If CapCut auto captions are not working on a phone, there are three different problems to separate: nothing generates, words generate incorrectly, or captions are correct in the editor but look wrong after export. Each needs a different test. Start with a short, safe project copy rather than changing your only edit.
Research scope: CapCut’s official mobile caption and troubleshooting guidance checked on November 4, 2026. Steps are researched, not a claim that we tested every app build. Tool names, language choices and payment restrictions can vary by device, version and account.
Diagnose the symptom before changing settings
| What happens | First useful check |
|---|---|
| Generate spins or fails without text | Internet connection, audio track and app version |
| Empty captions on a clip containing music | Confirm a clear, audible speech track |
| Captions appear in the wrong language | Select the actual spoken language before recognition |
| Several words are wrong | Edit the transcription and isolate noisy or overlapping audio |
| Captions are correct but badly timed | Correct segment boundaries and preview the timeline |
| Preview is correct but exported video differs | Check export results outside CapCut before deleting the project |
| The feature shows a subscription prompt | Check account entitlement rather than treating it as an audio bug |
Check the audio that CapCut is trying to transcribe
CapCut’s auto-caption troubleshooting page names muted tracks, unclear speech, background noise and wrong language selection as common reasons for empty or inaccurate captions.
Play the source section with sound inside the project. Is the speaker audible? Is the voiceover on a muted layer? Is music much louder than the speech? If your audio is unintelligible to a listener, changing the caption font will not fix the transcription.
A controlled diagnostic example: make a copy containing ten seconds of one clearly spoken sentence, with no music and no overlapping dialogue. Try recognition on that copy. If it succeeds while your full project fails, the contrast points toward an audio mix or project-specific issue. It does not prove that one particular effect is defective.
Run recognition again with the correct language
On iPhone or Android, CapCut’s mobile instructions describe opening a project, choosing Captions → Auto Captions, selecting a language and generating. Some versions show the caption tools under Text; follow the controls in the app you actually have.
Do not assume Auto Detect is selecting the right language. If the video contains Spanish speech and English on-screen text, the language for speech recognition is Spanish. A bilingual speaker may need extra review even if the tool offers translation. Translation and transcription are not identical operations.
After generating, read a few captions while listening. Confirm names, numbers and negatives before styling the entire sequence.
If generation hangs, isolate the connection
CapCut says cloud processing and unstable connectivity can interrupt caption generation. Check that ordinary internet access works, then try a different reliable connection. If the project is locally stored, preserve the draft while doing this.
CapCut’s network troubleshooting guide suggests checking mobile data permissions, restarting the app or device and updating through an official store. It also warns against clearing system App Data, which can delete local projects.
Do not uninstall CapCut simply to re-run captions. Before any maintenance step, use our missing-project safety checks to confirm where the editable draft is stored.
Correct bad captions instead of rerunning everything
CapCut’s inaccurate caption help documents correcting individual mobile caption lines and splitting long segments when needed. Tap the relevant subtitle segment in your timeline, open the text-editing control your version offers, and replace misrecognized words.
For example, the invented spoken line “I need fifteen clips, not fifty” has two high-impact recognition risks: the number and the word “not.” Compare the result with the audio. Do not correct it solely from what you think the speaker intended.
If two people speak over each other, a second recognition pass may still make mistakes. Separate the audio or temporarily mute unrelated sound in a duplicate project, if feasible, and regenerate a short section. Keep the original unmodified. CapCut’s guidance notes that targeted re-recognition may require a workaround rather than a dedicated mobile button.
Fix timing and text that cover the picture
A transcript can be accurate yet unreadable. Play the captioned section at normal speed. Break crowded text at natural speech pauses and check that no line appears too early or persists into the next speaker’s sentence.
Style comes last. Choose readable size and contrast, inspect the result at actual phone size and keep important text away from interface overlays used by the platform where you will publish. If a portrait video crops its subject, the issue is framing rather than speech recognition; see our vertical-video framing guide.
For a complete creation workflow, use CapCut Auto Captions on Mobile. This article focuses only on troubleshooting failures.
Check the finished export separately
Open the exported video in Photos, Gallery or a normal video player and inspect a caption near the beginning, one in the middle and the final segment. A text overlay in CapCut’s preview does not guarantee that the output saved correctly, and an SRT subtitle file is not the same as burned-in captions in an MP4.
If export itself fails or the saved video is missing, follow CapCut Won’t Export?. If you see a Pro or payment prompt, use the paid-feature checklist instead of repeatedly regenerating the words.
What to send support when captions still fail
Provide the app version, phone OS, chosen recognition language, whether speech is audible, whether a ten-second plain clip works, and the exact error message. A non-sensitive screenshot of the caption screen can help. Keep your original local footage and project intact throughout the investigation.
