Naming your language instead of leaving it on Auto can improve accuracy, because the model stops spending effort guessing. It helps most on short recordings.
Each row only appears when your current model can use it. If you cannot find
Translate to English, the model you are on cannot translate. Auto is
only offered for a model that can work out the language itself.
Your choice is kept when you switch models
The language you pick is saved as your choice, and switching models never changes it. The picker shows what the current model will actually use for it:- The language you picked, when the model hears it.
- Otherwise Auto, when the model can detect the language.
- Otherwise English, or the first language the model lists if it has no English.
Which model for which languages
Parakeet is English only. Nemotron covers 28 languages with automatic detection. Whisper Medium is the widest at 99.
To find every model for your language, use the language filter on the Speech to text tab. See Transcription models.
Which models translate to English
Translation is a separate task from recognition. Only some models support it, and the Translate filter on the Speech to text tab lists them.
Whisper Large v3 Turbo catches people out. It is a Whisper model, but it does not support translation.
Leave translation off and you get a transcript in the language you spoke. Turn it on and you get English. If you are already speaking English, the model transcribes as usual.
Cloud translation
Cloud providers that use the OpenAI multipart format (OpenAI, Groq, and any OpenAI-compatible server you set up yourself) support translation to English through a separate/audio/translations endpoint. Other cloud providers do not.
If translation is on but the selected cloud provider does not support it, SpeakoFlow transcribes normally and logs a warning. The setting row in the cloud section says so in place of a switch that would do nothing.
Simplified and Traditional Chinese
There are two ways to get the script you write in. Pick it as the language. The language list has both Simplified Chinese and Traditional Chinese. The model is toldzh either way, and SpeakoFlow converts every transcript into the script you picked afterwards, with OpenCC, on your computer.
Use Chinese script. With the language on Auto or Cantonese, the Chinese script row picks Simplified, Traditional, or As transcribed (the default, which leaves the text as the model wrote it). It converts Chinese only. On Auto, a dictation that comes out in English or Japanese is left alone, so you can switch languages freely and only your Chinese is converted. Traditional Cantonese follows Hong Kong’s character conventions, Traditional Mandarin Taiwan’s.
The row only appears where it can change something: for a model that can write Chinese, and in the cloud transcription settings, since every cloud provider can answer in Chinese. It is hidden while Simplified or Traditional Chinese is picked as the language, because that choice already decides the script.
Language as a hint for cloud providers
Language as a hint for cloud providers
The language setting is forwarded to cloud providers as a recognition hint:
language_code on ElevenLabs, language on the OpenAI schema and Deepgram.
The transcript still comes back in the language spoken. Script subtags like
zh-Hans and zh-Hant are stripped on the way out, because no cloud endpoint
here understands them. The Simplified or Traditional conversion still happens
on your computer.Custom words work differently per model
Custom words work differently per model
Whisper models receive your custom words as a recognition prompt, and models that accept a vocabulary receive them as vocabulary. Both bias the model before it decides. Every other model gets a fuzzy correction pass over the finished transcript instead.Both work. Biasing the model tends to be more accurate because nothing has to be guessed after the fact. See Custom words.
Transcription models
The full catalog and how to choose.
Cloud transcription
Using a cloud service instead of a local model.