How Should I Caption a Product Demonstration When Two People Speak at Once?

Separate simultaneous speech from a long sentence

Automatic captions often organize speech into a single chronological stream. That can become confusing when two people talk at once. A demonstrator might explain how a latch works while a second speaker asks whether an accessory is included. Both statements can matter to a shopper.

W3C's prerecorded-caption guidance establishes that captions should convey speech and identify speakers where needed. It does not prescribe one universal layout for every overlapping conversation. The editor must assess the final audio and the destination's caption capabilities.

First listen to the overlap without reading a transcript. Identify each speaker and the exact words that can be understood. Do not invent a clean sentence from indistinct speech. If an important phrase is unintelligible even in the source recording, flag the audio production problem rather than fabricating a caption.

Record whether the overlap is intentional, such as a natural interruption, or an editing mistake. That decision affects whether the best repair belongs in caption formatting or in the soundtrack itself.

Preserve timing and attribution

Mark the beginning and end of each speaker's utterance independently. If both speak from 3.0 to 3.6 seconds, do not automatically move one person's words to 3.6–4.2 seconds merely to avoid visual overlap. That would imply a different conversational sequence.

Use a supported caption format or player presentation that distinguishes the speakers and accommodates simultaneous information. WebVTT's specification permits overlapping cues, but the final rendering and positioning must be checked in the destination. Different players may present overlapping cues differently.

If the display becomes crowded, determine whether the original audio needs editorial repair. A revised mix or approved rerecording may be more intelligible than attempting to show several dense caption lines simultaneously. Preserve the actual meaning and conversational timing when making that decision.

Do not delete a product exclusion simply because it is spoken underneath another sentence. Such a change can materially alter the message.

Hypothetical case: an inclusion question overlaps an instruction

Hypothetical illustration; no caption player or audio was tested. A fictional storage-box demonstration contains two speakers. The Demonstrator says 'Slide the tray into the upper slot' from 2.8 to 4.0 seconds. The Narrator asks 'Is the tray included?' from 3.2 to 3.8 seconds.

Proposed version A places the entire question after 4.0 seconds, making it appear to follow the instruction. That may be easier to display, but it misrepresents the stipulated overlap. Proposed version B preserves both original intervals and uses distinguishable speaker cues, subject to the player's actual layout and readability.

If the two cues cover the tray action, the editor should evaluate a different supported caption placement or a revised audio edit. The worksheet describes possible treatments, not a tested accessible result.

Approve the actual conversation and product message

Play the final video with captions on and audio muted. Ask whether a viewer can identify who gives the instruction, who asks the question and whether the question was answered. Then compare against the final audio.

Check whether simultaneous text covers an essential control or product movement. If the player cannot render a readable arrangement, escalate the delivery design rather than silently omitting words.

The existing caption proofing guide covers transcription accuracy. This guide addresses the separate timing and presentation problem of two intelligible utterances occupying the same interval.

Record your own observations

The filled rows are illustrative examples. Use the blank row for your own files and observations.

Download the editable CSV worksheet

What the references establish

• W3C WAI — Understanding Captions (Prerecorded): Establishes the need to convey meaningful speech and speaker identification.
• W3C — WebVTT: The Web Video Text Tracks Format: Defines cue timing and explicitly permits overlapping cues; does not guarantee identical presentation in all players.