Wispr Flow Vietnamese dictation: setup and mixed-language limits

Wispr Flow supports Vietnamese, but that does not establish how well it will preserve an English identifier inside a Vietnamese instruction. On desktop or iOS, start by checking a Vietnamese-only session, then add the English terms you actually use at work. Android uses Auto-detect according to the current guide.

This is a setup guide based on Wispr documentation checked on September 11, 2026. The examples and downloadable worksheet are an original test kit, not recorded Flow output or an independent accuracy benchmark. For the broader buying decision, see our Wispr Flow review.

Vietnamese support and mixed-language dictation are different questions

Wispr explicitly includes Vietnamese in its supported-language research page. Its language settings guide says Flow detects a language at the beginning of each dictation session, rather than choosing again for every word.

That distinction matters for a sentence such as “Kiểm tra webhook trước khi deploy lên staging.” Most of the sentence is Vietnamese, while three technical terms are English. A longer passage alternating complete Vietnamese and English sentences presents a different task again. Do not treat success with one as proof that the other works.

Wispr recommends speaking mainly in one language with occasional words from another and documents limitations with rapid switching. It does not provide a Vietnamese–English benchmark in the sources used here. The accuracy and limitations page also distinguishes language support from formatting quality; it does not list Vietnamese among its dedicated-formatting languages.

Set the language for your device

On Mac or Windows, open Flow from the menu bar or system tray, then go to Settings → General → Dictation languages. Select Vietnamese, turn off Auto-detect for the initial check, and choose Save and close. Wispr says the change applies to the next dictation.

On iOS, open Flow → Settings → General → Set Language. Select Vietnamese and leave the language screen to close the sheet. Start a new dictation after making the change.

On Android, the current language guide says Flow uses Auto-detect only and has no language selection screen. You cannot run the manual-language comparison below on that platform. Record the detected output and your app version instead; do not infer Android controls from desktop instructions.

On desktop or iOS, once Vietnamese-only input is usable, compare it with Vietnamese and English selected together. You can also compare Auto-detect as a separate condition. Wispr recommends narrowing selection to the languages you use; this is vendor guidance, not a measured improvement for your voice.

Check meaning before punctuation

Use short passages with a specific thing to preserve. These are proposed inputs, not examples of successful transcription:

• Accents and names: “Nhờ Hương gửi bản tóm tắt cho Nguyễn vào chiều thứ sáu.” Check both names and Vietnamese diacritics.
• Negation: “Không xóa dữ liệu khách hàng khi thử lại.” Losing “không” reverses the instruction.
• Date and amount: “Chuyển hai triệu năm trăm nghìn đồng trước ngày mười lăm tháng chín.” Accept equivalent number formatting, but check the value and date.
• Developer terms: “Kiểm tra webhook trước khi deploy lên staging.” Check the intended terms as well as the action.
• Exact identifier: “Giữ nguyên biến userId trong payload.” Decide in advance that userId must keep its casing.

For an actual command, file path or identifier, type or paste the exact token when dictation cannot preserve it reliably. A readable paraphrase can be fine in an email and unusable in a terminal.

Run the 20-sentence check on your own setup

Download the Vietnamese and English sentence sheet and the blank results log. The sheet contains five Vietnamese sentences, five Vietnamese sentences with developer terms, five English sentences and five passages that switch languages. The final group deliberately probes a documented limitation; it is not a promise of support.

1. Record your OS, Flow version, microphone, target app and language selection. Use non-sensitive sample text in an empty document.
2. On desktop or iOS, read each Vietnamese and mixed sentence with Vietnamese selected. Test the English group with English selected. Record which language you choose for the switching group; analyze that stress case separately from the single-language baseline.
3. On desktop or iOS, read the same sheet with Vietnamese and English selected, then with Auto-detect as a separate condition. On Android, test Auto-detect only and mark manual selection as not applicable.
4. Save the inserted text before editing. Record whether the meaning, required terms and numbers survived, then time the corrections until the text is usable.
5. Repeat each condition if practical and rotate the order. Keep failed attempts in the log, including cases where no text appears.

The log uses not_run for trials you have not performed; empty measurements do not mean zero errors or instant output. Record your start and end points if you measure latency. For example, measure from releasing the dictation shortcut to all text appearing, consistently for every trial.

Report results separately for each language group. With only one speaker and one microphone, the conclusion applies to that setup. A combined percentage can hide the very mixed-language failures you wanted to investigate.

Choose the next fix from the error

On desktop or iOS, if Flow selects the wrong language, return to a single-language selection and start a new session. For a full English paragraph after Vietnamese, try ending the first session and changing the selected language before starting the next. Compare the editing work rather than assuming the extra steps are worthwhile.

If the language is right but a name or developer term is wrong repeatedly, use the developer dictionary guide. A vocabulary entry addresses a different problem from whole-session language selection.

If text is missing or recording fails, this sentence check cannot establish the cause. Flow requires an internet connection, and recording or insertion problems need separate troubleshooting. Avoid counting a failed insertion as a spelling error.

Before paying for another tool, run a short equivalent check with your device's built-in dictation. Choose the setup that preserves your instructions with acceptable correction time, using your own results rather than a general English accuracy claim.

To record repeatable comparisons across tools, the developer dictation benchmark kit extends this sentence set with a results template and local summary script.