# Veo 3 Fast on eroq — the video engine that renders its own sound > Veo 3 Fast inside eroq — 8 second clips with native ambience, effects and dialogue at 720p or 1080p, 256 credits a clip, and how to write prompts for sound. Published 2026-09-08 · eroq.ai — canonical: https://eroq.ai/blog/veo-3-fast-native-audio Six of the seven video engines on eroq return a silent clip. Veo 3 Fast returns a clip with sound — ambience, effects and dialogue rendered with the picture. It is the only engine on the roster that does, and it changes how you write the prompt. The rest of the spec is rigid: exactly 8 seconds, 720p or 1080p, 256 credits a clip, Creator plan and up, SFW only. This guide is mostly about the writing. ## What "native sound" means Native means the soundtrack is generated with the image, from the same prompt, as part of the same render. There is no library of stock clips laid under the video afterward, and no separate audio step to pay for. You describe a shot; you get the shot and what it sounds like. Three layers come back: - **Ambience** — the room tone. Rain, traffic, a crowd, wind, the hum of a fridge. - **Effects** — sounds tied to things happening on screen. A door, footsteps, a brake squeal, a glass set down. - **Dialogue** — a spoken line from a person in the frame. One boundary to say plainly: eroq does not offer a lip-sync feature. Dialogue on Veo 3 Fast is a line the engine performs inside the shot, not a tool for matching a recorded voice to a face. Keep lines short and treat the result as a rendered take. ## The numbers - **Length**: 8 seconds, fixed. Not 5, not 10. - **Resolution**: 720p or 1080p. - **Price**: 256 credits per clip, about $2.56 at the entry pack, where a credit is roughly a cent. - **Formats**: 16:9, 9:16, 1:1. - **Modes**: text-to-video and image-to-video. - **Plan**: Creator ($49, 6,000 credits) and up. - **Content**: SFW only; the provider moderates upstream. Mature fiction is [Motion One](/models/eroq-motion-one)'s job, silently. A Creator month is about 23 Veo clips if you spend it on nothing else. Failed renders refund automatically and unused credits roll over. ## Writing for audio The visual half of the prompt follows the same register as every engine on eroq: one flowing paragraph — subject, motion, camera, light, mood — 30 to 80 words. Then you add the sound, and the useful habit is to write it as a second, shorter sentence at the end, in the same order as the layers above. **Ambience first.** Name the space and what it sounds like. "Steady rain on a station canopy." "Low chatter and an espresso machine." One or two elements — an 8-second clip can't hold a soundscape. **Effects tied to actions you already wrote.** If the visual says "the doors hiss open," the sound sentence can say "the door chime and her footsteps on wet concrete." Sound with no visible cause tends to wander; sound that lands on an action tends to land. **Dialogue in quotes, attributed.** One line, under ten words, with who says it. Eight seconds is one line, maybe two. Write the line the way the character would say it, not the way a script summarizes it. **Silence is a choice.** If you want no dialogue, don't leave a person's mouth idle in the prompt — give them something to do. If you want a quiet room, say "quiet" and name the one sound that is there. What to skip: music. The engine renders what is in the scene; if you want a score, that is an edit you make afterward. And don't stack adjectives on the sound ("epic, cinematic, immersive audio") — name sources instead. ## Example prompts Ambience and effects, no dialogue, 16:9: > A night train pulls into a rain-soaked platform, brakes squealing as the doors hiss open and a single passenger steps down under the yellow station lamps, pulling her coat closed against the wind. Wide static shot from the far end of the platform, neon noir palette, calm and lonely. Sound of steady rain on the canopy, the brake squeal, the door chime and her footsteps on wet concrete. Settings: 8 s, 16:9, 1080p, Camera move Static, Lighting Practicals, Palette Neon noir. One line of dialogue, 9:16: > A barista in a small, quiet coffee shop slides a cup across the counter to a customer, glances at the rain outside and says, "Third one today, and it's not even noon." Close-up on the counter, soft cross lighting, warm film palette, gentle and dry. Ambience of an espresso machine hissing, low chatter and rain against the window. Settings: 8 s, 9:16, Camera move Close-up, Lighting Soft cross, Palette Warm film. Both prompts put the sound sentence last and tie every effect to something visible. That is the whole method. ## Where Veo fits Eight seconds with sound is a specific shape, and a few jobs fit it exactly: - **Ads.** The Ads mode in the Video tool ships templates like "UGC testimonial" for TikTok/Reels 9:16, Feed 1:1 and YouTube/Web 16:9. A testimonial is one person, one line, room tone — an 8-second Veo clip, in other words. The [ad generator](/tools/ai-ad-generator) is built around this and [UGC ads](/use-cases/ugc-ads) are the obvious use. - **Scenes in a film.** Film mode builds a storyboard scene by scene, up to 12, each rendering as you add it. A Veo scene brings its own sound into the assembled player. The API side is [/docs/films](/docs/films). - **Product moments.** A bottle opened, a zipper pulled, a laptop closed — actions with a signature sound. Where it doesn't fit: anything that isn't 8 seconds long, anything that needs a seed or an end frame, and anything mature. For long or precisely controlled shots, Seedance 2.5 is the engine — [/compare/seedance-vs-veo-3](/compare/seedance-vs-veo-3) has the side-by-side. The engine page at [/models/veo-3-fast](/models/veo-3-fast) keeps the current spec. ## FAQ ### Does Veo 3 Fast on eroq generate audio? Yes. It is the only engine on the roster that renders a soundtrack — ambience, effects and dialogue — with the picture, from the same prompt. ### How long is a Veo 3 Fast clip? Exactly 8 seconds. There is no shorter or longer option on this engine; for other lengths, pick another engine from the roster. ### How much does a Veo 3 Fast clip cost? 256 credits per 8-second clip, about $2.56 at the entry pack. It is unlocked on the Creator plan and up, and failed renders are refunded automatically. Write the shot, add the sound sentence, and render it — [/studio/video](/studio/video).