edytlab

Compare

AI audio editors compared: where they run and what they do

This page sets edytlab beside eight other tools that edit or clean up audio with AI help: where each runs, its license and price model, whether you can drive it by chat or through an agent, and whether it splits stems or transcribes speech. Every cell links the page that states it and the day we read it. edytlab has not shipped stem separation, and its row says so.

By · Prices and features checked on · How we compare

How to read the table

Each cell says what that product's own page says, in a line or two, and links the page and the day we read it. How we compare, and why, is on the about page. A few things to know:

  • “Not described” means the page we read does not say. It is a statement about that page, and not a claim that the product lacks the feature; the product may do it, or may have said so elsewhere.
  • “Chat or agent control” means you can ask in words for the editor to act, through a built-in assistant or an MCP client you connect.
  • Cells with no read date in edytlab's own row link edytlab's docs, which you can check against the app. The row is last, with the same columns as the rest.
  • Prices are the ones the page showed on the day, in the currency shown. Check the product's site before you decide.

The matrix

Product
Where it runs
License
Price model
Chat or agent control
Stem separation
Transcription
Platforms
Audacity
with its optional OpenVINO AI plugins
Its AI plugins “run 100% locally on your PC”, per its plugin page.
GPL v3.
audacityteam.org/audacity-4 · read
Free.
audacityteam.org · read
Not described
audacityteam.org · read
A plugin, Music Separation, splits a track into drums, bass, vocals and other instruments.
A plugin, Whisper Transcription, writes a label track.
Windows, macOS and Linux.
audacityteam.org · read
Audacity with the AudacityMCP server
a separate, third-party project
Drives Audacity on your computer through its named pipe, with no cloud for audio processing, per its README. The assistant is your own MCP client.
Apache License 2.0.
Its README speaks of keeping the project “maintained and free for everyone”.
Yes: MCP, with a tool set its README counts at 144. It works with Audacity 3.x and cannot control Audacity 4.
Not described
Experimental: faster-whisper, offline, set up separately.
Install steps for Windows, macOS and Linux.
Audionaut
an open-source multitrack editor
A desktop app. Its stems are made on your machine. The assistant is your own MCP client.
audionaut.app · read
GPLv3.
audionaut.app · read
Free.
audionaut.app · read
Yes: MCP. Each change from an assistant arrives as one undo step.
audionaut.app · read
Yes: drums, bass, other and vocals, on your machine.
audionaut.app · read
Not described
audionaut.app · read
macOS, Windows and Linux.
audionaut.app · read
Descript
with its Underlord AI co-editor
Cloud-based projects: its podcasting page says “Because it’s all in the cloud”. Where its AI runs is not stated.
descript.com/podcasting · read
No open-source license stated
descript.com/pricing · read
A free plan, and paid plans from $16 a month per person billed annually.
descript.com/pricing · read
Yes: Underlord, an AI co-editor in beta. Each prompt uses credits.
Not described
descript.com/pricing · read
Yes: transcription in 25 languages, and edits are made in the transcript.
descript.com/pricing · read
Desktop apps for macOS 14 or later and Windows 11 or later.
descript.com/download · read
Adobe Podcast
Enhance Speech and Studio
Cloud: “all on the web”, in the browser.
podcast.adobe.com/en/plans · read
No open-source license stated
podcast.adobe.com/en/plans · read
Premium comes with the Adobe Firefly Pro plan. The plans page shows no price.
podcast.adobe.com/en/plans · read
Not described
podcast.adobe.com/en/plans · read
Not described
podcast.adobe.com/en/plans · read
Not described
podcast.adobe.com/en/plans · read
Web.
podcast.adobe.com/en/plans · read
Auphonic
automatic audio post-production
Hosted: you upload audio or video to the service.
auphonic.com · read
No open-source license stated
auphonic.com/pricing · read
Free for 2 hours of audio a month; paid recurring plans and one-time credits.
auphonic.com/pricing · read
Not described
auphonic.com · read
Not described
auphonic.com · read
Yes: multilingual speech-to-text, with a transcript editor.
auphonic.com · read
Web, with an API, watch folders and a command-line tool.
auphonic.com · read
Cleanvoice AI
podcast cleanup as a service
Cloud: it stores your original and edited files for 7 days, then removes them.
cleanvoice.ai/pricing · read
No open-source license stated
cleanvoice.ai/pricing · read
A free trial, then pay-as-you-go credits or a monthly subscription.
cleanvoice.ai/pricing · read
Not described
cleanvoice.ai/pricing · read
Not described
cleanvoice.ai/pricing · read
Yes: transcription and summary.
cleanvoice.ai/pricing · read
A web app, plus an API.
cleanvoice.ai/pricing · read
ocenaudio
a simple cross-platform editor
A desktop app: native applications for each platform.
ocenaudio.com/features · read
No license stated
ocenaudio.com/features · read
A free download.
ocenaudio.com/download · read
Not described
ocenaudio.com/features · read
Not described
ocenaudio.com/features · read
Not described
ocenaudio.com/features · read
Windows, macOS and Linux.
ocenaudio.com/features · read
edytlab
this site’s project
A desktop app. Audio is processed on your computer; the chat goes to the AI provider you pick, or stays on your machine with Ollama.
The app is free. A hosted AI provider bills you for tokens; Ollama is free and runs on your machine.
Yes: a built-in agent calls the editing tools from your request. It is also an MCP client.
Not yet. The tool exists and returns an error today (#385).
Yes, on your computer, after a one-time download of the speech model (about 139 MB) that you start from Settings.
macOS, Windows and Linux. The installers are unsigned.
Not built yet
Stem separation has not shipped in edytlab. Its tool exists but returns an error today, so edytlab does not offer vocal isolation. Where another product does this, it does it and edytlab does not. It is not finished; progress is tracked on GitHub: stem separation.

Product by product

A few sentences on each, taken from its own pages. Where a product does something edytlab does not, the row above and the note below say so.

Audacity

Audacity is free and, per its Audacity 4 page, licensed GPL v3. Its AI features are optional OpenVINO plugins. Its plugin page lists music separation into drums, bass, vocals and other instruments, noise suppression, music generation and continuation, and Whisper transcription into a label track, and says they run locally. A June 2025 post says the plugins reached macOS with Audacity 3.7.4, as a beta of sorts, and none of the pages we read says whether they work with Audacity 4.

More: Audacity alternative.

Audacity with the AudacityMCP server

AudacityMCP is a separate project and not Audacity's own. Its README gives the license as Apache License 2.0 and says it connects MCP-compatible AI assistants to Audacity through a tool set that its README counts at 144. It also says it works with Audacity 3.x and cannot control Audacity 4, which removed the scripting interface it uses, and it lists local transcription with faster-whisper as an experimental feature that needs separate setup. We have not run it.

More: Chat control inside Audacity.

Audionaut

Audionaut's home page calls it a free, open-source desktop application, under the GPLv3 license, for macOS, Windows and Linux. It says the app speaks MCP, so assistants such as Claude can inspect and edit projects, with each change arriving as one undo step. It also lists splitting a clip into drums, bass, other and vocals, which it says runs entirely on your machine with a model downloaded once. The assistant is a client you connect, such as Claude Code; edytlab has its own agent built in.

Descript

Descript's pricing pagelists a free plan and paid plans, “Text-based editing: Edit video by editing a doc”, and transcription in 25 languages. Its Underlord AI co-editor is in beta: its help pagesays each prompt uses credits and that when Underlord runs a tool you are “charged for both the agent brain and the tool itself”. Its podcasting page says “Because it’s all in the cloud, you’ll be able to view different versions and restore them if you want”, and the pages we read do not say where its AI runs.

Adobe Podcast

Adobe Podcast runs in the browser. Its plans pagedescribes Enhance Speech, which it says will “Remove noise and echo from voice recordings”, and Studio (beta), which records, edits and enhances in the browser. Premium comes with the Adobe Firefly Pro plan: Enhance Speech up to 4 hours a day, with files up to 1 GB. The plans page shows no price. edytlab does not have a speech-enhancement model; its noise_reduction tool is spectral subtraction.

Auphonic

Auphonic is a hosted service: you upload audio or video. Its pricing page says it is free for 2 hours of processed audio a month and sells recurring and one-time credits. Its home pagelists noise and reverb reduction, an “Intelligent Leveler”, loudness specifications (“Define a target loudness, true peak limit, MaxLRA, and more”) and multilingual speech-to-text. edytlab also sets a track to a loudness target by name, with normalize_loudness, on your own computer.

Cleanvoice AI

Cleanvoice is a cloud service and API. Its pricing pagesays “We store both your original and edited files for 7 days. After 7 days, we remove your files permanently.” and lists a filler words remover, a background noise remover, a silence remover, a mouth sounds and breath remover, and transcription and summary. It offers a free trial, then pay-as-you-go credits or a monthly subscription.

ocenaudio

ocenaudio's download page offers a free download. Its features page lists VST plugin support, real-time preview of effects, multi-selection editing and a spectrogram view. The pages we read state no license and describe no AI features.

edytlab

edytlab is a free, MIT-licensed desktop app for macOS, Windows and Linux in which an AI agent calls local audio tools from a request you type. Audio is processed on your computer; the chat goes to the AI provider you pick, or stays on your machine with Ollama, and a hosted provider bills you for tokens. Transcription runs on your computer after a one-time download of the speech model that you start from Settings. Stem separation has not shipped, so the tools reference lists a tool that returns an error today.

More: Getting started.

Which should you pick?

It depends on the job. Find the sentence that describes yours.

  • You want to describe an edit in words and keep the audio on your computer. That is edytlab. Audionaut and AudacityMCP also let an assistant drive an editor on your computer, with the assistant a client that you bring. Descript's Underlord uses credits.
  • You need to split a mix into stems now, on your computer. Audionaut lists it as a built-in feature, and Audacity lists a music separation plugin, if the plugin supports the version of Audacity you run. edytlab has not shipped stem separation.
  • You want an assistant to drive an editor you already use. AudacityMCP connects assistants to Audacity 3.x, and its README says it cannot control Audacity 4. Audionaut speaks MCP itself. edytlab is its own editor, with the agent built in.
  • You want to edit audio by editing its transcript. Descript's pages describe a text-based editor, with projects in the cloud and credits for its AI. edytlab does it on your computer, after you download the speech model once. Auphonic and Cleanvoice list transcription too, as services that process what you upload.
  • You want one-click cleanup as a service. Auphonic, Cleanvoice and Adobe Podcast are hosted: you upload your files. Check each one's limits and how long it keeps them on its own site first.
  • You want a plain, free editor that takes plugins. Audacity lists VST3 and Nyquist plugin support on its home page, and ocenaudio lists VST support on its features page. edytlab has no plugin hosting.

Comparisons in detail

Two products have a page of their own, with a table of facts, a section on where the other product is better and a verdict: Audacity and Adobe Audition. Adobe Audition is not in the matrix: it is a multitrack workstation, and its product page does not describe chat or agent control. For how edytlab handles what stays on your machine and what leaves it, see the local-first page.

Sources

Everything stated about the products above comes from these pages, read on the dates shown.

  1. Audacity AI plugins page (plugins.audacityteam.org/ai-plugins/ai-plugins), checked
  2. Audacity 4 page (audacityteam.org/audacity-4), checked
  3. Audacity home page (audacityteam.org), checked
  4. AudacityMCP README (a third-party project) (github.com/xDarkzx/Audacity-MCP), checked
  5. Audionaut home page (audionaut.app), checked
  6. Descript podcasting page (descript.com/podcasting), checked
  7. Descript pricing page (descript.com/pricing), checked
  8. Descript help: Underlord (Beta) (help.descript.com/getting-started/underlord-beta-your-ai-co-editor-in-descript), checked
  9. Descript download page (descript.com/download), checked
  10. Adobe Podcast plans page (podcast.adobe.com/en/plans), checked
  11. Auphonic home page (auphonic.com), checked
  12. Auphonic pricing page (auphonic.com/pricing), checked
  13. Cleanvoice AI pricing page (cleanvoice.ai/pricing), checked
  14. ocenaudio features page (ocenaudio.com/features), checked
  15. ocenaudio download page (ocenaudio.com/download), checked
  16. Audacity blog: OpenVINO AI plugins for macOS (audacityteam.org/blog/openvino-mac), checked
About this comparison
edytlab is an independent open-source project. It is not affiliated with, endorsed by or sponsored by any of the other products on this page or their owners. Their names, and other product names, are trademarks of their respective owners and are used here only to identify the products being compared. Everything stated about the other products comes from their own pages, linked above, and was last checked on 24 October 2026. Features and prices change; check each product's site before you decide. Found a mistake? Tell us and we will correct it (how corrections work, how we compare).