Search documentation

Find setup, feature, and troubleshooting guides.

Documentation

Audio & Audio Processor

Record and transcribe audio, or turn audio and YouTube videos into structured audio notes without leaving Obsidian.

Audio Processor

Turn existing audio or a YouTube video into 2 linked Markdown notes. Choose the shape of the primary note before processing. SystemSculpt keeps a separate timestamped transcript for source checking and recovery. The transcript and any generated summary stay in the language spoken.

Choose the source
Run Open audio processor for the Audio and YouTube inputs, or Process YouTube video to open the YouTube input directly. Audio can come from your vault when available or from your device, up to 1 GB (1,000,000,000 bytes).
Let the server finish
After an audio upload finishes or a YouTube job is queued, processing continues on the SystemSculpt service. Stop watching closes progress in Obsidian without stopping the server job.
Keep the result in your vault
SystemSculpt saves 2 linked Markdown notes: the primary note in your selected format and a separate timestamped transcript for source checking and recovery. Completed server results remain available for 7 days, so reopen Obsidian within that window if you leave while processing continues.

Choose the note you want to read

The preset changes the primary note, not the underlying transcript. You can choose the fuller research format, a concise meeting handoff, or a clean transcript without timestamp clutter.

Detailed note
Keep the current structured note with the summary, key points, decisions, action items, discussion, open questions, and timestamped transcript together.
Meeting brief
Lead with the outcome, concise summary, decisions, action items, and open questions. The full timestamped transcript stays in its linked companion note.
Clean transcript
Read the transcript as plain paragraphs without timestamps or speaker labels in the primary note. The canonical timestamped transcript remains available separately.
2 linked notes
The selected primary note and separate timestamped transcript are saved together under SystemSculpt/Audio Notes and link to each other.
  • •A primary note in the format you selected before processing
  • •A separate canonical transcript with timestamps for checking the source
  • •Stable job markers that let SystemSculpt restore moved or interrupted results
  • •YouTube citations that open the matching source timestamp when a summary is included

YouTube citations open the matching source timestamp when the selected note includes a summary. Audio Processor opens the primary note when delivery finishes. Use Open transcript, or run Save audio transcript from any saved Audio Processor note, to open or restore the linked transcript. Run Save audio summary to create or open an optional summary-only note when the selected preset generates one. Completed server results remain available for 7 days.

Recording controls

Run Toggle audio recorder from the command palette or assign your own hotkey in Obsidian. The toolbar appears inside Obsidian so you never leave the note you are working on.

  • •The recorder remembers your preferred microphone, watches for device changes, and recovers if a track drops.
  • •The recording modal gives you transport controls, a waveform preview, and live status messages.
  • •Existing recordings upload through the SystemSculpt service so you can transcribe files you already captured.
Settings tie-ins
The Workflow tab controls recording, SystemSculpt transcription, and output behavior. Changes apply the next time you open the recorder.

Transcription flow

When you submit audio, the plugin cleans the file, uploads it, monitors progress, and finalises the transcript. Processing continues on the SystemSculpt service for recordings up to 8 hours long.

  • •Audio is normalised automatically (with optional resampling) before it uploads.
  • •SystemSculpt detects the spoken language automatically and keeps the transcript in that language, including code-switches.
  • •Progress updates stream back to the modal so you can see when a job starts, queues, and finishes.
  • •Optional post-processing prompts run after the transcript completes when the toggle is enabled in settings.
Command paletteTranscribe with SystemSculpt uses the same pipeline as the recorder, just without capturing new audio.
DiagnosticsErrors appear as Obsidian notices. Capture the notice and job state if a transcript fails repeatedly.

SystemSculpt transcription path

Managed upload
The plugin uploads audio through the SystemSculpt job flow and shows progress while the file is prepared.
Managed transcription
SystemSculpt runs transcription. You only choose recording and output preferences.
Vault output
Save the transcript as a note or attach it to Chat when the managed job completes.

Checklist

  • •Select your preferred microphone in Settings → SystemSculpt → Workflow.
  • •Run Toggle audio recorder from the command palette or assign your own hotkey in Obsidian.
  • •Open existing audio via the file context menu → “Transcribe with SystemSculpt”.
  • •Run Open audio processor when you want your selected primary output plus a separate canonical timestamped transcript from audio or YouTube.
  • •Choose whether recording should transcribe automatically, whether source audio stays in the vault, and whether the transcript is inserted only when the same note, editor, cursor or selection, or chat conversation remains active.

Previous

Workflow guides

Next

Embeddings & search