Speech-to-Text
Transcribe (REST)
`SpeechToText.Transcribe` for file-based speech recognition.
import "github.com/Shreehari-Acharya/sarvamai-go/stt"
func (c *STTClient) Transcribe(ctx context.Context, file io.Reader, opts ...Option) (*TranscribeResponse, error)
file (io.Reader) is required.
| Option | Type | Notes |
|---|
WithModel | ModelSaarika | ModelSaaras | If omitted, model is not explicitly sent by SDK. |
WithMode | ModeTranscribe | ModeTranslate | ModeVerbatim | ModeTranslit | ModeCodemix | Mode support depends on model validation rules. |
WithLanguage | languages.Code | Allowed values depend on selected model (or validation default when omitted). |
WithAudioCodec | speech.InputAudioCodec | SDK does not locally validate codec values for REST transcribe. |
| Combo | Result |
|---|
WithModel(ModelSaaras) + WithMode(...) | Valid |
WithModel(ModelSaarika) + WithMode(...) | Validation error (mode only supported with saaras:v3) |
WithMode(...) with model omitted | Valid in SDK validation (mode check uses saaras:v3 spec) |
WithModel(ModelSaarika) + WithLanguage in SaarikaLanguages | Valid |
WithModel(ModelSaarika) + WithLanguage outside SaarikaLanguages | Validation error |
WithModel(ModelSaaras) + WithLanguage in SaarasLanguages | Valid |
WithModel(ModelSaaras) + WithLanguage outside SaarasLanguages | Validation error |
WithLanguage with model omitted | Valid only when language is in SaarasLanguages |
file must be non-nil.- If
model is set, it must be saarika:v2.5 or saaras:v3.
- Mode is validated via
speech.ValidateMode.
- Language is validated via
speech.ValidateLanguageWithSpec.
| Field | Type | Notes |
|---|
Transcript | string | Main transcription output. |
RequestID | *string | Optional request identifier. |
LanguageCode | *string | Optional detected/selected language code. |
LanguageProbability | *float64 | Optional confidence for language detection. |
Timestamps | *speech.Timestamps | Optional word timing metadata. |
DiarizedTranscript | *speech.DiarizedTranscript | Optional diarized output when available. |
resp, err := client.SpeechToText.Transcribe(
ctx,
audioFile,
stt.WithModel(stt.ModelSaaras),
stt.WithMode(stt.ModeTranslate),
stt.WithLanguage(stt.LanguageHiIN),
stt.WithAudioCodec(stt.CodecWAV),
)
if err != nil {
panic(err)
}
fmt.Println(resp.Transcript)