Accurate Automatic Transcription

VoiceBase was conceived to do for voice what Google did for text: make it instantly searchable and shareable by creating a rich, queryable database. Our effective speech analytics begins with an accurate, automated transcript which enterprises can then index and analyze to deliver actionable insights.

Facilitate Discovery

Navigate recordings by jumping to the exact part of the speech you're looking for, through click-to-play keywords.

Improve User Experience

Recordings can be time consuming and tedious to listen to, even 20 second voicemails. Provide users with an instant transcription of voicemails, conference calls, and meetings to improve access to vital information.

Create a Valuable Archive

Make use of your storage fees and let users search through their archives of recordings. Audio and video archives hold useful information that businesses should be able to re-access over and over again.

Transcription
Stereo Speaker ID
Custom Vocabulary
Number Formatting
Instant Search

Discover More Features >

Languages

United States US English

Australia AUS English

United Kingdom UK English

Spain Spanish

Latin America Spanish

The Netherlands Dutch

Brazil Portuguese

France French

Germany German

Italy Italian

Features For Any Use Case

curl https://apis.voicebase.com/v3/media \
    --form media=@recording.mp3 \
    --form configuration='{
      "speechModel" : {
        "language" : "en-US"
      }
    }' \
    --header "Authorization: Bearer ${TOKEN}"

curl https://apis.voicebase.com/v3/media \

--form media=@recording.mp3 \

--form configuration='{

"speechModel" : {

"language" : "en-US"

}

}' \

--header "Authorization: Bearer ${TOKEN}"

Select a Language

Use the language configuration parameter to set the language. Omitting the parameter defaults the language to U.S. English (en-US).

curl https://apis.voicebase.com/v3/definition/vocabularies/earningsCalls \
  --request PUT \
  --header "Content-Type: application/json" \
  --header "Authorization: Bearer ${TOKEN}" \
  --data '{
      "vocabularyName": "earningsCalls",
      "terms": [
        {
          "soundsLike": [
            "A.F.F.O."
          ],
          "term": "AFFO",
          "weight": 2
        },
        {
          "soundsLike": [
            "Aypack"
          ],
          "term": "APAC",
          "weight": 2
        },
        {
          "term": "CapEx"
        }
      ]
    }'

curl https://apis.voicebase.com/v3/definition/vocabularies/earningsCalls \

--request PUT \

--header "Content-Type: application/json" \

--header "Authorization: Bearer ${TOKEN}" \

--data '{

"vocabularyName": "earningsCalls",

"terms": [

{

"soundsLike": [

"A.F.F.O."

"term": "AFFO",

"weight": 2

{

"soundsLike": [

"Aypack"

"term": "APAC",

"weight": 2

{

"term": "CapEx"

}

]

Add Custom Vocabulary

Accurately spot non dictionary terms like:
• Jargon
• Proper Nouns
• Acronyms
• Hyphenated words
• Multi-word phrases.

curl https://apis.voicebase.com/v3/media  \
    --form media=@recording.mp3 \
    --form configuration='{
        "ingest": {
          "stereo": {
            "left" : { "speakerName": "Customer" },
            "right": { "speakerName": "Agent"    }
          }
        }
      }' \
    --header "Authorization: Bearer ${TOKEN}"

curl https://apis.voicebase.com/v3/media \

--form media=@recording.mp3 \

--form configuration='{

"ingest": {

"stereo": {

"left" : { "speakerName": "Customer" },

"right": { "speakerName": "Agent" }

}

}' \

--header "Authorization: Bearer ${TOKEN}"

Define Channels

Recording and processing calls in stereo can significantly improve transcription accuracy and analytical insights. To realize the benefit, each speaker is recorded on a different channel (left or right), and the speaker metadata is provided to VoiceBase when uploading the recording.

1
00:00:02,76 --> 00:00:05,08
Agent: Well this is Michael
thank you for calling A.B.C.

2
00:00:05,08 --> 00:00:07,03
Cable services. How may I help you today.

3
00:00:08,28 --> 00:00:11,93
Customer: Hi I'm calling because I'm
interested in buying new cable services.

4
00:00:12,64 --> 00:00:16,43
Agent: OK great let's get started.

00:00:02,76 --> 00:00:05,08

Agent: Well this is Michael

thank you for calling A.B.C.

00:00:05,08 --> 00:00:07,03

Cable services. How may I help you today.

00:00:08,28 --> 00:00:11,93

Customer: Hi I'm calling because I'm

interested in buying new cable services.

00:00:12,64 --> 00:00:16,43

Agent: OK great let's get started.

Format for Captioning

VoiceBase can generate subtitles or closed captions for your video project, by allowing you to retrieve the transcript of your audio or video file using the WebVTT or SubRip Text (SRT) format.

No special configuration is required. All transcripts are always available in four formats: JSON word-by-word, plain text, WebVTT and SRT.

{  
   "speechModel" : {
         "language" : "en-US",
          "extensions" : [ "voicemail" ]
   },
   "knowledge": {
     "enableDiscovery" : false
   },
   "transcript": {
     "formatting" : {
       "enableNumberFormatting" : true
     }
  },       
  "publish":{  
     "callbacks": [  
        {  
           "url": "https://example.com/transcription",
           "method": "POST",
           "type": "analytics",
           "include":[  
              "transcript"
           ]
        }
     ]
  }
}

{

"speechModel" : {

"language" : "en-US",

"extensions" : [ "voicemail" ]

"knowledge": {

"enableDiscovery" : false

"transcript": {

"formatting" : {

"enableNumberFormatting" : true

}

"publish":{

"callbacks": [

{

"url": "https://example.com/transcription",

"method": "POST",

"type": "analytics",

"include":[

"transcript"

]

}

]

}

Optimize for Voicemail

VoiceBase is able to return automatic transcripts for voicemail recordings which can then be delivered via email or SMS. Here is an optimized configuration for fast and accurate voicemail transcription. This includes features that benefit shorter forms of audio, however they are all optional.

Disable Phrase Groups and Topic Extraction
Enable Number Formatting
Use Callbacks