Descript or Dalm: when you edit the recording, and when you only need the words

Descript and Dalm both turn an interview, a podcast episode or any other recording into a transcript with speakers and times. Descript is a video and audio editor that works through the transcript: delete a word and it disappears from the recording, and the same app removes filler words, records your screen and captures remote guests. Dalm keeps no recording to edit. It is a Mac app: the speech in your file is recognised on your Mac and no audio is sent anywhere, while the transcript goes to our servers automatically and a language model reads a meeting’s transcript to write its summary. If you edit episodes, choose Descript; if you only need the words, Dalm.

Dalm and Descript, row by row

	Descript	Dalm
What it is for	Editing video and audio by editing the transcript, and recording for publishing	Turning speech into text: files you bring in, calls, meetings in the room, dictation
Where speech becomes text	Its help says transcription starts once the file has finished uploading to the cloud; its security page names two transcription services: an outside company, and an open-source model hosted in its own infrastructure	Recognised on your Mac; the transcript syncs to our servers automatically
The media afterwards	Audio, video and transcription data are kept on Amazon S3 or Google Cloud	A file you bring in is read and left where it was, with no copy made; the audio of a recording made in the app is deleted once the text exists; for both, the one exception is below; no player and no audio export
Speakers	Detected in Zoom recordings it imports; Rooms gives each guest a separate file	For a file brought in as a meeting, voices are told apart on your Mac and you can name them; as a dictation, it is one voice
Languages	26, which its help says are limited to the Latin alphabet; Russian is listed as not supported; one language per file	Two speech models: “European languages”, which include Russian, and “More languages”, over a hundred; you do not pick the language
Fixing the words	Edit the text and the recording follows; filler words found and removed with AI credits	Edit a line of the transcript; Dalm does not remove filler words
Saving the text	.docx, .txt, .rtf, .md or .html; subtitles as .srt or .vtt with speaker labels	Documents as Markdown, Word or PDF
Calls	Imports Zoom cloud recordings; Rooms records you and your guests	Records calls from the Mac and meetings in the room
Apps	Mac with macOS 14 or later, Windows 11, a web version	Mac with Apple Silicon · macOS 15.0 or later; iPhone app in development

The audio is deleted once the text exists. What Descript does is taken from its own help pages, privacy policy, security page and pricing page, checked on 2 October 2026.

Where Descript is better than Dalm

• It edits the recording. Delete a sentence in Descript’s transcript and that moment is gone from the audio and the picture; it finds filler words such as “um” and “uh” and removes them, using AI credits. Dalm has nothing to cut: it keeps no audio to edit, has no player and exports no sound.
• The media stays with the project. After transcribing, Descript stores your audio, video and transcript on Amazon S3 or Google Cloud, so you can come back and cut the episode again. In Dalm the audio of a recording made in the app is deleted once the text exists, and a file you bring in stays where it was, with no copy in Dalm to cut again.
• Recording made to be published. Rooms records you and your guests, each locally in up to 4K, and the files arrive in the project already transcribed; there is a screen recorder, and Studio Sound takes background noise and echo out of a voice. Dalm records calls and rooms for notes, not for broadcast.
• Windows and a browser. Descript runs on a Mac with macOS 14 or later, on Windows 11, and in a web version. Dalm needs a Mac with Apple Silicon · macOS 15.0 or later; its iPhone app is in development.
• More ways out, and Zoom recordings in. A transcript saves as Word, rich text, Markdown, HTML or plain text, subtitles as .srt or .vtt with speaker labels, and Descript pulls your Zoom cloud recordings in by itself. Dalm saves documents as Markdown, Word or PDF; a Zoom recording comes in as a file you already have.
• Written statements about training, and an audit. Descript’s help says the AI models it runs use no user data, that research models use data only from people who opted in to sharing, and that its outside AI providers are bound not to train on it; its privacy policy adds that projects may be used to improve the service unless you turn that off. Its security page lists a SOC 2 Type II audit. Dalm does not train models of its own, but what the provider of its language model does with the text is not set out on this site, and Dalm has no certification.

Where Dalm is better than Descript

• The audio is not uploaded. Descript’s help says transcription starts once the file has finished uploading to the cloud, and its security page names two transcription services, an outside company and an open-source model hosted in Descript’s own infrastructure. In Dalm the speech is recognised on your Mac and no audio is sent anywhere. The text does leave: the transcript syncs to our servers automatically, and for a file brought in as a meeting, a language model reads it to write the summary.
• A file without a language model, if you choose. Bring a recording in as a dictation and it becomes one block of text that only you can see, and no language model reads it; the voices are not split, and the text still syncs to our servers. Bring it in as a meeting when you need the voices apart and a summary.
• Russian, and more than the Latin alphabet. Descript’s help lists 26 languages, limited to the Latin alphabet, with Russian among those it does not support. In Dalm the default speech model covers English, Russian, German, French, Spanish and twenty more, a second one over a hundred languages, and you do not set the language. How a recording in another language comes out.
• No monthly allowance of minutes. Descript’s Free plan includes 60 media minutes a month, and its Hobbyist, Creator and Business plans 10 to 40 media hours a month. No plan puts a limit on recording minutes, on the length of a recording, or on how long anything is kept. A file you bring in can run up to four hours.
• Notes from the conversation, not only its text. A file brought in as a meeting gets a summary without being asked, and each point keeps the words it came from. Every quote shows you where it was said — open it and the transcript opens at that place.

What Descript and Dalm cost, and what Free includes

	Descript	Dalm
Free plan	60 media minutes a month; 100 AI credits, once; exports at 720p with a watermark	A plan, not a trial; bringing in a file is part of it
Billed monthly	Hobbyist $24 a month for each person; Creator $35; Business $65	Pro $16 a month; Team $40 for each member (paid plans are not open yet)
Billed annually, per month	Hobbyist $16; Creator $24; Business $50; Enterprise on request	Pro $12; Team $32 for each member
Media each month on paid plans	Hobbyist 10 hours, Creator 30, Business 40	See below

For Dalm, the Free plan is a plan rather than a trial, and transcribing a file you bring in is part of it. No plan puts a limit on recording minutes, on the length of a recording, or on how long anything is kept. The pricing page sets out Pro and Team, and what each one adds, and how a team shares one dictionary and one set of templates has a page of its own. While Dalm is not public, the way in is the waitlist. Descript’s figures are its regular prices from its pricing page, checked on 2 October 2026.

Bringing a recording in: one voice or a whole conversation

In Dalm you drag a file onto the window or choose it with ⌘O, and say what it is. Any file macOS treats as audio or video will do, from a third of a second to four hours; for a video, only the sound is read. As a Dictation, it becomes one block of text in your dictations: the voices are not split, only you can see it, and no language model reads it. As a Meeting (several voices), the voices are told apart on your Mac, you can name them, and a summary is written from the transcript without being asked — that is when a language model reads the text. Either way the speech is recognised on your Mac, your file is left where it was, and the text syncs to our servers. Descript starts from the other end: its help says transcription starts once the file is in its cloud, and the transcript becomes the handle for editing it.

Editing an episode, or only needing the words

• Descript, if you edit podcasts or videos, record your screen or remote guests, work on Windows or in a browser, or want the recording kept with the project to cut again later.
• Dalm, if from an interview, an episode or a lecture you need the words and not the edit, you work on a Mac, and the audio should stay off other people’s computers; the text still syncs to our servers. And if files are only part of your day: Dalm also records calls and meetings in the room, and types what you dictate into any app. Everything you say lands in one archive you can search: meetings, phone calls, conversations in the room, and what you dictate to a model.
• Neither, if the audio must not leave your computer and you also need to edit it by its words: Descript’s help says the file is uploaded to the cloud for transcription, and Dalm keeps no recording to edit.

Other transcription tools, compared by name

Weighing more than two? Seven transcription tools side by side, each with where it beats us. If what you have is a podcast episode, getting its transcript, split by voice, step by step, and what a podcaster gets after an interview. Where each step of transcription runs, locally or in the cloud.