Merge pull request #26 from deepgram-devs/adding-toSRT-and-WebVTT

Adding to srt and web vtt

deepgram/deepgram.go

@@ -5,7 +5,7 @@ import (
 	"net/http"
 )
 
-var sdkVersion string = "0.4.0"
+var sdkVersion string = "0.5.0"
 var dgAgent string = "deepgram-go-sdk/v" + sdkVersion
 
 type Client struct {

deepgram/prerecorded.go

@@ -1,58 +1,58 @@
 package deepgram
 
 type Metadata struct {
-	RequestId string `json:"request_id"`
-	TransactionKey string `json:"transaction_key"`
-	Sha256 string `json:"sha256"`
-	Created string `json:"created"`
-	Duration int `json:"duration"`
-	Channels int `json:"channels"`
+	RequestId      string  `json:"request_id"`
+	TransactionKey string  `json:"transaction_key"`
+	Sha256         string  `json:"sha256"`
+	Created        string  `json:"created"`
+	Duration       float64 `json:"duration"`
+	Channels       int     `json:"channels"`
 }
 
 type Hit struct {
 	Confidence float64 `json:"confidence"`
-	Start float64 `json:"start"`
-	End float64 `json:"end"`
-	Snippet string `json:"snippet"`
+	Start      float64 `json:"start"`
+	End        float64 `json:"end"`
+	Snippet    string  `json:"snippet"`
 }
 
 type Search struct {
 	Query string `json:"query"`
-	Hits []Hit `json:"hits"`
+	Hits  []Hit  `json:"hits"`
 }
 
 type WordBase struct {
-	Word string `json:"word"`
-	Start float64 `json:"start"`
-	End float64 `json:"end"`
-	Confidence float64 `json:"confidence"`
-	Punctuated_Word string `json:"punctuated_word"`
-	Speaker int `json:"speaker"`
+	Word            string  `json:"word"`
+	Start           float64 `json:"start"`
+	End             float64 `json:"end"`
+	Confidence      float64 `json:"confidence"`
+	Punctuated_Word string  `json:"punctuated_word"`
+	Speaker         int     `json:"speaker"`
 }
 
 type Alternative struct {
-	Transcript string `json:"transcript"`
-	Confidence float64 `json:"confidence"`
-	Words []WordBase `json:"words"`
+	Transcript string     `json:"transcript"`
+	Confidence float64    `json:"confidence"`
+	Words      []WordBase `json:"words"`
 }
 
 type Channel struct {
-	Search []Search `json:"search"`
+	Search       []Search      `json:"search"`
 	Alternatives []Alternative `json:"alternatives"`
 }
 
 type Utterance struct {
-	Start float64 `json:"start"`
-	End float64 `json:"end"`
-	Confidence float64 `json:"confidence"`
-	Channel int `json:"channel"`
-	Transcript string `json:"transcript"`
-	Words []WordBase `json:"words"`
-	Speaker int `json:"speaker"`
-	Id string `json:"id"`
+	Start      float64    `json:"start"`
+	End        float64    `json:"end"`
+	Confidence float64    `json:"confidence"`
+	Channel    int        `json:"channel"`
+	Transcript string     `json:"transcript"`
+	Words      []WordBase `json:"words"`
+	Speaker    int        `json:"speaker"`
+	Id         string     `json:"id"`
 }
 
 type Results struct {
 	Utterances []Utterance `json:"utterances"`
-	Channels []Channel `json:"channels"`
-}
+	Channels   []Channel   `json:"channels"`
+}

deepgram/transcriptions.go

@@ -3,20 +3,22 @@ package deepgram
 import (
 	"bytes"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"log"
 	"net/http"
 	"net/url"
+	"strings"
 
 	"github.com/google/go-querystring/query"
 	"github.com/gorilla/websocket"
 )
 
 type PreRecordedResponse struct {
-	Request_id string      `json:"request_id"`
-	Metadata   interface{} `json:"metadata"`
-	Results    interface{} `json:"results"`
+	Request_id string   `json:"request_id"`
+	Metadata   Metadata `json:"metadata"`
+	Results    Results  `json:"results"`
 }
 
 type LiveTranscriptionOptions struct {
@@ -85,7 +87,6 @@ func (dg *Client) LiveTranscription(options LiveTranscriptionOptions) (*websocke
 		log.Fatal("dial:", err)
 	}
 	return c, resp, nil
-
 }
 
 func (dg *Client) PreRecordedFromURL(source UrlSource, options PreRecordedTranscriptionOptions) (PreRecordedResponse, error) {
@@ -130,3 +131,46 @@ func (dg *Client) PreRecordedFromURL(source UrlSource, options PreRecordedTransc
 	}
 
 }
+
+func (resp *PreRecordedResponse) ToWebVTT() (string, error) {
+	if resp.Results.Utterances == nil {
+		return "", errors.New("This function requires a transcript that was generated with the utterances feature.")
+	}
+
+	vtt := "WEBVTT\n\n"
+
+	vtt += "NOTE\nTranscription provided by Deepgram\nRequest ID: " + resp.Request_id + "\nCreated: " + resp.Metadata.Created + "\n\n"
+
+	for i, utterance := range resp.Results.Utterances {
+		utterance := utterance
+		start := SecondsToTimestamp(utterance.Start)
+		end := SecondsToTimestamp(utterance.End)
+		vtt += fmt.Sprintf("%d\n%s --> %s\n%s\n\n", i+1, start, end, utterance.Transcript)
+	}
+	return vtt, nil
+}
+
+func (resp *PreRecordedResponse) ToSRT() (string, error) {
+	if resp.Results.Utterances == nil {
+		return "", errors.New("This function requires a transcript that was generated with the utterances feature.")
+	}
+
+	srt := ""
+
+	for i, utterance := range resp.Results.Utterances {
+		utterance := utterance
+		start := SecondsToTimestamp(utterance.Start)
+		end := SecondsToTimestamp(utterance.End)
+		end = strings.ReplaceAll(end, ".", ",")
+		srt += fmt.Sprintf("%d\n%s --> %s\n%s\n\n", i+1, start, end, utterance.Transcript)
+
+	}
+	return srt, nil
+}
+
+func SecondsToTimestamp(seconds float64) string {
+	hours := int(seconds / 3600)
+	minutes := int((seconds - float64(hours*3600)) / 60)
+	seconds = seconds - float64(hours*3600) - float64(minutes*60)
+	return fmt.Sprintf("%02d:%02d:%02.3f", hours, minutes, seconds)
+}

examples/preRecordedFromUrl_example.go

@@ -0,0 +1,43 @@
+package main
+
+import (
+	"fmt"
+	"log"
+	"os"
+
+	"github.com/deepgram-devs/go-sdk/deepgram"
+)
+
+func main() {
+	dg := deepgram.NewClient("YOUR_API_KEY")
+	// Feel free to use this as your audio url to test https://anchor.fm/s/3e9db190/podcast/play/22624519/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2020-10-15%2F128822202-44100-1-79cab5de0d7af3c9.mp3
+	res, err := dg.PreRecordedFromURL(deepgram.UrlSource{Url: "AUDIO_FILE_URL"}, deepgram.PreRecordedTranscriptionOptions{Punctuate: true, Diarize: true, Language: "en-US", Utterances: true})
+	if err != nil {
+		fmt.Println("ERROR", err)
+		return
+	}
+	// Log the results
+	log.Printf("recv: %+v", res.Results)
+	f, err := os.Create("transcription.vtt")
+	if err != nil {
+		fmt.Printf("error creating VTT file: %v", err)
+	}
+	// Convert the results to WebVTT format
+	vtt, err := res.ToWebVTT()
+	if err != nil {
+		log.Fatal(err)
+	}
+	f.WriteString(vtt)
+
+	// Convert the results to SRT format
+	srtF, err := os.Create("transcription.srt")
+	if err != nil {
+		fmt.Printf("error creating SRT file: %v", err)
+	}
+	srt, err := res.ToSRT()
+	if err != nil {
+		log.Fatal(err)
+	}
+	srtF.WriteString(srt)
+
+}