Guide for transcribing Plaud recordings using ElevenLabs Scribe v2. Use when the user wants a transcript, asks to transcribe a recording, or needs speaker-diarized text from their recordings. IMPORTANT -- transcription costs API credits.
Transcribe Plaud recordings using ElevenLabs Scribe v2 with speaker diarization, or trigger Plaud's built-in (free, lower quality) processing.
transcribe() calls ElevenLabs API and costs real credits. Before transcribing:
get_recording(file_ref) -- if transcript_cached: true, the call is FREE (returns from cache)memory_ingest(dry_run=true) to estimate cost first (see bulk-operations skill)transcribeHigh-quality transcription via ElevenLabs Scribe v2. Synchronous -- returns full transcript in one call.
| Parameter | Type | Default | Description |
|---|---|---|---|
file_ref |
| string |
| null |
| Plaud recording ID, prefix, or row number |
file_path | string | null | Local audio file path (MP3, WAV, M4A, AAC, OGG, FLAC) |
language | string | "eng" | 3-letter ISO 639-3 code |
Provide file_ref OR file_path, never both.
Response: { "status": "complete", "text": "...", "speakers": [...], "language": "eng" }
Timing: Cached transcripts return instantly. Uncached recordings take 60-120s (downloads audio, calls ElevenLabs, caches result). No polling needed.
trigger_processingTrigger Plaud's built-in server-side transcription + summarization. Free but lower quality.
| Parameter | Type | Default | Description |
|---|---|---|---|
file_ref | string | required | File ID, prefix, or row number |
language | string | "en" | 2-letter Plaud code (NOT 3-letter ElevenLabs) |
diarization | bool | true | Enable speaker identification |
Idempotent -- safe to call on already-processed files. Processing takes 2-5 minutes on Plaud servers.
get_processing_statusCheck Plaud's AI processing queue. No parameters. Returns list of files currently being processed with status.
list_languagesShow all supported transcription languages with codes. No parameters.
ElevenLabs (transcribe) | Plaud (trigger_processing) | |
|---|---|---|
| Quality | High accuracy, speaker diarization | Basic |
| Cost | ElevenLabs API credits | Free |
| Speed | ~30s-2min per recording | 2-5 minutes |
| Language codes | 3-letter (eng, fra, deu) | 2-letter (en, fr, de) |
| Caching | RustFS local cache | Plaud servers |
| Local files | Supported (file_path) | Not supported |
Common ElevenLabs language codes: eng (English), fra (French), deu (German), jpn (Japanese), zho (Chinese), spa (Spanish), kor (Korean)
Plaud S3 --> RustFS audio cache --> ElevenLabs Scribe v2 --> RustFS transcript cache --> Cognee ingestion
get_recording(file_ref="<id>") -- check transcript_cachedtranscribe(file_ref="<id>") -- instant, freetranscribe(file_ref="<id>")transcribe(file_path="/path/to/recording.mp3")trigger_processing(file_ref="<id>", language="en")get_processing_status() -- confirm queuedget_content(file_ref, content_type="summary") after processing completeslist_languages() -- returns full list with codes for both systemstranscribe() without checking cache first -- always get_recording first to avoid unnecessary costeng), Plaud uses 2-letter (en)trigger_processing for local files -- it only works with Plaud recordingsdry_run=true first