Review Packet
Paraspeech review kit
Use these facts and test steps before recording a Paraspeech review, demo, or comparison. The goal is simple: test the exact mode and device you mention, then keep local, offline, privacy, and platform claims inside that evidence.
Mac and iOS dictation
Review the exact mode you test.
Core facts
What it is
A Mac and iOS dictation app for inserting spoken text where you write.
Last reviewed
2026-07-07. Recheck platform, pricing, and competitor claims before publishing.
Best first audience
Mac users who dictate emails, notes, documents, chats, and drafts.
Mac support
Requires 14.0 or later. Apple Silicon Macs can use local modes; eligible Intel Macs use cloud-backed subscription models.
iOS support
Paraspeech for iPhone and iPad is available on the App Store for iOS 17.0 or later.
Local claim boundary
Local transcription and local rewriting can run offline after setup; initial model downloads and cloud-backed features require internet.
Privacy claim boundary
Local transcription and local rewriting stay on your Mac. Cloud processing is explicit and account-dependent.
Who it is for
- Writers, founders, students, operators, and teams who dictate into Mac apps.
- People comparing Mac dictation apps on privacy boundary, setup, and daily writing fit.
- Reviewers who can test a real writing task instead of only reading a feature list.
Setup requirements
- Mac users need 14.0 or later.
- Local Mac tests require a supported Apple Silicon Mac and completed model setup.
- Cloud-backed modes require internet and an account-dependent plan.
- iPhone and iPad tests require iOS 17.0 or later.
Repeatable review test
- 1
Install Paraspeech on a supported Mac and complete required permissions.
- 2
Download any required local model before testing offline behavior.
- 3
Dictate the same short email, note, browser field, and document paragraph.
- 4
Run one test with internet on and one local-mode test after setup with internet off.
- 5
Record where the text lands, how much editing is needed, and which mode was used.
- 6
Do not publish speed, accuracy, or compliance claims unless you measured and sourced them.
Reviewer proof checklist
Use this checklist before publishing screenshots, videos, or claims. It gives readers enough context to reproduce the result without turning one mode, device, or network state into a universal product claim.
System-wide writing
Record one email, one document paragraph, and one browser text field. Note whether text lands at the cursor without copy/paste.
Local-mode boundary
Use a supported Apple Silicon Mac, finish model setup first, then repeat one short local-mode test with internet off.
Cloud-backed boundary
If you test an account-backed cloud mode, label it separately and do not present it as an offline result.
iOS boundary
If you include iPhone or iPad, identify the device and iOS version and keep the claim separate from Mac local-mode behavior.
Comparison boundary
When comparing Wispr Flow or another tool, cite the competitor's current docs and avoid universal winner language.
Screenshot and video guidance
- Use the app icon and screenshots from public Paraspeech surfaces or screenshots you capture during your own review.
- Show the cursor-based test: put the cursor somewhere, hold the shortcut, speak, release, and inspect the inserted text.
- Label local/offline tests with the exact Mac model and whether setup/model downloads were already complete.
- Avoid screenshots that expose private dictated content, emails, customer data, or license keys.
Claims to avoid
- Do not say every Paraspeech feature works offline.
- Do not say Paraspeech is HIPAA, SOC 2, or enterprise-compliance certified.
- Do not claim a speed or accuracy multiplier unless your review measured it.
- Do not say Lifetime includes all cloud features; Lifetime covers local/on-device models only.
- Do not say Intel Macs support offline local mode.
Need clarification before publishing?
Send the draft claim, screenshot, or test setup before publication if you are unsure whether a local/offline/privacy statement is framed correctly.
