Trans live English → Português captionsconnecting…

Microphone

Speak in English; each pause flushes a caption. First start loads the models (a few seconds).

Speech detection

Sentence pause (debounce) — how long a pause ends the sentence Max sentence length — force-translate a run-on sentence Realtime update rate — how often live captions refresh Realtime captions (show words while still speaking)

Saved automatically and applied live — no restart needed.

Speech recognition engine

Bigger = catches more obscure words. large-v3 is the most accurate; large-v3-turbo is nearly as good and much faster (recommended with an NVIDIA GPU). Models not yet downloaded are fetched on first switch — the big ones take a few minutes.

Custom vocabulary — names, spells, slang the recognizer should expect

Whisper biases toward these words, so rare terms come out spelled right instead of guessed. Saved automatically.

Overlay designer

Style Languages shown Position Scale — Time on screen Text size — Max width — Font Text color Accent (EN line + border) Background Background opacity — Border width — Corner radius — ALL CAPS Text shadow

Drag the overlay in the preview to place it anywhere — the URL updates with the exact spot. In OBS: + → Browser, paste the URL, size 1920×1080. Transparent background — it composites over the scene.

Trans live English → Português captionsconnecting…

Microphone

Speech detection

Speech recognition engine

Overlay designer

Test without the mic

Live feed