AI video with sound: native ambience, effects, music and when to mute
Models marked for sound in the AI Video Generator make the audio with the picture: rain that matches the window, traffic that matches the street, music that follows the cut. Here is how to ask for it, which models have it, and when a silent clip is the better choice.
· 5 min read · 1K views · by the Cortex Digital Hub team
Pick a model marked for sound, leave the sound toggle on, and write the ambience into the same sentence as the motion. The credit total shows before you press Generate. From ₹24 a clip · no watermark · runs in your browser
Make a video → → Opens AI Video Generator · runs in your browser
A clip of a Chennai street at dusk is half a clip without the scooters, the horns and the vendor calling out. Models marked "sound" in the AI Video Generator make the audio with the picture instead, so the rain starts when the clouds open and the sizzle begins when the dosa hits the pan. This post covers what native sound does, how to prompt for it, which models have it, and when a muted clip is the smarter choice.
What native sound means on a model card
When a model is marked for sound, it generates ambience, effects and music that match what moves on screen. A dolly through a wedding hall in Hyderabad gets crowd murmur and a band in the distance; a slow orbit of a pickle jar gets kitchen room tone and a light musical bed. The audio is produced with the frames, not laid on afterwards, which is why it lines up with the motion.
Two limits are worth knowing. First, there is no voice-over unless the prompt asks for speech, and only Lumen Motion 2.6 is built for dialogue; the other sound models make ambience, effects and music. Second, the model decides the mix. You can ask for "quiet" or "loud traffic", but there is no volume slider.
Which models have sound
As of October 2026, the models marked for sound are Lumen Motion 3 and Lumen Motion 2.6, Vela 2.5 and Vela 2.0, Orbit 3, Orbit Prime and Orbit 2.6, Pulse Fast and Pulse Pro, Prism 6, Lumen Frames and Director 4. That covers the popular picks: Lumen Motion 3 as the all-rounder from 115 credits for 5 seconds, Orbit 3 as the best-value draft from 70 credits, and Vela 2.5 for cinematic 30-second clips.
Nova 2K and Nova Lite are not marked for sound, so plan on a silent clip from them; the same goes for Lumen Motion 2.5, Orbit 2.7, Lumen Omni, Canter, Spark and Morph. The sound toggle sits with the other options next to the prompt box.
How to write sound into the prompt
Put the sound in the same sentence as the motion, as a thing that happens, not a request for audio. "Rain falls on a tin roof while the camera pushes in on a tea stall, steam rising" tells the model what to show and what to hear at once. "Add rain sound" tacked on at the end is weaker, with nothing on screen to tie it to.
Be specific with nouns. "Traffic" could be anything; "two-wheelers and an auto horn on a Bangalore main road at 6 pm" gives the model a texture to match. Name the surface a sound comes from (tin, marble, tar road), the distance (a temple bell far off, a pressure cooker in the next room) and the energy (a lazy afternoon, a busy morning). For music, describe the mood and the instruments rather than a genre label: "a slow sitar line under the shot" beats "Indian music". For dialogue on Lumen Motion 2.6, put the line in quotation marks and say who says it: a woman at the counter says, "Two masala chai, less sugar."
Keep it simple in a short clip: one ambience and one effect is plenty in five seconds; three compete and the model muddles them.
Sound-prompt phrases that work
- Steady rain on a tin roof, distant thunder, water dripping from the edge
- Two-wheelers and auto horns on a busy main road, a bus pulling away
- A temple bell ringing slowly, birds in the courtyard, quiet otherwise
- Dosa batter sizzling on a hot tawa, a ladle scraping, kitchen room tone
- Waves on the Marina beach at night, a light wind, no music
- A ceiling fan humming, a pressure cooker whistle in the next room
- Wedding band playing in the distance, crowd murmur, a child laughing
- Soft piano under the shot, slow and warm, no drums
- A packet tearing open and a single crunch, then silence
- Footsteps on a marble floor echoing in an empty flat
Use the Avoid box for sounds you do not want: "no music", "no voices", "no wind noise".
When to turn sound off
Reels and Shorts are the big one. If you will add a trending track inside Instagram or YouTube, a clip with its own baked-in music fights the one you choose. Turn the toggle off and add the track in the app. The same applies when a voice-over is coming later: a clean silent clip gives your editor a quiet bed to work on.
Turn it off, too, for product listing videos and WhatsApp catalogues, where almost everyone watches muted, and for anything that ends up as a GIF, since the video to GIF step drops audio altogether. For a video made from several clips, keep the toggle the same for every clip; room tone next to silence sounds like a cut even when the pictures match.
Audio references and music rights
The References area accepts audio as well as images and video. Drop an audio file and the model keeps to it, which is how to carry a brand jingle or a rhythm across a set of clips instead of getting a fresh musical idea each time. Trim the reference to the useful part first.
On rights, the generator does not add songs; it makes its own ambience and music. If you drop a commercial track as a reference, or lay one over the finished clip, the usual rules apply: a copyrighted song in an advertisement generally needs permission, and each platform treats music in business posts differently. As of 2026, check the platform's help page before publishing a paid promotion with a track you did not make. A model-generated bed, or a jingle you own, avoids the question.
Check the sound before you export
Download the clip within seven days and play it with headphones once. If the ambience is right but the clip runs long, the Trim Video tool cuts without re-encoding, so the audio track stays exactly as generated. Converting to WebM for a website with the MP4 to WebM tool keeps the audio too, as does compressing for WhatsApp with compress video to 16 MB. The format guide for WhatsApp Status and Reels covers which container each platform prefers. Then open the AI Video Generator, pick a model with the sound mark, and put the rain in the same sentence as the camera move.
Frequently asked questions
Does AI video come with sound?
On models marked for sound, yes. The model generates ambience, effects and music that match the motion in the clip. Models without the mark, such as Nova 2K and Nova Lite, give you a silent clip.
Can the AI video speak a line of dialogue?
Only if the prompt asks for speech. Lumen Motion 2.6 is the model built for dialogue and sound; put the line in quotation marks and say who speaks it. Other sound models make ambience, effects and music, not a voice-over.
Should I turn sound off for Instagram Reels?
Usually. If you plan to add a trending track inside Instagram, a clip with its own music fights the one you pick. Turn the toggle off, or leave it on only when the ambience itself is the point of the Reel.
Can I use a popular song in my AI video?
The generator does not add songs; it makes its own ambience and music. Using a commercial song in an ad generally needs a licence, and as of 2026 each platform has its own rules for music in business posts, so check the platform's help page first.