Veo 3 Dialogue and JSON Prompts: Examples With Sound

Veo 3.1 makes speech and sound from the prompt itself: put spoken lines in quotes after the speaker, and list sounds as plain phrases ("SFX: rain on the window"). A JSON-style prompt is the same information split into labelled fields (scene, camera, character, dialogue, audio), which keeps long prompts organised. It is not a special mode. On VdoBloom each Veo 3.1 clip is 8 seconds with sound and costs 25 credits on Veo 3.1 Fast or 100 on Veo 3.1 High Quality. No Google subscription is needed and new accounts get 10 free credits.

Updated October 31, 2026.

Quick facts

• Length: 8 seconds per clip, so keep dialogue to about 15 to 20 words.
• Aspect: 16:9 or 9:16 (no 1:1). See Veo 3 vertical video with dialogue.
• Speech: speaker + "says" + line in quotes.
• Sound: describe ambience and effects in plain words, or in an "audio" field.
• No subtitles: add "no subtitles, no on-screen text" or Veo may print the words.
• Cost: 25 credits (Fast) or 100 (High Quality). Failed renders are refunded.

Does Veo 3 need JSON prompts?

No. Veo reads plain sentences and JSON-style text the same way, as text. JSON helps you because each field reminds you to fill in camera, sound and dialogue, and you can swap one field without rewriting the rest. Keep it short: a few fields with one or two sentences each. Long nested JSON wastes words the model does not need.

Style	Use it when
Plain sentence	One speaker, one action, quick tests
JSON-style fields	Two speakers, timed beats, a series of clips with the same look

JSON-style prompt template

Copy this and replace the values:

{ "scene": "small bakery at 7am, flour on the counter, warm window light", "camera": "handheld medium shot, slow push-in", "character": "baker in her 40s, green apron, flour on her cheek", "action": "she lifts a tray of croissants toward the camera", "dialogue": "she says: "First batch of the day. Smell that?"", "audio": "oven hum, tray clatter, soft street noise outside", "style": "natural documentary look, no music", "avoid": "no subtitles, no on-screen text" }

Vlog and talking-to-camera prompts

1. A young man holds the phone at arm's length on a windy cliff and says, "Okay, I made it to the top. Look at this." Wind gusts on the mic, distant waves. Vertical selfie framing, no subtitles.
2. A woman in her car, parked, looks into the dashboard camera and says, "Quick update: the interview went really well." Rain taps on the roof, indicator ticking. No music.
3. A traveller walking through a night market turns the camera to herself and says, "This is the best thing I've eaten all week," then takes a bite. Crowd chatter, sizzling grills.
4. { "scene": "home office, morning", "camera": "static desk-level webcam angle", "character": "developer in a hoodie", "dialogue": "he says: "Day one of building my app. Let's go."", "audio": "keyboard clicks, coffee mug set down", "avoid": "no subtitles" }

ASMR and sound-first prompts

1. Extreme close-up of a knife slicing through a crisp red apple on a wooden board. Sound: sharp crunch, juice drip, knife tapping wood. No voice, no music.
2. Hands slowly kneading soft dough on a marble counter, flour dusting the air. Sound: soft squish and pat, quiet room tone.
3. Rain running down a window at night with a candle flickering inside. Sound: steady rain, distant thunder, faint crackle of the wick.
4. { "scene": "close-up of a pen on paper", "action": "a hand writes a short note in cursive", "audio": "pen scratch, paper rustle, clock ticking", "style": "warm macro, shallow focus", "avoid": "no voice, no music, no text" }

Ad and product prompts with voice

1. A barista slides an iced coffee across the counter toward the camera and says, "Cold brew, twenty hours. Worth the wait." Ice clinks, cafe chatter. Medium shot, slow push-in.
2. A runner stops on a bridge at sunrise, checks her watch and says, "Five kilometres before work. Every day." Breathing, distant traffic, birds.
3. { "scene": "bright bathroom shelf", "camera": "slow slide right along the shelf", "character": "voice-over only, warm female voice", "dialogue": "voice-over says: "Three ingredients. Nothing else."", "audio": "soft water drip, light room echo", "avoid": "no on-screen text" }

For full ad scripts, see cinematic AI commercial prompts.

Two-person dialogue prompts

Name each speaker by what they look like and give them separate lines in order. Two short lines fit in 8 seconds; three is the limit.

1. In a diner booth, the man in the grey jacket says, "You told them?" The woman in the yellow sweater sighs and says, "I had to." Low diner noise, cutlery clinks. Static two-shot.
2. Two kids on a treehouse floor. The girl with braids whispers, "Did you hear that?" The boy with glasses whispers back, "It's just the wind." Creaking wood, wind in leaves.
3. { "scene": "TV-style street interview, sunny sidewalk", "camera": "handheld, interviewer's mic in frame", "dialogue": "interviewer says: "What's your secret to staying calm?" Older man in a cap laughs and says: "Never check my email before breakfast."", "audio": "street ambience, passing bus", "avoid": "no subtitles" }

Why does Veo 3 ignore my dialogue or add subtitles?

• Line too long. Over about 20 words gets cut off or rushed. Split it across two clips.
• No speaker named. "says" needs a clear subject, or the wrong person talks.
• Subtitles appear. Add "no subtitles, no on-screen text".
• Music drowns speech. Add "no music" or "quiet background".
• Accent or tone wrong. Describe the voice: "calm, low voice", "excited, fast".

Still stuck? Read the Veo 3.1 guide or AI video not following the prompt.

FAQ

How do I make Veo 3 characters talk?

Write the speaker, then "says", then the line in quotes: The chef looks at the camera and says, "This is the trick." Keep it under about 20 words for an 8-second clip.

Is a JSON prompt better for Veo 3?

Not by itself. It is a way to organise the same details. Use it for multi-speaker scenes or a series of clips that must match.

Can Veo 3 make ASMR videos?

Yes. Describe the close-up action and list the sounds, and add "no voice, no music" so only the effects play.

How do I stop Veo 3 adding subtitles?

Add "no subtitles, no on-screen text" at the end of the prompt or in an "avoid" field.

How much does a Veo 3 clip with sound cost?

On VdoBloom, 25 credits for Veo 3.1 Fast and 100 for Veo 3.1 High Quality, 8 seconds each with audio. No subscription; one-time pack credits never expire.