Studio turns long scripts (chapters, articles, podcast scripts) into narrated audio. Izwi splits a script into paragraphs that you can edit, listen to and re-render one at a time, then export as one file. Open Studio from the sidebar. Each project opens at /studio/:projectId.

Create a project

  1. Select New project (or press ⌘N / Ctrl+N).
  2. Paste your script, or select Import a file for a .txt, .md or .html file.
  3. Choose a model and a voice. Select the play button next to a voice to hear it first.
  4. Select Create project.
If the model isn’t downloaded yet, Studio offers to download it. Paragraphs render once it is on this device.

Edit the script

A project reads like a document: one block per paragraph. Shift-click or ⌘-click (Ctrl-click) a paragraph’s play button to select it. With paragraphs selected, a bar offers Render, Export and Delete.

Render

Rendering runs on the Izwi server, one paragraph at a time. It keeps going when you leave Studio, and picks up where it stopped after the server restarts. The button at the top right always shows the next step:
  • Render 12 paragraphs for a new project.
  • Render 3 changes after edits: only paragraphs whose audio is missing, out of date or failed.
  • Rendering 4 of 12, with a stop button. Stopping finishes the paragraph in progress and leaves the rest.
  • Export when every paragraph is current.
Each paragraph shows where it stands: Select a paragraph’s render button to render just that one, or Render again in its ⋯ menu.

Listen

The player at the bottom plays the whole project, paragraph after paragraph, and highlights the one playing. Use it to skip between paragraphs, scrub through the project and change the playback speed.

Voice, speed and pronunciation

Select the voice button at the top of a project to open its settings:
  • Voice: the model and voice every paragraph uses.
  • Speed: for models that support it.
  • Pronunciation: replace a word with how it should be said, for names, acronyms or product terms. Rules match whole words in any letter case.
Saving a new voice or speed marks the affected paragraphs as changed, so you choose when to render them again. To give one paragraph a different voice (a second speaker, say), open its ⋯ menu and select Voice for this paragraph. Choose Project voice to return it to the project’s voice.

Export

Select Export to save the project as one WAV file.
  • Pause between paragraphs adds a short silence at each break.
  • Include script saves the text as a .txt file next to the audio.
If any paragraph isn’t rendered or has changed, the button reads Render and export: Studio renders those paragraphs first, then saves the file. Paragraphs rendered with different models are matched to one sample rate.

Using the API

Studio’s preview API is documented in the API Reference.

See Also