Live Captions Show Up, but the Translation Is Missing on Mac: What to Check
Seeing the spoken words but no translated line? Check whether the tool translates, its display mode, source and target languages, and downloaded resources.
2026-10-09 · 7 min
The short answer
When original captions appear but translation is missing, start with the translation and display settings. Speech recognition has already produced text, so repeatedly changing audio permissions is usually the wrong first move for this particular symptom. First confirm that the tool actually translates, then check whether translated text is hidden, a target language is selected, and the language pair is ready.
Alison Workspace separates recognition, translation and subtitle display. Those choices can explain why you see one language even though the audio is working. Test a short complete sentence after each change so you can tell which step helped.
Make sure your caption tool includes translation
Captions are not automatically a translation. Apple's Live Captions transcribes supported speech. Apple's Live Translation is a different feature for supported communication apps and language pairs; turning on Live Captions does not turn every Mac video into translated audio.
The same distinction applies to platform controls: a meeting's ordinary Captions switch may only show the spoken language. Translation can have separate account, organizer or language requirements. If your current tool only transcribes, use its original captions as intended or choose a workflow that explicitly translates the audio source you need.
Check what the window is allowed to show
A translation can be available while the visible layout shows only original text. In Alison's subtitle settings, inspect the display mode: original only, translation only or bilingual. For troubleshooting, bilingual mode gives you the clearest comparison.
Keep the subtitle window large enough to show both parts at your chosen font size. Read a complete sentence before deciding the lower area is missing. If you selected translation only but see the same words as the speaker, check the target language as well: identical-looking output is not proof of a successful translation.
Check recognition and translation as separate choices
Recognition language means the language being spoken. Translation target means the language you want to read. Choosing English recognition for Japanese audio does not request a Japanese-to-English translation; it asks the recognizer to interpret the wrong speech.
- Confirm that the original line is plausible for the speaker's actual language.
- Set recognition to that language rather than the language of the website interface.
- Select the reading language in the translation control, rather than the no-translation option.
- Use a different source and target for a clear test.
- Check the app's indication of whether that pair is available.
Recognition choices and translation targets are different catalogs. Seeing a language in recognition settings does not prove it is a translation target or that every pair is available on your Mac.
Prepare the translation resources
Alison uses local recognition resources and Apple's Translation framework. The initial preparation can require an internet connection to download resources even though daily processing can then run on the Mac. A downloaded recognition model does not by itself prove translation is ready.
Connect to the internet for setup, review the app's language-readiness indication, and complete the requested resource preparation. Then test the same source and target again. Availability can vary with macOS version, region and installed resources. Alison requires macOS 15 or later and Apple silicon, M1 or newer; do not infer support on an older or Intel Mac from a general caption guide.
If you use both inputs, check each source
System audio captures accessible sound playing on the Mac; microphone input captures the selected nearby voice source. Alison's both-input mode labels them separately and allows separate recognition and translation languages.
If remote speech is translated but your own speech is not, inspect the microphone source's settings rather than replacing the remote speaker's language. Test the sources separately with headphones to reduce playback leaking into the microphone. Do not assume that one correct language pair applies to every person in a mixed-language conversation.
Use the symptom to choose the next check
No original text and no translation: inspect listening state, selected input, macOS permission and actual playback first.
Wrong original text: fix the recognition language or audio quality before judging the translated line.
Good original text but no translated text: inspect display mode, selected target, pair availability and resource readiness.
Translation that appears after a pause: let the sentence finish, then judge whether the delay is usable for your task. Recognition can revise drafts, and translation follows recognized speech.
Only some passages fail: replay a clean segment with one speaker. Names, mixed languages, interruptions and overlapping voices are more difficult than a predictable sentence. These cases call for different checks; they are not all permission failures.
Retest without changing everything at once
Use a permitted clip with one speaker, one language and about a minute of clear speech. Change one setting, restart listening if needed, and compare the result. Include a short name and number so you can see whether the recognized line is trustworthy before assessing its translation.
If the failure persists, note the macOS version, selected input, source and target languages, display mode and the exact message the app shows. Report the visible symptom rather than inventing an error code. Share a neutral test sentence instead of confidential meeting text when asking for support.
Keep the successful result in perspective
A visible translation is still automatic text. Confirm important names, dates, figures and instructions against the source or a reviewed transcript. Once the test works, return to the actual meeting, video or course and check that its audio route behaves the same way.
Alison processes accessible audio locally after resource preparation and keeps local text history under your control; that history may contain sensitive speech. Follow participant and organizational rules, and delete text when you no longer need it. The goal of troubleshooting is a usable reading aid, not a claim that every sentence is accurate.
Official references checked October 9, 2026
Apple: Live Captions on Mac
Apple: Translate messages and calls
Frequently asked questions
Why do I see captions but no translation?
Common causes include a transcription-only tool, original-only display, no target selected, an unavailable language pair or translation resources that are not ready.
Can Apple Live Captions translate every Mac app?
Live Captions transcribes supported speech. Apple Live Translation is a separate feature for supported communication apps; it is not a general translation switch for every audio source.
Should I change system-audio permission if the original text works?
Check translation and display first. A complete capture failure normally affects original captions too, although different inputs may have different permission states.
Does downloading a speech model also prepare translation?
Not necessarily. Recognition and translation use separate resources and settings. Check translation readiness for the selected pair.
Why is only one side of a conversation translated?
In both-input mode, the system-audio and microphone sources can use separate languages. Check the source that is missing its translation.
Is a short translation delay always a failure?
No. Translation follows recognized speech and drafts may change. Test complete sentences and decide whether the observed delay fits your task.
What information helps support diagnose the issue?
Provide the macOS version, input mode, language pair, display mode and visible message, using a non-sensitive sample rather than private meeting text.