feat - add --srt flag (#87)

* feat - add --srt flag

* feat - add srt to readme

* feat - add srt to readme

* feat - add srt to readme

* fix - improve error message

Author: fcastillo90

.gitignore

@@ -3,4 +3,5 @@ testfile.*
 !.github/*
 brew/*
 tmp-*
-vendor
+vendor
+*.srt

README.md

@@ -211,6 +211,11 @@ assemblyai transcribe [local file | remote url | youtube links] [--flags]
 > example: `--custom_spelling "[{\"from\": [\"ariana\"], \"to\": \"Arianna\"}]"` or  `--custom_spelling ./custom_spelling.json`
 > Specify how words are spelled or formatted in the transcript text.
 
+> **--srt**  
+> default: false  
+> example: `--srt`  
+> Create an SRT file named `[id].srt` in the current directory.
+
 </details>
 
 ### Get
@@ -234,6 +239,11 @@ assemblyai get [id]
 > example: `--poll=false`  
 > The CLI will poll the transcription every 3 seconds until it's complete.
 
+> **--srt**  
+> default: false  
+> example: `--srt`  
+> Create an SRT file named `[id].srt` in the current directory.
+
 </details>
 
 ## Contributing
@@ -242,7 +252,7 @@ We're more than happy to welcome new contributors. If there's something you'd li
 
 ## Telemetry
 
-The AssemblyAI CLI includes a telemetry feature that collects usage data, and is enabled by default.
+The AssemblyAI CLI includes a telemetry feature that collects usage data and is enabled by default.
 
 To opt out of telemetry, set the telemetry variable in the `config.toml` file to false.
 

cmd/get.go

@@ -31,6 +31,7 @@ var getCmd = &cobra.Command{
 		id := args[0]
 		flags.Poll, _ = cmd.Flags().GetBool("poll")
 		flags.Json, _ = cmd.Flags().GetBool("json")
+		flags.Srt, _ = cmd.Flags().GetBool("srt")
 
 		U.Token = U.GetStoredToken()
 		if U.Token == "" {
@@ -51,5 +52,6 @@ func init() {
 	getCmd.Flags().BoolP("json", "j", false, "If true, the CLI will output the JSON.")
 	getCmd.Flags().BoolP("poll", "p", true, "The CLI will poll the transcription until it's complete.")
 	getCmd.Flags().Bool("test", false, "Flag for test executing purpose")
+	getCmd.PersistentFlags().BoolP("srt", "", false, "Generate an SRT file for the audio file transcribed.")
 	getCmd.Flags().MarkHidden("test")
 }

cmd/transcribe.go

@@ -40,6 +40,7 @@ var transcribeCmd = &cobra.Command{
 
 		flags.Json, _ = cmd.Flags().GetBool("json")
 		flags.Poll, _ = cmd.Flags().GetBool("poll")
+		flags.Srt, _ = cmd.Flags().GetBool("srt")
 		params.AutoChapters, _ = cmd.Flags().GetBool("auto_chapters")
 		params.AutoHighlights, _ = cmd.Flags().GetBool("auto_highlights")
 		params.ContentModeration, _ = cmd.Flags().GetBool("content_moderation")
@@ -265,8 +266,8 @@ func init() {
 	transcribeCmd.PersistentFlags().BoolP("auto_chapters", "s", false, "A \"summary over time\" for the audio file transcribed.")
 	transcribeCmd.PersistentFlags().BoolP("auto_highlights", "a", false, "Automatically detect important phrases and words in the text.")
 	transcribeCmd.PersistentFlags().BoolP("content_moderation", "c", false, "Detect if sensitive content is spoken in the file.")
-	transcribeCmd.PersistentFlags().BoolP("dual_channel", "d", false, "Enable dual channel")
 	transcribeCmd.PersistentFlags().BoolP("disfluencies", "D", false, "Include Filler Words in your transcripts")
+	transcribeCmd.PersistentFlags().BoolP("dual_channel", "d", false, "Enable dual channel")
 	transcribeCmd.PersistentFlags().BoolP("entity_detection", "e", false, "Identify a wide range of entities that are spoken in the audio file.")
 	transcribeCmd.PersistentFlags().BoolP("format_text", "f", true, "Enable text formatting")
 	transcribeCmd.PersistentFlags().BoolP("json", "j", false, "If true, the CLI will output the JSON.")
@@ -276,6 +277,7 @@ func init() {
 	transcribeCmd.PersistentFlags().BoolP("redact_pii", "r", false, "Remove personally identifiable information from the transcription.")
 	transcribeCmd.PersistentFlags().BoolP("sentiment_analysis", "x", false, "Detect the sentiment of each sentence of speech spoken in the file.")
 	transcribeCmd.PersistentFlags().BoolP("speaker_labels", "l", true, "Automatically detect the number of speakers in your audio file, and each word in the transcription text can be associated with its speaker.")
+	transcribeCmd.PersistentFlags().BoolP("srt", "", false, "Generate an SRT file for the audio file transcribed.")
 	transcribeCmd.PersistentFlags().BoolP("summarization", "m", false, "Generate a single abstractive summary of the entire audio.")
 	transcribeCmd.PersistentFlags().BoolP("topic_detection", "t", false, "Label the topics that are spoken in the file.")
 	transcribeCmd.PersistentFlags().StringP("boost_param", "z", "", "Control how much weight should be applied to your boosted keywords/phrases. This value can be either low, default, or high.")

schemas/types.go

@@ -179,6 +179,7 @@ type UploadResponse struct {
 type TranscribeFlags struct {
 	Poll bool `json:"poll"`
 	Json bool `json:"json"`
+	Srt  bool `json:"srt"`
 }
 
 type TranscribeParams struct {

utils/transcribe.go

@@ -357,6 +357,9 @@ func getFormattedOutput(transcript S.TranscriptResponse, flags S.TranscribeFlags
 		fmt.Fprintf(os.Stdin, "\033[1m%s\033[0m\n", "Summary")
 		summaryPrintFormatted(transcript.Summary)
 	}
+	if flags.Srt {
+		srtDownloadTranscript(*transcript.ID, transcript.Words)
+	}
 }
 
 func textPrintFormatted(text string, words []S.SentimentAnalysisResult) {
@@ -589,6 +592,28 @@ func summaryPrintFormatted(summary *string) {
 	fmt.Println()
 }
 
+func srtDownloadTranscript(id string, words []S.SentimentAnalysisResult) {
+	// Create a new file.
+	f, err := os.Create(fmt.Sprintf("%s.srt", id))
+	if err != nil {
+		fmt.Println("Could not create .srt file, please check your permissions")
+		return
+	}
+	defer f.Close()
+
+	output := GetSrtText(words)
+
+	// Write bytes to file.
+	_, err = f.WriteString(output)
+	if err != nil {
+		fmt.Println("Could not write to .srt file, please check your permissions")
+		f.Close()
+		return
+	}
+
+	fmt.Printf("Successfully created file %s.srt\n", id)
+}
+
 type ArrayCategories struct {
 	Score    float64 `json:"score"`
 	Category string  `json:"category"`

utils/utils.go

@@ -288,6 +288,26 @@ func GetSentenceTimestamps(sentences []string, words []S.SentimentAnalysisResult
 	return timestamps
 }
 
+func GetSrtText(words []S.SentimentAnalysisResult) string {
+	srtText := ""
+	index := 1
+	sentence := ""
+	var start string
+	for _, word := range words {
+		if sentence == "" {
+			start = TransformMsToTimestamp(*word.Start)
+		}
+		sentence += word.Text + " "
+		if strings.Contains(word.Text, ".") {
+			srtText += fmt.Sprintf("%d\n%s --> %s\n%s\n\n", index, start, TransformMsToTimestamp(*word.End), sentence)
+			sentence = ""
+			index++
+		}
+	}
+
+	return srtText
+}
+
 func GetSentenceTimestampsAndSpeaker(sentences []string, words []S.SentimentAnalysisResult) [][]string {
 	var lastIndex int
 	timestamps := [][]string{}