Captions vs subtitles: what to choose for speech, sound and translation

Dylan de Heer
Captions usually help someone understand the audio in the same language, including relevant sounds and speaker information. Subtitles often translate dialogue into another language. Both appear alongside the media, but the labels are not used consistently: in the UK, “subtitles” can also mean the same-language text used for accessibility.
Choose by what the audience needs, rather than the label alone. Check the language, whether important sounds are described, how the text follows the audio and whether the person needs a record to revisit later.
For example, a translated conversation can still leave someone unable to understand a doorbell that changes what happens next. A same-language transcript can preserve the words without showing each line at the moment it is spoken. These are different needs, and one option may not meet all of them.
Captions vs subtitles at a glance
Choice | Main question it answers | What to check |
|---|---|---|
Same-language captions | What is being said, and what relevant audio information would I otherwise miss? | Dialogue, speaker identification where needed, meaningful sounds, accuracy and timing |
Translated subtitles | What does the dialogue mean in a language I understand? | Translation, readable phrasing, timing and whether non-speech information is also needed |
Open captions | Must the text always be visible in this version? | It stays visible rather than being a viewer-controlled option |
Closed captions | Can the viewer choose whether to show the text? | The player or platform must expose and support that option |
Transcript | Can I read, search or revisit the content outside the moving text? | Completeness, speaker context, access and whether the record is actually retained |
Open and closed describe how captions are displayed, not whether the text is translated. A translated subtitle track can also be selectable. A viewer-controlled track still needs accurate content; “closed” does not mean checked or complete.
Why the terminology can be confusing
W3C's media guidance uses “captions” for same-language text and “subtitles” for translated speech, while explicitly recognising that some regions use “subtitles” for both. It also explains that translated subtitles may include relevant non-speech information rather than being limited to dialogue. W3C: captions and subtitles.
That regional difference matters when someone asks for “English subtitles”. They might want an English translation, same-language text for English speech, or a fuller accessible presentation of the audio. Clarify the intended audience and outcome before deciding what to produce.
A useful brief would say: “The audience needs to follow the English dialogue without hearing the soundtrack, including the sounds that affect the explanation.” Another would say: “The audience can hear the Spanish audio but needs an English translation.” These are clearer than relying on the word “subtitles” to carry the whole requirement.

Captions need more than recognised dialogue
When sound contributes meaning, the audience needs that information too. A useful caption may identify an off-screen speaker, describe a significant alarm or indicate that a remark was spoken rather than appearing on screen.
Consider a fictional training video. The presenter explains a task, an off-screen colleague interrupts and a timer signals that a stage has ended. A word-perfect rendering of the presenter's speech could still omit the interruption and timer. The text should preserve the information needed to understand the sequence.
This does not mean describing every incidental noise. Ask what someone would miss if they could not hear the audio, then provide the relevant information clearly and at the appropriate moment. A decorative sound that adds nothing to the explanation has a different role from an alarm that changes the required action.
Automatic speech recognition can help generate a starting point, but it is not proof that this work has been completed. It may mishear a name, omit a short word or fail to describe a non-speech event. W3C's guidance distinguishes automatically generated text from the editing needed to make captions accurate. W3C: caption quality and automatic captions.
Translated subtitles solve a language problem
Translated subtitles help someone follow speech in another language. A useful translation preserves the intended meaning while remaining readable at the pace of the media.
In a fictional research discussion, a participant might use a local term that has no simple one-word equivalent. A fluent-looking sentence can still change that meaning. Review the translation in context, paying particular attention to names, numbers, negation, specialist terms and expressions whose meaning depends on the situation.
Do not assume that changing the language makes the material accessible in every other respect. A viewer may need a translated account of a significant sound, clearer speaker identification or a transcript they can read at their own pace. Establish those needs separately.
For a live conversation, also distinguish text translation on your own screen from interpretation for everyone in the call. An app may provide one without transmitting translated speech or captions to other participants. The available output matters more than the presence of a translation setting.
Open captions and closed captions are display choices
Open captions remain visible in the version being shown. Closed captions can be turned on or off by the viewer. W3C uses this distinction when describing caption delivery. W3C: open and closed captions.
Always-visible text can be useful when a distribution method does not reliably expose a caption control. It also fixes choices such as placement and appearance into that version. Check that it does not obscure an important visual detail and that it remains readable at the size people will actually use.
A selectable track gives the viewer control, but only if it reaches the final destination and works in the player. Inspect the delivered media, not just the editor. Confirm that the expected language is available, the control can be found and the text remains synchronised after an upload or export.
Neither delivery choice fixes an inaccurate transcript, a misleading translation or an omitted speaker change. Treat content review and delivery review as separate checks.
When you also need a transcript
A transcript lets someone read the content as a document instead of following short lines while the audio progresses. Depending on the purpose, it may need relevant sound information and descriptions of visual information as well as speech. W3C distinguishes basic and descriptive transcripts and notes that interactive transcripts depend on the player's capabilities. W3C: transcripts.
For a recorded conversation, a retained transcript can help you revisit a passage, check a quotation or prepare follow-up work. It is different from a summary that selects only the main points; see meeting notes vs transcript.
Do not infer persistence from a feature name such as “Live transcript”. That name may describe a scrollable current-session view. Check whether the text survives stopping the session, whether audio is retained and what the app actually allows you to save or export.
If you need to substantiate a quotation later, plan for an authorised saved source before the conversation begins. A transient line on screen is not automatically a durable record. Where timestamps and audio are available, use them to check the passage rather than assuming the recognised words are exact.
A worked choice: three readers, different requirements
Imagine a fictional organisation sharing a recorded demonstration and holding a follow-up discussion.
A viewer watching the demonstration without sound needs the dialogue and important audio context at the right time. Review same-language captions, including speaker changes and sounds that affect the instructions. A translation is not the missing feature in this case.
A viewer who hears the audio but does not understand the spoken language needs a suitable translation. Check that the subtitle timing and wording work together. If that viewer also cannot hear the soundtrack, translated dialogue alone may leave important information out.
A colleague preparing the follow-up discussion may want to search a retained transcript and revisit the source. Captions can help during viewing, but a document or interactive transcript may better support this later task. The colleague should still distinguish recorded statements from proposals or conclusions drawn afterwards.
The choices can be complementary. The right result may be captions, translated subtitles and a transcript, each checked for its intended purpose. Avoid creating extra versions without knowing who needs them and how they will be maintained when the source changes.

How this maps to Weeve's live text and saved recordings
Weeve's Subtitles feature displays recognised speech from audio on your Mac. Its current interface offers Subtitles and Live transcript views, and a Caption language choice for original-language text or available local translation. The product feature name does not itself establish that every meaningful sound has been described or that a video caption file has been produced. Follow live audio with subtitles.
The standalone feature keeps temporary session text. Stopping or restarting clears that history. If you need a saved recording and transcript, start Recording separately before the speech you need to keep. A saved recording is a separate source for later review; it does not save the translated caption text. Translate live audio explains that distinction and the model requirements.
Live recognition and this translation workflow run on your Mac with downloaded models. The source app may still use online services, and other optional Weeve workflows have their own processing boundaries. Do not turn a local overlay into a claim about the entire call or every integration.
Use live text to help follow the current speech, a separately saved source when you need to revisit it, and reviewed caption/subtitle delivery tools when you are producing media for others. These are related workflows with different outputs.
Make the final choice by checking the result
Before you share or rely on the text, confirm the intended language, the audio information the audience needs, timing and readability, and whether the material must remain available afterwards. Review a representative passage in the final destination, including an important name, a speaker change and a meaningful sound where those occur.
That check is more useful than choosing whichever menu item sounds closest. A clear brief and a verified output help you decide whether you need captions, translated subtitles, a transcript or a combination.



