Veo 3.1: dialogue and sound in one generation
Veo 3.1 generates the picture and the audio together — spoken dialogue, room tone and score — so a scene with a line of speech in it comes out finished in a single run.
Real results
Made in solorender with these models. Nothing retouched.
Best for
- ✓Any shot where somebody speaks. Write the line in the prompt and it is performed, lip-synced, in the clip
- ✓Emotional beats — a face lighting up, a gasp, a pause before a reply
- ✓Scenes that need a cut in the middle. It will cut, and the cut lands
- ✓Ambient sound and music as part of the same generation, not a separate pass
Where it struggles
- —Clips run up to 8 seconds. Plan scenes in 8-second units and assemble with Render Film
- —Veo 3.1 Fast costs well under half as much as the full model. Use Fast to find the shot and the full model only for the keeper
- —Audio is always on in solorender, which is why it costs what it costs — the silent tier is half the price but defeats the purpose of choosing Veo
Prompts that work
Each one opens the studio with the prompt already in the box. They are also one click away inside the app, under the prompt field.
A line of dialogue
Try it →An elderly man's face lights up with surprise as his family presents a birthday cake beside a family album on the table. Warm lighting fills the room. Gentle piano music plays softly in the background. He gasps and says with emotion, "You remembered."
Note the order: action, then light, then music, then the spoken line in quotes. The quoted words are what gets said.
Action beat
Try it →Cinematic dynamic scene starting close on the face of a man driving a minimalist vintage car, then zooming into the gear stick as he changes gear at high speed.
Two-hander
Try it →She turns from the window and says quietly, "They already know." The camera cuts to a man in the doorway who answers, with tension in his voice, "Then we move tonight." Low evening light, faint street noise outside.
Two speakers, two quoted lines, an explicit cut between them.
Mood with score
Try it →A slow push-in across an empty diner at night, rain streaking the window, neon sign flickering outside. Low strings build underneath. No dialogue.
"No dialogue" still gets you score and room tone — useful for an opener.
Put the spoken words in quotes
The single highest-value habit with Veo is to write dialogue as dialogue: the character, how they say it, then the words themselves inside quotation marks. "He says, with a worried tone, 'A huge storm is coming.'" comes out as a performance. "He warns about a storm" comes out as a man moving his mouth.
Accent and tone belong in the same sentence as the line, before the quotes. A northern British accent and a worried tone are both instructions the model acts on, and they change the delivery more than any camera direction will.
Find it on Fast, finish on full
Veo 3.1 Fast and Veo 3.1 answer the same prompt with the same understanding; Fast just spends less compute getting there. So the economical loop is to iterate the prompt on Fast until the beat and the delivery are right, then run that exact prompt once on the full model.
For a lot of social work the Fast output is the deliverable and the second run never happens. Check before you pay for it.
Run these for real
Every model in this guide is live in solorender. Credits never expire.
Open Veo 3.1 →