action-transcribe-event.xml

<?xml version="1.0" encoding="UTF-8" ?>
<definition xmlns="http://workflow.opencastproject.org">

  <id>Transcribe</id>
  <title>Generate Subtitles using Whisper AI (Experimental)</title>
  <tags>
    <tag>archive</tag>
  </tags>
  <displayOrder>700</displayOrder>
  <description>
	  Generate and replace existing subtitles using Whisper-AI, this is an experimental AI model that can transcribe
    speech from almost any language to text. It is also possible to translate to English with the model.

    After running this workflow, you need to republish the event in order to add the subtitles to the search service 
  </description>

  <configuration_panel>
  <![CDATA[
    <div id="workflow-configuration">

	    <fieldset>
        <legend style="color:seagreen">Select the Transcription model: </legend>
        <ul>
          <li>
            <input id="mTiny" name="transcriptionModel" type="radio" class="configField" value="TINY">
	          <label for="mTiny">Tiny (Speed ~32x) [Disabled]</label>
          </li>
          <li>
            <input id="modelBase" name="transcriptionModel" type="radio" class="configField" value="BASE">
	          <label for="modelBase">Base (Speed ~16x)</label>
          </li>
          <li>
            <input id="mSmall" name="transcriptionModel" type="radio" class="configField" value="SMALL">
	          <label for="mSmall">Small (Speed ~6x) [Disabled]</label>
          </li>
          <li>
            <input id="mMedium" name="transcriptionModel" type="radio" class="configField" value="MEDIUM">
	          <label for="mMedium">Medium (Speed ~2x)</label>
          </li>
          <li>
            <input id="mLarge" name="transcriptionModel" type="radio" class="configField" value="LARGE">
	          <label for="mLarge">Large (Speed ~1x) [Disabled]</label>
          </li>
        <ul>
        <legend style="color:seagreen">Automatic translation (Do not add for events in english) </legend>
        <br>
          <li>
            <input id="transEn" name="translEn" type="checkbox" class="configField" value="false" />
	          <label for="transEn">Add automatic transcription to english</label>
          </li>
        <!-- Place to keep the layoutType state --> 
          <!-- Place to keep the layoutType state --> 
          <input type="hidden" id="layoutType" class="configField" value="NONE" />
      </fieldset>

      <script>
          document.getElementById("mTiny").disabled = true;
          document.getElementById("mSmall").disabled = true;
          document.getElementById("mLarge").disabled = true;


      </script>
      
     </div>

  ]]>

  </configuration_panel>
  
  <operations>
  <state-mappings>
    <state-mapping state="running">EVENTS.EVENTS.STATE_MAPPING.PUBLISHING</state-mapping>
    <state-mapping state="failing">EVENTS.EVENTS.STATE_MAPPING.PUBLISHING</state-mapping>
  </state-mappings>


    <!-- Set translation parameters-->

    <operation
      id="conditional-config"
      exception-handler-workflow="partial-error"
      description="Setting transcription variables">
      <configurations>
        <configuration key="configuration-name">transcription-model</configuration>
        <configuration key="condition-1">${mTiny}</configuration>
        <configuration key="value-1">{{ whisper_server }}:9000</configuration>
        <configuration key="condition-2">${modelBase}</configuration>
        <configuration key="value-2">{{ whisper_server }}:9001</configuration>
        <configuration key="condition-3">${mSmall}</configuration>
        <configuration key="value-3">{{ whisper_server }}:9002</configuration>
        <configuration key="condition-4">${mMedium}</configuration>
        <configuration key="value-4">{{ whisper_server }}:9003</configuration>
        <configuration key="condition-5">${mLarge}</configuration>
        <configuration key="value-5">{{ whisper_server }}:9004</configuration>
        <configuration key="no-match">no-model</configuration>
 
      </configurations>
    </operation>

    <!-- Analyze media assets -->

    <operation
      id="analyze-tracks"
      exception-handler-workflow="partial-error"
      description="|__   Analyze tracks in media package and set control variables">
      <configurations>
        <configuration key="source-flavor">*/source</configuration>
      </configurations>
    </operation>

    <!-- Delete old subtitles --> 

     <operation
      id="tag"
      exception-handler-workflow="partial-error"
      fail-on-error="false"
      description="Remove old archived subtitles ">
      <configurations>
        <configuration key="source-flavors">captions/vtt+de, captions/vtt+en</configuration>
        <configuration key="target-tags">-archive,-engage-download</configuration>
      </configurations>
    </operation>

    <!-- Generate the subtitles - only if Presenter is available --> 

    <operation
      id="execute-once"
      if= "(NOT ${presentation_source_audio}) AND ${presenter_source_audio}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presenter">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presenter/source)} #{id} #{out}</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+de</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>

    <operation
      id="execute-once"
      if= "((NOT ${presentation_source_audio}) AND ${presenter_source_audio}) AND ${transEn}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presenter (English Translation)">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presenter/source)} #{id} #{out} translate</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+en</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>

    <!-- Generate the subtitles - only if Presentation is available --> 

    <operation
      id="execute-once"
      if= "(NOT ${presenter_source_audio}) AND ${presentation_source_audio}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presentation">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presentation/source)} #{id} #{out}</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+de</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>

        <operation
      id="execute-once"
      if= "((NOT ${presenter_source_audio}) AND ${presentation_source_audio}) AND ${transEn}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presentation (English Translation)">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presentation/source)} #{id} #{out} translate</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+en</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>


    <!-- Generate the subtitles - if Presenter and Presentation are available --> 

    <operation
      id="execute-once"
      if= "(${presenter_source_audio}) AND ${presentation_source_audio}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presentation and presenter together">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presenter/source)} #{id} #{out}</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+de</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>

        <operation
      id="execute-once"
      if= "((${presenter_source_audio}) AND ${presentation_source_audio}) AND ${transEn}"
      fail-on-error="true"
      exception-handler-workflow="error"
      description="Run Whisper AI for presentation and presenter together (English Translation)">
      <configurations>
        <configuration key="exec">/etc/opencast/scripts/whisper-getvtt.sh</configuration>
        <configuration key="params">${transcription-model} #{flavor(presenter/source)} #{id} #{out} translate</configuration>
        <configuration key="output-filename">subtitle-output.vtt</configuration>
        <configuration key="target-flavor">captions/vtt+en</configuration>
        <configuration key="target-tags">archive, engage-download</configuration>
        <configuration key="expected-type">Attachment</configuration>
      </configurations>
    </operation>

    <!-- Archive current state -->

    <operation
      id="snapshot"
      exception-handler-workflow="partial-error"
      description="Archive current mediapackage">
      <configurations>
        <configuration key="source-tags">archive</configuration>
      </configurations>
    </operation>

    </operations>

</definition>