101 lines
2.2 KiB
Go
101 lines
2.2 KiB
Go
package speech
|
|
|
|
import (
|
|
"context"
|
|
|
|
"github.com/flowy-live/llink/internal/utils/flog"
|
|
|
|
dgapi "github.com/deepgram/deepgram-go-sdk/v3/pkg/api/listen/v1/rest"
|
|
interfaces "github.com/deepgram/deepgram-go-sdk/v3/pkg/client/interfaces"
|
|
client "github.com/deepgram/deepgram-go-sdk/v3/pkg/client/listen"
|
|
)
|
|
|
|
type TranscriptWord struct {
|
|
Word string
|
|
Start float64
|
|
End float64
|
|
}
|
|
|
|
type TranscriptSentence struct {
|
|
Text string
|
|
Start float64
|
|
End float64
|
|
}
|
|
|
|
type TranscriptParagraph struct {
|
|
Sentences []TranscriptSentence
|
|
Start float64
|
|
End float64
|
|
}
|
|
|
|
type TranscriptResult struct {
|
|
Transcript string
|
|
Words []TranscriptWord
|
|
Paragraphs []TranscriptParagraph
|
|
}
|
|
|
|
type SpeechService interface {
|
|
Transcribe(ctx context.Context, mediaUrl string) (*TranscriptResult, error)
|
|
}
|
|
|
|
type speechServiceImpl struct {
|
|
deepgramClient *dgapi.Client
|
|
}
|
|
|
|
func NewSpeechService(ctx context.Context) SpeechService {
|
|
return &speechServiceImpl{
|
|
deepgramClient: dgapi.New(client.NewRESTWithDefaults()),
|
|
}
|
|
}
|
|
|
|
func (s *speechServiceImpl) Transcribe(ctx context.Context, mediaUrl string) (*TranscriptResult, error) {
|
|
options := &interfaces.PreRecordedTranscriptionOptions{
|
|
Model: "nova-3",
|
|
SmartFormat: true,
|
|
Paragraphs: true,
|
|
}
|
|
|
|
response, err := s.deepgramClient.FromURL(ctx, mediaUrl, options)
|
|
if err != nil {
|
|
flog.Error("failed to transcribe prerecorded media", "error", err)
|
|
return nil, err
|
|
}
|
|
|
|
alt := response.Results.Channels[0].Alternatives[0]
|
|
|
|
words := make([]TranscriptWord, len(alt.Words))
|
|
for i, w := range alt.Words {
|
|
words[i] = TranscriptWord{
|
|
Word: w.PunctuatedWord,
|
|
Start: w.Start,
|
|
End: w.End,
|
|
}
|
|
}
|
|
|
|
var paragraphs []TranscriptParagraph
|
|
if alt.Paragraphs != nil {
|
|
paragraphs = make([]TranscriptParagraph, len(alt.Paragraphs.Paragraphs))
|
|
for i, p := range alt.Paragraphs.Paragraphs {
|
|
sentences := make([]TranscriptSentence, len(p.Sentences))
|
|
for j, s := range p.Sentences {
|
|
sentences[j] = TranscriptSentence{
|
|
Text: s.Text,
|
|
Start: s.Start,
|
|
End: s.End,
|
|
}
|
|
}
|
|
paragraphs[i] = TranscriptParagraph{
|
|
Sentences: sentences,
|
|
Start: p.Start,
|
|
End: p.End,
|
|
}
|
|
}
|
|
}
|
|
|
|
return &TranscriptResult{
|
|
Transcript: alt.Transcript,
|
|
Words: words,
|
|
Paragraphs: paragraphs,
|
|
}, nil
|
|
}
|