Translation

Speakly deals with two languages: the one you speak and the one it types. They are separate settings, and both matter — the transcriber has to be told which language the audio is in, because guessing costs it accuracy, while translating into the language you want happens afterwards.

Where the output language is set

1. Open Settings → General.
2. Under Languages, the current choice is shown as Output language: English, with the line "Speak in any language, get transcription in English" beneath it.
3. Click Change.
4. In the Output Language dialog, use Search languages... and click the language you want.

The language you speak

It sits on the left of the same dialog, under You speak. Speakly fills it in from your system language the first time it runs, so most people never touch it. There is an Auto-detect option, but a specific language is more reliable: detection struggles when your speech mixes languages — Portuguese sentences full of English technical terms, for example. When the two languages differ, Speakly transcribes what you actually said and then translates it.

How many languages

99 — the complete set of languages OpenAI Whisper supports. Every one of them is selectable in the Output Language dialog.

Custom prompts have their own language

A custom prompt carries its own Target Language field (sidebar → Custom Prompts → New Prompt). The transcription itself still comes out in the language from Settings → General → Languages; then, while that prompt is active, post-processing is told to write the result in the prompt's language, translating if needed. So if the two differ, the active prompt's language is what you get in the end. That is how a prompt can act as a dedicated translator for one app.

Where else the language matters

• Style examples are generated in your output language, so the hint sent to the transcription model matches the language you write in.
• Call Translation has a separate target language of its own — see Call Translation.