Transcribe: audio-to-text tool with diarization and speaker role labelling

New sixth tool in the hub. Accepts MP3/WAV/OGG/M4A/FLAC/WEBM up to 200 MB,
proxies to Whisper on cuttlefish GPU. Optional speaker separation with LLM
role labelling (dommer, advokat, forelder, sakkyndig, etc. via GPT-4o-mini).
Client-side TXT / SRT / VTT download from segment data.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-05-13 18:43:22 +02:00
parent bddafea049
commit d425c99e8e
4 changed files with 600 additions and 2 deletions
+37 -1
View File
@@ -61,7 +61,7 @@ $authenticated = dbnToolsIsAuthenticated();
<section class="cap-section">
<div class="section-inner">
<h2 class="section-heading">Five tools, one corpus</h2>
<h2 class="section-heading">Six tools, one suite</h2>
<div class="cap-grid">
<div class="cap-card">
<span class="cap-label">Ask</span>
@@ -88,6 +88,11 @@ $authenticated = dbnToolsIsAuthenticated();
<h3>Redact</h3>
<p>Remove sensitive personal data with configurable Nordic / ECHR / Global profiles.</p>
</div>
<div class="cap-card">
<span class="cap-label">Transcribe</span>
<h3>Transcribe</h3>
<p>Convert audio recordings to text with optional speaker separation and Norwegian role labelling.</p>
</div>
</div>
</div>
</section>
@@ -189,6 +194,10 @@ $authenticated = dbnToolsIsAuthenticated();
<span>Redact</span>
<small>Privacy</small>
</button>
<button type="button" class="tool-tab" data-tool="transcribe" aria-pressed="false">
<span>Transcribe</span>
<small>Audio</small>
</button>
</nav>
<section class="tool-panel" aria-labelledby="toolTitle">
@@ -207,6 +216,20 @@ $authenticated = dbnToolsIsAuthenticated();
<label><input type="radio" name="language" value="no"> Norsk</label>
</div>
<div class="control-row is-hidden" id="transcribeLangControl">
<span class="control-label">Language</span>
<label><input type="radio" name="transcribeLang" value="auto" checked> Auto-detect</label>
<label><input type="radio" name="transcribeLang" value="no"> Norsk</label>
<label><input type="radio" name="transcribeLang" value="en"> English</label>
</div>
<div class="control-row is-hidden" id="diarizeControl">
<span class="control-label">Speakers</span>
<label><input type="checkbox" id="diarizeCheck" name="diarize"> Separate speakers</label>
<span class="control-label" style="margin-left:1.25rem">Count</span>
<input type="number" id="numSpeakersInput" name="num_speakers" min="2" max="10" placeholder="auto" class="num-speakers-input" aria-label="Expected speaker count">
</div>
<div class="control-row is-hidden" id="redactionControl">
<span class="control-label">Mode</span>
<label><input type="radio" name="redactionMode" value="standard" checked> Standard</label>
@@ -218,6 +241,19 @@ $authenticated = dbnToolsIsAuthenticated();
<label><input type="radio" name="redactionRegion" value="global"> Global</label>
</div>
<div class="upload-zone is-hidden" id="audioZone" role="region" aria-label="Audio upload">
<input type="file" id="audioInput" accept="audio/*,video/mp4,video/webm" aria-label="Choose audio file">
<div id="audioPrompt" class="upload-prompt">
<span class="upload-icon" aria-hidden="true">&#9654;</span>
<p>Drop audio file here, or <label for="audioInput" class="upload-browse">browse</label></p>
<p class="upload-hint"><strong>MP3</strong>, <strong>WAV</strong>, <strong>OGG</strong>, <strong>M4A</strong>, <strong>FLAC</strong>, <strong>WEBM</strong> &mdash; max 200&thinsp;MB</p>
</div>
<div id="audioFileInfo" class="upload-file is-hidden">
<ul class="upload-file-list"><li id="audioFileLine"><span id="audioFileName" class="upload-filename"></span><span id="audioFileSize" class="upload-chars"></span></li></ul>
<button type="button" id="audioClear" class="upload-clear" aria-label="Clear audio file">&times;</button>
</div>
</div>
<div class="upload-zone is-hidden" id="uploadZone" role="region" aria-label="File upload">
<input type="file" id="uploadInput" multiple accept=".pdf,.docx,.txt" aria-label="Choose files">
<div id="uploadPrompt" class="upload-prompt">