UE5 Sequencer: Play a Voice Line in a Cutscene from a DataTable

The problem: a cutscene needs the right line, not a random one

Bark systems are easy. You filter a dialogue DataTable by context, pick a random row, and play it. Cutscenes are the opposite problem. When you want UE5 Sequencer to play a voice line in a cutscene, you need one exact clip on one exact frame, every time the scene plays, so the camera move, the subtitle and the lip flap all land together. Randomness is the enemy of cinematics.

This tutorial uses the Deity Dialogue Pack as the source library. It is a DataTable-driven NPC voiceover pack for a deity/war-god archetype, shipping a thunderous, commanding voice performance plus cinematic divine narration that suits gods, oracles and otherworldly bosses addressing the player from on high. Its 566 voice lines live in a 'DT_Dialogue' DataTable where each row carries a 'VoiceAudio' reference, which is exactly the shape you want for a deterministic cutscene pull.

The technique below works for any pack in the same collection because they share the same row schema, so once you have wired one cutscene you can reuse the pattern for every speaker in your game.

Step 1 - Pull a specific line by row name

Every row in 'DT_Dialogue' has a row key, an FName, that uniquely identifies it. For a cutscene you do not search by context tag; you address the line you have already chosen by its known row name, so the result is deterministic.

1. Open your dialogue DataTable from the Content Browser and find the line you want for the scene. Note its row name (the leftmost 'Row Name' column).

2. In your cutscene setup Blueprint, add a 'Get Data Table Row' node. Set its 'Data Table' pin to the deity 'DT_Dialogue' asset and type the row name into the 'Row Name' field.

3. From the 'Out Row' struct, break out the 'VoiceAudio' member. In this collection 'VoiceAudio' is a soft object pointer to a USoundWave, which means the clip is not in memory until you ask for it.

4. Add a 'Load Synchronous' node off the 'VoiceAudio' soft reference. Its return value is the resolved USoundWave you will hand to Sequencer. Because cutscene lines are intentional and few, a synchronous load on the frame you build the scene is acceptable; for long lines you can instead async-load ahead of the cut.

Doing the lookup by FName keeps the cutscene stable across content updates as long as the row name persists, which is far safer than relying on a row index or a random pick.

Step 2 - Add the SoundWave to a Sequencer audio track

Sequencer drives cutscenes from a Level Sequence, and audio lives on an Audio track. You can author this entirely in the editor for a fixed scene, or build it at runtime if your cutscene is data-driven.

For a fixed, hand-authored cutscene, the simplest path is to open your Level Sequence, click the green 'Track' button, choose 'Audio Track', then on that track click the '+' and pick the resolved USoundWave from the asset list. Drag the resulting audio section so its start frame matches the moment the deity should speak. Because these lines are one-shot with no loop, the section length equals the clip length and you do not need to worry about loop seams.

For a runtime or templated cutscene, build the audio section from the USoundWave you loaded in Step 1. Get the Level Sequence's movie scene, add a movie scene audio track, add an audio section to it, and set the section's sound on the 'Sound' property using your loaded SoundWave. Set the section range so it begins on your chosen frame. This is the same data the editor writes when you drop a clip by hand, so the two approaches are interchangeable.

If the line is diegetic, for example a war-god speaking from a physical altar in the scene, attach the audio to the altar actor so it plays in 3D with attenuation. For a non-diegetic 'voice of god' narration over a prologue, leave it as a 2D master-track sound and skip attenuation entirely so the proclamation sits cleanly over the mix.

Preview by scrubbing the Sequencer playhead. The line should fire exactly when the playhead crosses the section start, which is the determinism a cutscene needs.

Step 3 - Sync subtitles to the clip

A cinematic line needs on-screen text, and the dialogue row already carries it. The 'ResponseText' field on the same row is the spoken line as written, so you never have to retype or risk drift between audio and caption.

The cleanest sync is to tie the subtitle to the audio section's lifetime rather than to a separate timer. In Sequencer, add an Event track and place a key on the frame the audio section starts; fire an event that pushes 'ResponseText' to your subtitle widget. Place a second key on the audio section's end frame that clears the subtitle. Because the section's start and end are the clip's true bounds, the caption appears and disappears with the voice with no magic numbers.

If you prefer to derive timing from the audio itself, read the duration of the loaded USoundWave and hold the subtitle for that length. Either way, pull the caption text straight from 'ResponseText' on the row you fetched in Step 1 so the words on screen are guaranteed to be the words being spoken.

For the longer divine proclamations, lean on the pack's larger length tiers; those LG and XL lines are written as multi-sentence narration, so consider breaking 'ResponseText' across a couple of subtitle cards keyed to natural pauses rather than dumping a paragraph on screen at once.

Step 4 - Reuse the line library across cutscenes

The reason to drive cutscenes from a DataTable instead of dragging loose WAVs is reuse. Once one cutscene addresses lines by row name and pulls captions from 'ResponseText', every future cutscene is the same three moves: pick a row name, load 'VoiceAudio', key the audio and subtitle into Sequencer.

Wrap Steps 1 to 3 in a small helper, for example a Blueprint function or a C++ utility that takes a DataTable and a row name and returns the loaded USoundWave plus the caption text. Call it from every cutscene builder so the lookup logic lives in one place; if the row schema ever changes you fix it once.

The same helper works across the whole collection, because the deity pack shares its 'DT_Dialogue' schema with the other Lore Packs. Point the same code at the Bard Dialogue Pack for a theatrical quest-giver, at the free Assassin Dialogue Lore Pack for a rogue, or migrate a single archetype out of the Fantasy NPC Voices complete bundle, and your cutscene pipeline does not change at all.

Performance-wise, cache the DataTable reference rather than reloading it per cutscene, and pre-load the specific 'VoiceAudio' clips a scene needs just before the cut so the synchronous load never stalls a frame mid-cinematic.

DT_Dialogue fields you use in a cutscene

The row schema is shared across the collection, so the same field names apply to every pack referenced here.

FAQ