Skip to content
KaveoKaveo

Translating subtitles

What it is for

Section titled “What it is for”

Translate a subtitle with an AI model, online or on your own computer. Each line keeps its timing.

Settings › Subtitles › Translation set up for Ollama (local): Provider, Endpoint, Model with translategemma and 3 models found, Translate into on Italiano, Read the original audio, Re-read the translation, Already translated set to Ask, and an empty API key that this server does not need, numbered 1 to 4

  1. Provider.
  2. API key.
  3. Model.
  4. Translate into.

How to use it

Section titled “How to use it”

Set up a provider

Section titled “Set up a provider”
  1. In Settings › Subtitles › Translation, choose a Provider, or Other… and type its Endpoint.
  2. Paste the API key if it needs one, and pick or type a Model.
  3. Choose a language in Translate into, such as Italian, and press Save. If your computer has no password store, a line under the key says it is kept only until you restart.

Translate

Section titled “Translate”

The progress panel over the player, reading Translating to Italian — 30 of 60 lines, with a half-filled bar, a Cancel button, and under them Block 2 of 2, how long the reply in flight has been waited for, and how long the last one took

  1. With a subtitle file on screen, right-click the picture and open Subtitles › Tools.
  2. Choose Translate to Italian…, named after your Translate into language, or Translate while I watch to start where the film is.

Good to know

Section titled “Good to know”
  • A track inside the video has to be extracted first — see Loading subtitles. Which model to pick: Choosing a translation model.
  • A panel at the top of the player shows the progress, with Cancel; the result is added under Beside the video.
  • Lines that still fail after Kaveo retries are left in their original language.
  • While the translation queue works, the player offers to pause it.
  • Turn on Re-read the translation for a second pass that corrects mistranslated lines. It helps with a large model; with a small one on your own computer it can add as many mistakes as it fixes.