File size: 5,073 Bytes
336f974 aa287b7 336f974 aa287b7 336f974 aa287b7 336f974 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | ---
license: cc-by-4.0
language:
- en
pretty_name: Mac Dictation Privacy Matrix
size_categories:
- n<1K
tags:
- dictation
- speech-to-text
- macos
- privacy
- on-device-ai
configs:
- config_name: default
data_files:
- split: train
path: data/mac-dictation-privacy-matrix.csv
---
# Mac Dictation Privacy Matrix
An open, source-reviewed dataset comparing the documented privacy boundaries of
18 Mac dictation products.
The matrix separates questions that are often collapsed into one label:
- where microphone audio becomes a transcript;
- whether an optional cloud, cleanup, assistant, or agent path exists;
- whether the documented speech path works offline after setup;
- what the publisher says it retains;
- which account, subscription, license, or provider boundary applies; and
- which platform requirements and caveats materially qualify the claim.
This repository is a distribution mirror for the canonical research page:
**[Private Mac Dictation Apps: Local vs Cloud (2026)](https://iravoice.com/research/mac-dictation-privacy-matrix)**
## Data
| File | Use |
|---|---|
| [`data/mac-dictation-privacy-matrix.csv`](data/mac-dictation-privacy-matrix.csv) | Flat rows for spreadsheets and analysis |
| [`data/mac-dictation-privacy-matrix.json`](data/mac-dictation-privacy-matrix.json) | Structured metadata, methodology, product rows, caveats, and source records |
| [`DATA_DICTIONARY.md`](DATA_DICTIONARY.md) | Field definitions and interpretation rules |
Current release: **1.5.0**
Last reviewed: **2026-07-26**
Products: **18**
First-party sources: **33**
## Quantitative snapshot
The JSON edition now publishes named cohorts so every count can be recomputed
without interpreting product prose or accepting a hidden score.
| Documentation cohort | Count | Named products |
|---|---:|---|
| At least one documented local speech-to-text path | **16/18** | IraVoice, Apple Dictation, Google AI Edge Eloquent, BetterDictation, Spokenly, Superwhisper, MacWhisper, VoiceInk, Voice Type, Whryte, TypeVox, Dicta, MacParakeet, Susurr, Sotto, Megaphone |
| No documented local speech-to-text path in the reviewed sources | **2/18** | Raycast Dictation, Wispr Flow |
| Locally capable and also documents an optional connected path | **10/16** | IraVoice, Google AI Edge Eloquent, BetterDictation, Spokenly, Superwhisper, MacWhisper, VoiceInk, Voice Type, MacParakeet, Sotto |
| First-party sources reviewed | **33** | Product, help, security, privacy, launch, App Store, and repository pages |
The 16-product cohort includes configuration-dependent and optional local
paths; it does not mean every mode or feature is local. The connected paths do
not all send microphone audio: some send transcript text, selected text,
vocabulary data, or a finished specification. These are documentation cohorts,
not security certifications or product rankings.
## Method
1. Review current first-party product, help, security, and privacy pages.
2. Separate speech transcription from later cleanup, formatting, assistant,
sync, or destination processing.
3. Describe configurable modes explicitly instead of transferring a local or
retention claim from one mode to every mode.
4. State what the reviewed documentation does not establish.
5. Publish the same product order in the HTML, CSV, and JSON editions.
## Important limitation
This is documentation research published by IraVoice, not an independent
security audit, packet inspection, source-code review, or product ranking.
IraVoice appears as one row and is identified as the publisher. Verify
time-sensitive claims against the linked first-party sources and the active
product configuration.
“Local,” “cloud,” “offline,” and “zero retention” are not interchangeable:
- local transcription does not prove that optional cleanup or the destination
is local;
- zero retention does not mean audio was never transmitted;
- offline availability may depend on model, app, or license setup; and
- the destination app creates a separate privacy and retention boundary after
text is inserted.
## Integrity
SHA-256:
```text
ff47165c59fd92996df2dd4a0578bebd8885dd4985e925d7693d69b986e272ec data/mac-dictation-privacy-matrix.csv
fb1112bfb9fac171a91e98eb9e31422ec054ca8c2a64d1153832cb9e5d89952d data/mac-dictation-privacy-matrix.json
```
## Corrections
Open an issue with:
1. the product and field that needs correction;
2. the exact first-party URL;
3. the relevant dated language; and
4. the proposed replacement wording.
Unsupported marketing claims, affiliate roundups, and undated third-party
summaries are not accepted as primary evidence.
## Citation
Use the repository's [`CITATION.cff`](CITATION.cff), or cite:
> IraVoice. *Mac Dictation Privacy Matrix: Local vs Cloud*. Version
> 1.5.0, reviewed 2026-07-26; IraVoice row updated 2026-10-01.
> https://iravoice.com/research/mac-dictation-privacy-matrix
## License
The dataset and documentation in this repository are licensed under
[Creative Commons Attribution 4.0 International](LICENSE).
|