Reel Transcript Guide · Published 26 October 2026

How to Keep Reported Speech Separate from the Current Speaker in an Instagram Transcript

Keep the person who is actually speaking as the speaker of the turn, even when they recount someone else’s words. Preserve the wording that introduces the reported speech, and do not create a new live speaker turn just because a name or quotation appears. Check the recording before deciding where the quotation begins and ends; keep any uncertain attribution visible in a separate review note.

Watch the silent video guide: reported speech

A quoted person is not automatically a new voice

A speaker can describe what a colleague said without that colleague appearing in the recording. The transcript needs to represent the words that were actually spoken. A label for a live voice and an attribution inside a story do different jobs. W3C’s transcription guidance addresses accurate spoken wording and relevant speaker identification. The worksheet below is our original editorial method for applying those principles to reported speech; it is not an official notation standard or a claim that a tool can identify an absent person.

A fictional attribution trap

Imagine one person saying, “Maya told me, send the draft on Friday. I asked her to confirm the date.” These words are invented for teaching; we did not obtain them from a Reel or test a service with them. Creating a separate “Maya:” turn would suggest that Maya’s voice is present, although this example contains only one current speaker. Removing “Maya told me” would also remove the spoken attribution. Preserve that introduction while reviewing the actual wording and punctuation.

Our current-voice review workflow

1. Locate the passage: save the transcript version and a usable source location before editing. A name in the text is a search clue, not evidence of a voice change.
2. Review the current voice: replay the relevant passage and its surrounding context. Use only speaker identities supported by your evidence; keep an unknown speaker label when identity is unresolved.
3. Retain the spoken introduction: keep phrases such as “she said” or “I told him” when they occur. Do not silently rewrite indirect speech as a direct quotation.
4. Check the return boundary: establish where the current speaker resumes their own account. Punctuation is an editorial decision to review, not audible proof of a quotation boundary.
5. Record uncertainty separately: write a review note if the wording, attribution or boundary remains unclear. Keep the note distinct from spoken text so another reviewer can see what still needs checking.

When a real second voice appears

A Reel may cut from a person recounting an exchange to a recording of another person. Review the actual source transition rather than assuming that every quoted sentence has the same status. A second audible speaker may need a separate turn; a name mentioned by the narrator alone does not establish that turn. If an edit or overlapping audio prevents a confident decision, preserve the uncertainty and retain the source reference. Do not invent a voice change to make the transcript look tidy.

Keep a summary separate from the spoken account

For working notes, you might write a summary such as “The narrator reports an instruction attributed to Maya.” Label that as your summary rather than replacing the transcript with it. The recording establishes what the current speaker said; it does not by itself verify that the recounted event happened exactly as described. Our Instant Reel Transcript tool returns text from supported public links. It does not independently verify reported events, identify absent people or certify quotation boundaries.

Questions about reported speech

Should every quoted name become a speaker label?

No. First establish whether that person actually speaks in the recording. Preserve names and attribution inside the current turn when they are part of the spoken account.

Can I turn “she said she would return” into “I will return”?

Do not present newly written direct speech as words spoken in the recording. Keep the original wording in the transcript and put any paraphrase in a clearly separated summary.

What if I cannot find where the reported words end?

Keep an open review note with the source location. Do not silently assign the rest of the passage to the quoted person.

Sources and scope

Checked: 26 October 2026. Primary guidance: W3C WAI — Transcribing Audio to Text, on accurate spoken wording and relevant speaker identification. The R01 example, three-layer card and review workflow are original editorial guidance, not official punctuation rules, a legal opinion or an accuracy benchmark.