VEO 3.1: write the sound into the prompt (4s, 6s, 8s, 4K, 40cr)
VEO 3.1 has native audio. A silent prompt still gets a soundtrack you did not choose.
VEO 3.1 has native audio. A silent prompt still gets a soundtrack you did not choose.
VEO 3.1 is Google's enhanced VEO 3 row: text-to-video, no still required, 4K ceiling, clips of 4s, 6s, or 8s, 40 credits. The catalog line is "Enhanced VEO 3 with improved quality and features." None of that is a mute switch. Audio is on. If the prompt never names a voice, a room, or a hit, the model still writes one.
Native audio is already on
Native audio means picture and soundtrack leave in the same pass. You do not get a silent master as the default. You get a clip whose air, footsteps, and maybe dialogue were decided at generate time. The AI video generator will run that pass. It will not invent the mix brief for you.
Write the sound the way you write the lens. Quote the line. Name the room tone. Say whether the street is wet enough to hiss. A picture-only sentence is how a "quiet product hero" comes back with a trailer score you then have to fight in the edit. Dialogue and audio prompting is the longer version of that rule.
4, 6, or 8 seconds — not a slider
Duration is a closed set: 4s, 6s, 8s. There is no 5-second step and no 10-second hold on this record. If the line needs a breath you cannot fit in eight seconds, this is the wrong row, not a prompt problem. Pick the length before you spend the 40 credits. Trimming a padded eight-second take is cheaper than pretending the catalog hid a twelve-second option.
Aspect is 16:9, 9:16, or auto. Quality settings include 720p, 1080p, and 4K. Styles listed are Cinematic, Documentary, and Abstract. Those are the knobs. They do not include "silent."
The 40-credit 4K clip
Forty credits is the catalog price for this generate. It is a short cinematic clip with audio baked in — not a minute of coverage, not a talking-head pipeline, not a music-video renderer you plan to mute. Use it when the shot is a scene: physics, light, a line that belongs in the picture.
If you need a longer single pass, leave this row. If you need a product that must match a locked still, this is not reference-to-video. Captions still happen after. The complete VEO 3.1 guide is the map. This page is only the audio rule.
Write the room, write the line
A useful VEO 3.1 prompt names who is in frame, what they do inside four, six, or eight seconds, the camera, the light, the words in quotes if anyone speaks, and the air: HVAC, rain, crowd, none. Leave any of those out and the model fills them. Filling them is the feature. It is also how you pay 40 credits for a soundtrack you throw away.
Google's other video rows are a different brief. This one is enhanced VEO 3, 4K, native audio, 40 credits. Stay on it when that is the job.
FAQ
If I omit sound from the prompt, is the clip silent?
No. Native audio is on. A prompt that never mentions sound still returns a soundtrack the model chose. Write the mix, or plan to replace it.
Can I generate a 10-second VEO 3.1 clip?
Not on this catalog record. Supported durations are 4s, 6s, and 8s. Ten seconds is a different model, not a hidden setting.
Do I need an image to start?
No. This row does not require an image. It is text-to-video. A still is a different row if the plate must survive.
What does 4K mean here?
Max output resolution is 4K, with 720p and 1080p also listed. The 40 credits are for this generate, including the audio pass. 4K does not turn the clip into a silent master.