> Word error rates and wait times for Caddie, Wispr Flow, Kivi and Apple dictation on the same eight clips of Indian English and Hinglish speech. Method, data (CSV) and how to cite.

# Dictation accuracy benchmark: six Mac dictation apps on Indian English and Hinglish

On the same eight clips of everyday Indian English and Hinglish, Caddie Scribe made the fewest word errors (7.0%), Wispr Flow was close behind (7.9%), and Apple's built-in dictation got about one word in four wrong (25.5%).

Measured 13 October 2026 by the team that makes Caddie. 8 clips, one speaker, about 70 seconds of audio. [Download the data (CSV)](https://usecaddie.app/research/dictation-accuracy.csv)

## Key findings

- In Caddie's test of six Mac dictation setups on the same eight clips, Caddie's Scribe mode made the fewest word errors: 7.0%.
- Wispr Flow had a 7.9% word error rate on the same clips, less than a point behind Caddie Scribe, which is within this test's noise.
- Apple's built-in macOS dictation got about one word in four wrong (25.5% word error rate) on everyday Indian English and Hinglish speech.
- Apple dictation was the quickest to show text, a median 0.23 seconds after the key was let go, against 1.01 seconds for Caddie Scribe.
- Sending the whole clip once the speaker stopped, instead of streaming it, cut ElevenLabs Scribe's word error rate from 13.3% to about 7%.
- Compressing the audio as AAC while the speaker talked made the upload 7 times smaller (2.3 MB to 0.33 MB) and cut Scribe's wait from 1.24 seconds to 0.91 seconds.
- A relay server in Mumbai cut the network round trip for a speaker in India from 0.54 seconds to 0.29 seconds.

## Word error rate by app

Lower is better. Word errors are the wrong, missing and extra words, as a share of the words the speaker meant to type.

**Caddie Scribe (Most accurate)**7.0%

**Wispr Flow**7.9%

**Kivi**9.1%

**Caddie Default**10.9%

**Caddie Turbo**15.8%

**Apple dictation**25.5%

## The data

| App and mode | Word error rate | Median wait | Engine and notes |
| --- | --- | --- | --- |
| Caddie Scribe (Most accurate) | 7.0% | 1.01 s | ElevenLabs Scribe, with cleanup |
| Wispr Flow | 7.9% | 0.84 s | As shipped |
| Kivi | 9.1% | 1.01 s | by Sarvam; Hindi converted from Devanagari to Roman letters before scoring |
| Caddie Default | 10.9% | 0.46 s | Whisper large v3 on Groq, with cleanup |
| Caddie Turbo | 15.8% | 0.32 s | Whisper turbo on Groq, with cleanup |
| Apple dictation | 25.5% | 0.23 s | built into macOS, no cleanup |

Wait is the median time from letting go of the dictation key to the text appearing in the app. [Download this table as CSV](https://usecaddie.app/research/dictation-accuracy.csv).

## Method and sample

- **Sample.** Eight everyday clips from one speaker, about 70 seconds of audio in all: Indian English, Hinglish (Hindi and English in one sentence), tech terms, people's names, a self-correction, an email address and a number.

- **Same sound for every app.** The clips were played into each app through a virtual microphone, so every app heard exactly the same audio.

- **Scoring.** Wrong, missing and extra words were counted against what the speaker meant to type, and divided by the number of words meant.

- **Cleanup counts.** Caddie's three modes were scored after their cleanup step, because that is the text a user gets. Apple dictation has no cleanup.

- **Script.** Kivi writes Hindi in Devanagari, so its output was converted to Roman letters before scoring.

- **Setups.** Six setups, each as shipped on 13 October 2026: Caddie Scribe, Default and Turbo, Wispr Flow, Kivi by Sarvam, and Apple dictation on macOS.

## Limits

- Eight clips from one speaker is a small sample. With 70 seconds of audio, a gap of a point or two is within noise, so treat close rankings (Caddie Scribe and Wispr Flow) as a tie.

- Caddie makes one of the apps tested. We published the method and the data so anyone can check it.

- It doesn't yet separate how much of Scribe's lead comes from knowing a user's own words and how much from the engine.

- A larger run, on 100 clips from public datasets (AI4Bharat Svarah for Indian English and Microsoft MUCS for Hinglish), is next. This page will be updated with it.

- Superwhisper and other apps weren't in this run.

## How to cite

You're welcome to use these numbers, under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Please link to this page.

`Caddie (2026). Dictation accuracy on Indian English and Hinglish: six Mac dictation setups compared. https://usecaddie.app/research/dictation-accuracy`

## Related

[The story behind the testWhy dictation that knows your notes gets your words right](https://usecaddie.app/blog/dictation)
[Dictation for MacHold fn anywhere to dictate with Caddie](https://usecaddie.app/mac-dictation/)
[Caddie vs Wispr FlowPrice, privacy and accuracy side by side](https://usecaddie.app/vs/wispr-flow)
[Caddie vs SuperwhisperHow the two Mac dictation apps differ](https://usecaddie.app/vs/superwhisper)
[Check your own dictationFree word error rate calculator](https://usecaddie.app/tools/dictation-accuracy-checker)