Help & answers

A 1080p NDI source with true alpha transparency — your switcher composites it like any lower-third graphic. No NDI gear? Use the chroma-key HDMI output instead. There is no burn-in and no browser overlay.

Translated captions reach the venue screen in about 1.7 seconds (under 2.5 s at the 95th percentile), measured from audio in to caption out — around 0.7 s with translation off.

Transcription and translation across 58 languages, with an auto-detect source option and code-switching (multiple languages in one talk). Set any source › target pair per session. Full list: English · Spanish · French · German · Italian · Portuguese · Dutch · Russian · Japanese · Korean · Chinese · Polish · Turkish · Ukrainian · Swedish · Danish · Norwegian · Romanian · Czech · Greek · Finnish · Indonesian · Bulgarian · Slovak · Arabic · Belarusian · Bangla · Bosnian · Catalan · Croatian · Estonian · Gujarati · Hebrew · Hindi · Hungarian · Kannada · Latvian · Lithuanian · Macedonian · Malay · Marathi · Persian · Serbian · Slovenian · Filipino · Tamil · Telugu · Thai · Urdu · Vietnamese · Afrikaans · Georgian · Armenian · Punjabi · Nepali · Traditional Chinese · Basque · Galician.

OpenAI reads context and idiom the way a person would, so captions sound natural on fast, informal live speech — not stiff, word-for-word output. If OpenAI is ever slow or errors, SCAPTION automatically switches to a second European provider mid-session — no operator action, no gap in the captions. Each caption line is translated on its own, with no shared history between sessions, so nothing from one event can leak into another’s translation.

The caption engine and NDI output run locally on your machine. The internet only carries audio and text to the relay — your show isn’t hostage to a browser tab. If the connection drops, the audience output fades to transparent, never a frozen caption or error card.

Nothing on screen. Hitting a limit warns and auto-bills overage — a live session is never interrupted. Entitlement is only re-checked between events.

The usual cause is Windows’ “Public” network profile, which blocks inbound NDI mDNS discovery for every app on the machine. SCAPTION detects this and warns you, and always shows its own source name + ip:port so you can add it manually in vMix / TriCaster.

One hour of full-featured captioning and translation, no expiry, no card required — enough to prove it on your own rig before you pay.

Yes. SCAPTION reads ASIO drivers directly — up to four discrete channels from one console, which is what multi-mic events need, since a WDM endpoint only ever gives you a stereo pair. WDM works too, so your interface shows up either way, and almost every one exposes a WDM input (Behringer, Yamaha, Allen & Heath, Focusrite). For Dante, run Dante Virtual Soundcard in WDM mode. If a device is ASIO-only, bridge it with VB-Audio Voicemeeter or ASIO Link Pro. One thing to check at load-in: some interfaces close their WDM side while a DAW holds the ASIO driver exclusively — route through a virtual device if you need both at once. Advanced mode needs one distinct WDM device per microphone, so a multi-channel interface that only exposes stereo pairs may need a virtual router to split it.