Focus
Click into any normal writable field on macOS.
Local-first dictation for macOS
Local whisper or cloud — your call. Clean transcripts land in the field already under your cursor, with no app switching.
macOS 13+ · Apple Silicon · 7-day Pro trial · Card required
hey team, the new onboarding flow ships tuesday — can you both review the figma file and leave comments by end of day monday?
Hey team, the new onboarding flow ships Tuesday — can you both review the Figma file and leave comments by end of day Monday?
Tested across the apps you actually use. Direct accessibility insertion where supported, clipboard fallback where it's not.
App names belong to their respective owners. Listed here to indicate compatibility, not endorsement.
VoiceBass is designed for the system-wide macOS workflow: keep focus where the words need to land.
Click into any normal writable field on macOS.
Use the global shortcut in push-to-talk or toggle mode.
VoiceBass records only during the active dictation session.
The transcript is placed back into the focused target.
Switch between fully offline transcription and our cloud whenever you want — per session, no settings dance.
Offline-first, embedded model
99+ languages, top accuracy
| Dimension | Local | Cloud |
|---|---|---|
| Privacy | Audio never leaves your Mac | TLS in transit, transcripts never logged |
| Speed | Real-time, on-device | Median < 1.5 s round-trip |
| Languages | English + 8 common languages | 99+ languages with auto-detect |
| Works offline | ✓ | — |
| Uses credits | — | ✓ |
| Best for | Sensitive content, plane mode, fixed costs | Multilingual work, longer sessions, top accuracy |
Pick a workflow - see how VoiceBass fits in, what it solves, and where it shows up.
PR descriptions, terminal commands, issue replies — all friction-heavy when you're already deep in code.
Speak through a Cursor or VS Code session: dictate code comments, draft PR bodies in GitHub, fill terminal prompts in Warp without reaching back to the keyboard.
Cloud accuracy on technical terms, local mode for proprietary code, and clipboard-paste fallback that just works in Warp and Ghostty.
Real answers to the questions buyers ask before installing a microphone-aware tool.
Local mode runs entirely offline through embedded whisper.cpp. Audio never leaves the device.
Cloud transcription stores only the metadata needed for billing — duration, model, timestamp — never the words you spoke.
Password, OTP and other secure inputs are detected and skipped. The accessibility text inserter never writes into them.
Local mode uses whisper.cpp — auditable, community-reviewed code. No proprietary black box on your machine.
Distributed directly from VoiceBass with Apple notarization. No third-party store, no silent updates without consent.
Start with a 7-day Pro trial. A card is required; cancel before day 7, no charge.
pro
Card required. Cancel before day 7, no charge.
power
Card required for trial · Cancel before day 7, no charge · Stripe-secured checkout
Privacy, languages, distribution, refunds - short answers for the things buyers ask first.
Yes. VoiceBass targets normal writable fields across every macOS app and falls back to clipboard paste where direct accessibility insertion is not reliable.
No. Audio is captured only while an active dictation session is running. The microphone indicator on macOS confirms this — VoiceBass won't show it outside an active session.
Local mode supports English plus 8 commonly used languages. Cloud mode supports 99+ languages with automatic language detection through Whisper.
Yes. Local mode runs whisper.cpp entirely on your Mac and needs no network. Cloud mode requires an internet connection to reach the VoiceBass API.
VoiceBass works system-wide with a global shortcut, offers a dual local/cloud engine choice, supports 99+ languages in cloud mode, and is account-aware so credits and devices stay synced. Apple Intelligence is a great default for short utterances inside Apple's apps; VoiceBass is built for sustained, cross-app workflows.
Not in MVP1. VoiceBass is distributed directly as a signed and notarized macOS DMG so updates can ship without store review delay.
Cloud transcription stops until the next billing period or a plan change. Local mode keeps working. MVP1 does not include rollover or top-up credits.
No. Cloud transcription stores only the metadata needed for billing — duration, model, timestamp — never the audio or transcript content.
macOS 13 or later. Apple Silicon is recommended — local mode runs significantly faster on M-series chips.
Yes. Paid plans are managed through the Stripe billing portal. You can cancel before day 7 of the trial to avoid the first charge.
On the roadmap. The current cloud model handles most technical vocabulary well; a user-tunable dictionary is planned for a follow-up release.
Yes. VoiceBass detects secure input fields through the Accessibility API and skips them — the transcript is never written into a password or OTP input.
Native macOS dictation with local-first transcription. Start a 7-day Pro trial with a card and cancel before day 7, no charge.