Record and transcribe audio
Updated for SystemSculpt 6.18 ·
Use the recorder for a new capture, file transcription for Markdown or subtitles, and Audio Processor for longer audio or a YouTube source. Transcription and Audio Processor use hosted SystemSculpt AI and credits; recording in Obsidian does not make transcription local.
On this page
Record in Obsidian
- Review Settings → SystemSculpt AI → Workflow for microphone and transcription preferences.
- Run Toggle audio recorder to open the compact recorder and start your capture.
- Stop when finished and wait for Recording saved before closing or suspending Obsidian.
Keep Obsidian active while recording. Background recording is not guaranteed: hiding or suspending the app can prevent completion. Workflow settings include Transcribe automatically, Keep source audio, Insert transcript at origin, cleanup controls and Send after dictation. Choose what you want saved rather than assuming source audio is retained.
Transcribe an existing file
Run Transcribe an audio file and select the source. Choose Markdown for an editable note or SRT for a subtitle file. To make SRT the default, use Workflow → Transcript output → Default file format → Subtitle file (.srt). SRT is saved as a .srt file; it does not go through Markdown cleanup or note insertion. Convert Audio to Markdown is also available from a supported audio file's menu. When several audio files arrive at once, Not now leaves them waiting with nothing uploaded; run Resume inbox transcription when you are ready.
Use Audio Processor
- Run Open audio processor, or Process YouTube video for that source.
- Choose Audio or YouTube and supply the source.
- Under Output, choose Detailed note, Meeting brief or Clean transcript, then Process audio (Process video for YouTube).
- Open the result in the vault. Each preset provides a primary note and a canonical timestamped companion transcript.
Save audio transcript opens or restores the companion transcript. Save audio summary is available for detailed-note and meeting-brief output, not clean transcript. Audio Processor continues server-side after upload and queueing; Stop watching does not cancel the job. A finished result waits on the server for up to 30 days until SystemSculpt saves it to your vault; after that, the server keeps its copy for seven days. SRT support above belongs to audio-file transcription, not a promised Audio Processor export.
File and recording limits
These are existing operation-specific limits, not interchangeable capacities. They do not promise a particular recording duration.
| Operation | Current limit / qualification |
|---|---|
| Recorder on append-capable Obsidian, documented from 1.12.3 | 512 MiB encoded capture, with incremental writes/recovery |
| Older in-memory recorder | 64 MiB desktop; 24 MiB mobile |
| Standard transcription source | 500 MiB; service duration maximum 8 hours |
| Device-file import on older hosts | 128 MiB desktop; 32 MiB mobile; vault-file slicing is a separate path |
| Audio Processor | 1,000,000,000 bytes; service duration maximum 8 hours |
A recording that fits the recorder can exceed the transcription source limit. Follow the current operation's format/size error rather than applying one number to all paths. Check the source file and the specific error before retrying.
Questions
Can I keep the source audio?
Yes. Review Keep source audio under Workflow before recording/transcribing, and wait for the capture to save.
Does Stop watching stop Audio Processor billing or execution?
No. It stops watching the queued job; it is not a cancel action. Save the completed result when ready.
Can I make subtitles?
Yes, choose SRT for audio-file transcription. That is separate from Audio Processor's note and companion-transcript outputs.
Still need help? Contact SystemSculpt.