Waveform Studio

Waveform Studio is available in two forms: a cloud web app and a downloadable desktop app. Both offer the same core editing experience, but they differ in how they are accessed, how projects are stored, and what features are available.

Waveform Studio is open source. The desktop app exists so that anyone can use the full editing environment for free, without a cloud subscription or an internet connection. Rather than running in a browser, the desktop app bundles an internal server that runs locally on your machine.

The desktop app is currently available on Windows only (Windows 10 and Windows 11, 64-bit).

Automatic transcription is available in the cloud web app, where it is handled server-side using Whisper and Pyannote. It is not yet available in the desktop app. To work with a transcript in the desktop app you will need to generate one using an external tool and import it as a CSV file with start, end, speaker, and text columns.

Waveform Studio was designed around Whisper (speech-to-text) and Pyannote (speaker diarization), both of which are free and open source and produce output compatible with the import format. Other transcription tools can also be used as long as their output is converted to the expected CSV structure.

Waveform Studio is built around the concept of a project — a folder that contains your audio file, transcript data, and any project settings. Follow these steps to create and load your first project.

An account is required to use the Waveform Studio cloud web app. The desktop app does not require one.

Waveform Studio is currently in closed beta. Account creation is not open to the public — new accounts require administrator approval before they can be used.

With an account you can:

All accounts start on the Free tier. Paid tiers unlock higher usage limits, more storage, and access to larger Whisper models. Billing can be monthly or yearly; yearly billing includes a 20% discount.

Whisper model availability also varies by tier — larger, more accurate models are only available on higher tiers. See Transcription for model details. Your current tier and usage can be viewed and changed in the Subscription tab of Account Settings.

Open Account Settings by clicking the avatar icon in the top bar of the sidebar. Settings are divided into tabs.

There are a number of different terms used throughout the application, and this guide. This is a list of terms and their meanings.

The workspace is divided into three key areas. Each panel is resizable by dragging the dividers between them.

The sidebar on the left edge of the app lists your projects. When connected to a server, projects are stored in your account on that server. The sidebar can be collapsed with the toggle button (◀) on its right edge.

A row of utility buttons runs along the very top of the sidebar.

When signed in, the project list is divided into two sections:

The ↓ Save button in the title bar downloads the current project as a .wfs file. This can be used at any time to take a local copy of a project for backup or for use in the desktop app.

Server projects are saved automatically — a few seconds after each edit, changes are pushed to the server. This is the default and intended behaviour. The ↑ Upload button in the title bar is available as a manual trigger, useful as a confirmation that changes have been saved or when auto-save has been turned off. Auto-save can be disabled in Settings.

Server projects can be organized into folders. Folders are nested — you can create folders inside other folders — and the current location is shown as a breadcrumb trail at the top of the project list.

The waveform panel is the primary audio interface. It displays a zoomable waveform of the loaded audio file, lets you control playback, and shows speaker regions aligned to the transcript. Every component is described below.

The panel is divided into two areas stacked vertically:

Below the controls bar is the minimap — a full-width thumbnail of the entire waveform. The shaded rectangle shows which portion of the audio is currently visible. Drag the rectangle to pan, or drag its handles to resize the view.

Click anywhere on the waveform display or on a region in the region lane to move the playhead to that position. The transcript panel will scroll to and highlight the corresponding segment.

Use the ↩ 5s and 5s ↪ buttons in the controls bar to skip backward or forward by five seconds. The keyboard equivalents are ← and →.

Zooming in reveals more detail in the waveform and makes precise seeking easier.

The region lane is the coloured strip directly below the waveform display. It provides a timeline view of every transcript segment laid out to scale against the audio.

Each block in the lane is a region — it maps exactly to one transcript segment. The block's horizontal position and width represent that segment's start and end times, so longer blocks mean longer speech. Regions are colour-coded by speaker: the colour of each block matches the hue assigned to that speaker in the Speakers panel, making it easy to see at a glance who is speaking and when.

Regions are grouped by paragraph. All segments within the same paragraph are joined into a single continuous bar with rounded outer corners and the speaker's name drawn inside it. A small gap separates each paragraph from the next. A speaker block in the transcript may contain multiple paragraphs, so it is possible to see several bars of the same colour in a row — one per paragraph — before the colour changes for a different speaker.

When you zoom in on the waveform, the region lane zooms with it, keeping regions aligned to the audio beneath them. If no transcript has been loaded, the region lane is hidden.

The speakers panel lists every speaker in the project. Each speaker has a name, a colour, an internal ID, and an optional voice sample. Speakers are listed in the order they first appear in the transcript.

Click ＋ Add Speaker at the top of the panel to create a new speaker. A new row appears with an auto-generated ID (e.g. SPEAKER_04) and the name field is immediately opened for editing so you can type a name right away.

Click a speaker's name to enter inline edit mode. Type the new name, then press Enter or click away to confirm. Press Esc to cancel and restore the previous name. The new name is reflected everywhere in the transcript immediately.

If you clear the name field and confirm, the name resets to the speaker's raw ID.

Click the colour swatch to the left of a speaker's name to open the hue picker. Drag the handle around the colour wheel to choose a new hue. The change is applied live — waveform regions, region lane bars, and transcript speaker labels all update instantly.

Hovering over a speaker's name highlights all of that speaker's regions in the waveform panel, making it easy to see at a glance how much of the audio they occupy and where their turns fall.

Each speaker row has a voice sample cell. A voice sample is a short audio clip that represents a speaker's voice — useful for confirming identity when labelling speakers. When no sample has been set, three buttons are shown for adding one:

Once a sample is set, the cell shows a mini waveform visualisation drawn in the speaker's colour. Click ▶ to play the sample — click it again to stop early. Click the ✕ button to remove the sample.

Click the ✕ button at the right of a speaker row to delete that speaker. A confirmation dialog appears. If the speaker has segments assigned to them, the dialog will ask you to choose another speaker to reassign those segments to before deletion proceeds. If the speaker has no assigned segments, you can delete them directly.

The transcript panel displays the full diarized transcript as readable text, grouped by speaker. It is the primary interface for reviewing, navigating, and editing the content of a transcript.

The transcript is organised into three levels:

The coloured bar on the left edge of each paragraph is its handle. Its colour matches the speaker. Interacting with the handle affects the entire paragraph:

While audio is playing, the segment corresponding to the current playhead position is highlighted and the transcript scrolls automatically to keep it in view.

A search bar appears at the top of the transcript panel once a transcript is loaded. Type to search — matching segments are highlighted and a match counter shows how many results were found. Use the ◀ and ▶ arrows to step through matches, or press Enter to advance to the next one. Press Esc or click the clear button to dismiss the search.

Use the speaker filter dropdown to restrict results to a specific speaker.

Click the replace toggle button to expand a replace bar beneath the search field. Enter a replacement term and use Replace to substitute the current match, or Replace All to substitute every match at once. Press Tab to move between the search and replace fields.

You can select text freely within the transcript using click-and-drag. Selections can span multiple segments. Hold Ctrl while dragging to snap the selection boundaries to whole words.

Once text is selected, two actions are available:

Any span of text in the transcript can have a hyperlink attached to it. Linked text is displayed with a distinct style so it stands out from the surrounding content.

Select the text you want to link, then press Shift+K or right-click and choose Add link. If no text is selected, the link covers the entire segment. To edit an existing link, select or click on the linked text and use the same shortcut — the dialog opens pre-filled with the existing values.

The link dialog contains four fields:

When a new link is added over a range that already contains one or more links, conflicts are resolved automatically:

Hovering over linked text shows a tooltip with the link's display name, description, URL, and any editor notes (in edit mode). The tooltip appears after a short delay, or immediately if Ctrl is already held.

To follow a link, hold Ctrl and click the linked text. The URL opens in a new tab.

Waveform Studio can automatically transcribe an audio file and generate a speaker-diarized transcript using Whisper (speech-to-text) and Pyannote (speaker diarization). Transcription runs on cloud GPU infrastructure and is available in the web app only — see Cloud vs Desktop for details.

The dialog displays an estimated cost and processing time based on your audio duration and chosen models. These are calculated assuming an A10G GPU and will update live as you change the model selections. Actual cost and time may vary.

The transcription pipeline code is included in the project source. If you are self-hosting Waveform Studio or want to run transcription jobs independently, you can adapt or invoke this code directly rather than using the hosted cloud service.

See the developer documentation for a full walkthrough: Transcription Pipeline Tutorial.

Presentation mode is a read-only, shareable view of a project designed for playback and transcript review. It strips away all editing controls and presents the audio player and transcript in a clean, scrollable layout that anyone can use — no account required, depending on your sharing settings.

Presentation mode is available for projects in the web app only.

Click the Present button in the workspace header. This opens the presentation view in a new browser tab at a unique URL for your project. You can copy and share this URL with others using the Copy link button inside the presentation view.

The presentation view has three main areas:

The transcript is interactive even in read-only mode:

Whether viewers need to sign in depends on the project's sharing setting:

A Live Quote is an interactive audio-and-text widget that you can embed on any webpage. It shows a highlighted excerpt from your transcript alongside a playable audio clip — viewers can read the words, click any segment to seek, and play or pause the clip directly in the page, with no account or sign-in required.

Live Quotes are available in the web app only. The source project must be saved to the server.

Select any text in the transcript panel, then open the Live Quote dialog using one of two methods:

The dialog has two panels: a configuration panel on the left and a live preview of the final widget on the right.

The left panel lists every transcript segment that falls within your selection. Each segment has a checkbox. Unchecking a segment removes its text from the displayed quote and replaces it with an ellipsis (…) — but the underlying audio still plays through uninterrupted. This lets you condense a quote without creating jarring audio gaps.

When the first or last segment in your selection is only partially selected, a trim checkbox appears at that boundary:

Live Quote clips are capped at 60 seconds. A duration bar below the segment list shows the current clip length. As you approach the limit the bar turns amber; if you exceed it the bar turns red and the Create Embed Code button is disabled until you reduce the selection.

The right panel shows an exact preview of the widget as it will appear on the page. It updates in real time as you check or uncheck segments and toggle options. You can play the audio clip directly in the preview — clicking any segment seeks to that point, and clicking the waveform or progress bar seeks to that position in the clip.

Click Create Embed Code to submit the configuration to the server. The server processes the audio clip and stores the widget data, then returns an <iframe> snippet. The right panel switches to show the snippet in a read-only text area with a Copy to clipboard button.

Paste the snippet into any webpage — a blog post, CMS, or plain HTML file — and the widget will render at 660 × 280 px, scaling down to fit narrower containers automatically.

The embedded widget is fully self-contained and interactive. It includes:

When signed in to a Waveform Studio server, your projects are stored in your account and synced automatically. You also gain access to server-side transcription powered by Whisper and Pyannote.

Waveform Studio is designed for keyboard-driven workflows. The full shortcut reference is also accessible via the legend at the bottom of the sidebar.

A complete reference for every interactive element in Waveform Studio, organized by area of the interface.

Server features are unlocked by signing in. See Server Integration for the full workflow.

The Ⓘ icon appears next to certain settings and labels throughout the interface. Hovering over it shows a short tooltip explaining the adjacent control. Clicking the icon opens a relevant page in this documentation in a new tab.