a mastered file. The stretch is one stage in a chain of 37 DSP modules, and the whole chain exists to keep the result interesting for as long as you asked for.
Where Audacity is the right answer
There are real cases where reaching for anything else is a mistake, and we would rather say so than pretend otherwise.
It costs nothing, forever
Free, GPL, no account, no license server, no tier. For a student, a hobbyist, or anyone stretching audio twice a year, that is the end of the argument.
You need to edit, not generate
Trimming, splicing, fading, noise reduction, spectral repair, converting formats, batch processing a folder. Reverie does none of this and is not trying to.
You want the 2006 algorithm exactly
If you are recreating a specific documented result, or you want the unmodified reference behaviour of Paul Nasca's original, Audacity gives you that with two knobs and no interpretation on top.
Short stretches, where stasis is the point
Stretch four seconds to forty and you get a beautiful frozen bloom. This is the case Paulstretch was designed for, and it does it perfectly.
Reach for Audacity when:
- •The task is editing rather than generating
- •You need a short, extreme, frozen stretch
- •Budget is zero and must stay zero
- •You are on Linux, ARM, or an old machine and want something that definitely runs
What Reverie adds on top
Dreamtime is our implementation of the Paulstretch idea, rewritten in native C++ with six documented differences from the 2006 baseline. Every one of them exists to solve the same problem: what happens when you ask for thirty minutes instead of thirty seconds.
The read head wanders
The original reads through the source linearly. Dreamtime follows three sinusoids at golden ratio frequencies with plus or minus 30 percent drift, so the material under the window keeps changing. This is the single biggest reason a long Dreamtime render stays alive where a long Paulstretch render goes static.
It reads several places at once
For sources over three seconds, two to four non-contiguous micro segments are extracted and crossfaded, with silent zone detection and retry, instead of one contiguous window. Since 1.14 those segments are also loudness matched, so a quiet intro and a loud chorus no longer produce audible level steps at the boundaries.
Phase is blended, not re-rolled
Rather than randomizing phase from scratch on every frame, Dreamtime blends 70 percent new random phase with 30 percent of the previous frame, wrapping by the shortest arc. You keep the drone character and lose most of the grainy digital artefacts.
You get the length you asked for
You set a target duration, not a stretch factor, and the engine orchestrates pre-loop, re-stretch and sigmoid fades to land on it exactly. In Audacity you compute a factor, apply, discover you overshot, undo, and try again.
The stretch is not the whole chain
Downstream of Dreamtime sit 36 other modules: reverb, spectral freeze, spectral cross-synthesis, resonators, breathing, micro drift. A style like Cathedral or Deep Space is a designed chain, not a preset on one effect. What lands on disk is finished, loudness normalized and true-peak limited.
The whole thing is one number
Every render carries a USeed, a portable code capturing the full chain, every parameter and the seed. Paste it on another machine and you get the same chain back. Audacity has no equivalent: your settings live in your memory or in a text file.
Reach for Reverie when:
- •You need twenty or thirty minutes of ambience that does not repeat or go static
- •You want a finished, mastered file rather than raw material
- •You are producing a lot of these and want reproducibility across machines
- •You want the stretch combined with reverb and spectral processing without building the chain by hand
Side by side
Audacity 3.x built-in Paulstretch vs Reverie 1.14.0. Both were checked against current documentation.
| Feature | Audacity | Reverie |
|---|---|---|
| Price | Free, GPL | Free tier, Pro 29 EUR once |
| Platforms | macOS, Windows, Linux | macOS, Windows, Linux, iPadOS |
| Stretch algorithm | Paulstretch, 2006 reference | Dreamtime, six documented differences |
| Stretch controls | Stretch Factor, Time Resolution | Target duration, variation, stretch, brightness |
| Set output length directly | No, you compute a factor | Yes, you set minutes |
| Movement across a long render | Static by design | Spectral drift and multi-segment reading |
| Effects after the stretch | Apply them yourself, one at a time | 36 further modules, chained automatically |
| Mastering | None, you do it | Gated LUFS normalization, tone EQ, true-peak limiting |
| Reproducible settings | Manual notes | USeed, one portable code |
| Preview before rendering | Limited | Yes, plus per-card looping playback |
| Editing, recording, repair | Yes, extensive | No |
| Batch and scripting | Yes, Macros and Nyquist | No |
| Runs offline | Yes | Yes, after license validation |
What it costs
Audacity: free
Free and open source under the GPL, on macOS, Windows and Linux. No account, no tier, no upsell. Paulstretch is included at no cost because everything in Audacity is included at no cost.
Reverie: free tier, or one payment
The free tier renders up to 3 minutes with 5 presets, which is enough to decide whether the sound is for you. Pro is 29 EUR paid once, with 30 minute renders, all 20 presets, the Custom Chain Builder and four export formats. No subscription, and 1.x updates are included.
Both run entirely on your machine. Neither uploads your audio anywhere, and neither is AI generated. These are signal processing tools, not models.
They work well together
This is not an either-or, and in practice most of our users have both installed. Audacity is where you prepare a source, and Reverie is where you turn it into a piece. Since 1.14 Reverie also has its own sample region selection, so trimming a source no longer requires the round trip, but everything else in the preparation stage is still Audacity territory.
- 01Record or import your source in Audacity
- 02Clean it up: noise reduction, de-click, level it, cut the parts you do not want
- 03Export as WAV, 24-bit, at the source sample rate
- 04Load it into Reverie, set the region, pick a style, set the length, render
- 05Take the result back into Audacity if you need to trim, splice or convert it
Common questions
Is Paulstretch built into Audacity, or do I need to install it?
It is built in. Open the Effect menu, then Pitch and Tempo, and Paulstretch is in the list. There is no download and no plugin folder to manage. If you have been searching for a Paulstretch plugin for Audacity, you already have the effect.
What is Paulstretch in Audacity?
It is an extreme time-stretch effect based on the algorithm Paul Nasca published in 2006. It works in the frequency domain and deliberately randomizes phase, which smears transients into a continuous wash instead of trying to preserve them. That is why it turns music into pad textures rather than into slowed-down music.
What is the best Time Resolution setting?
The Audacity manual recommends 0.25 seconds for most music, and it is a good default. Go higher, 0.5 to 1 second, for smoother pad textures with less rhythmic residue. Go lower, 0.1 to 0.15, if you want some of the original rhythm to survive as texture. Whatever you pick, the value must be shorter than your selection or the effect will not run.
Why does my long Paulstretch render sound boring?
Because that is what the algorithm does. It reads through the source linearly at a constant rate, so once the spectral content settles, nothing new arrives. The Audacity manual itself describes the result as a stasis effect. For a two minute render that is the appeal. For a thirty minute one it is a limitation, and it is the specific problem Dreamtime's spectral drift and multi-segment reading were built to solve.
Can Audacity stretch audio 100 times?
Yes, the Stretch Factor will accept it. What you should expect is a very long render time and a result that is essentially one sustained texture. Whether that is what you want depends entirely on the source and the intent.
What are Audacity's other time stretch options?
Change Tempo changes length without changing pitch, Change Speed changes both together like a tape machine, Change Pitch does the opposite, and Sliding Stretch ramps continuously from one value to another across the selection. These are conventional time-stretch tools that try to preserve transients. Paulstretch is the one that deliberately does not.
Is Reverie a replacement for Audacity?
No, and it would be a strange claim to make. Reverie has no editor, no recorder, no repair tools and no batch processing. It generates soundscapes. If you need to cut a file in half, you still want Audacity.
Is Reverie AI-generated?
No. It is a deterministic signal processing chain: 37 DSP modules, a seeded random number generator for parameter selection, and no model anywhere in the pipeline. The same seed produces the same file. Nothing is sent to a server.
Try the stretch you already own first
Genuinely: open Audacity, select ten seconds of something, run Paulstretch at factor 10 with Time Resolution at 0.25, and listen. If that sound is what you were looking for, you are done and it cost you nothing. If you liked it but found it ran out of ideas when you pushed it to twenty minutes, that gap is the reason Reverie exists, and the free tier will tell you in three minutes whether we closed it.