Dictation and Subtitles in DepCut and Where Each Runs

Speech to text appears in two places in DepCut, and each one runs somewhere different. Knowing which is which explains why dictation can be free and instant on one setup, and why subtitles show up in the background and use credits.

Two jobs, two places

Dictation is the Dictate microphone button in prompt boxes: the AI chat, and the prompts for image, video and audio generation. You speak, and the words land in the box.

Subtitles are the captions you create in the Subtitles tab with Generate subtitles. They turn the speech in your cut into timed captions on the timeline.

Dictation with the Mac app running

When the DepCut Mac app is running, dictation runs on your Mac. Your browser captures the microphone and streams the audio to the app, and the transcript builds live in the box as you talk. This holds whichever home the project has, and it costs nothing.

While you dictate, the box shows a level meter and your words so far. Press the check button, titled Use transcription, or press Enter to add the text after anything already typed. Press the x button, titled Cancel, or Escape to throw the take away.

Dictation without the app

With no app running, DepCut records your take in the browser. When you confirm, it sends the recording in one piece for hosted transcription. There are no live words while you speak, so the box shows Listening… and the text arrives after you confirm.

Either way, your browser must have microphone permission for the site. If it is blocked, the editor says to allow the microphone for this site and try again.

Subtitles, always hosted

Subtitles use a hosted speech model, with or without the Mac app, and they spend account credits. This is why the result is the same on every machine. The process:

1. In the Subtitles tab, pick a language. Detect language leaves the choice to the model, and the list also names English (US), English (UK), Español, Français, Deutsch, Italiano, Português (BR), 日本語, 한국어, 中文 and Tiếng Việt.
2. Press Generate subtitles.
3. Your browser mixes the audio of the cut the way it will sound, with your trims, speeds and volumes applied, and sends it in short chunks. Silent chunks are skipped.
4. The model returns text with timing, and DepCut places the captions on the timeline.

A line reading Transcribing… with a clock runs while this happens. It continues in the background, so keep editing. If the cut contains no speech, the panel says "No speech was found in this cut, so no subtitles were added to the video" and the button changes to Try again.

Check the timing

A transcription model can place a caption slightly early or late. After the transcript arrives, DepCut compares caption times with the speech envelope of your audio and moves an edge only when the audio shows where the speech starts or stops. A caption boundary in the middle of a sentence keeps the time the model gave it.

Cut your video first and generate subtitles last, so the captions are built from the audio you will export. Then read the transcript and fix any misheard words.

Compare the two

Open a project in DepCut, generate subtitles for a clip with speech, and try the Dictate button in the AI chat to compare the two.