Frequently Asked Questions
Everything you need to know about SwitchScribe, our agent switchboard, Singapore PDPA privacy, and SGD billing.
The Switchboard
What is the Switchboard?
The Switchboard models every transcription and enrichment step as an independent, database-backed agent (e.g. Speaker Diarization, PII Redaction, Translation). You can toggle agents on/off per job or set account-level defaults to optimize cost and latency.
Can I re-run agents without re-transcribing the entire audio?
Yes! If you want to change your summary style or translate an already transcribed file, simply change the Switchboard toggles on the recording and re-run. Only the affected agents will execute, saving time and SGD cost.
What are Switchboard Presets?
Presets are pre-configured combinations of agents and providers for specific workflows (e.g. "Fast & cheap", "Balanced", "Meeting notes", "Compliance"). You can also save your own custom presets.
Privacy & Singapore PDPA
How does SwitchScribe comply with Singapore PDPA?
We adhere to the Consent, Purpose Limitation, Retention Limitation, and Transfer Limitation principles of the Personal Data Protection Act. We provide automated media purges (7 to 90 days), disclose all vendor processors, and offer one-click permanent erasure of all user data.
Is my voice data used to train AI models?
No. All transcription and LLM inference is performed under strict zero-data-retention enterprise agreements. Your recordings and transcripts are never used for model training.
How does PII Redaction work?
Our Redaction Agent automatically detects and masks Singapore NRIC/FIN numbers (e.g. S****567A), phone numbers, email addresses, and custom confidential keywords.
Editor & Exports
What formats can I export my transcripts in?
You can export in TXT, SubRip (.srt), WebVTT (.vtt), Microsoft Word (.docx), and structured JSON (.json) with word-level timestamps and confidence metrics.
Can I rename speakers in bulk?
Yes. Simply click any speaker badge (e.g. "Speaker 1") to open the rename dialog and change it across all utterances in the transcript instantly.
What are Confidence Flags?
Our speech models assign confidence scores to every word. Any word with confidence below 80% is highlighted with a dotted underline so you can quickly proofread challenging words.