How to make a generated ASMR video with convincing sound

Last updated 2026-08-16 · 6 min read

What is the short version?

ASMR is an audio format that happens to have pictures. The visuals are the easy part now; what separates a satisfying clip from a dead one is a sound bed with no gaps in it, and gaps are exactly what generated footage leaves behind.

1. Go to Vizard Agent and describe the action and the material.
2. Say how close and how slow the shots should be.
3. Say the length and the feed.

What do you need before you start?

A description. There is no footage to supply, so the brief is everything — and the details that matter are the ones that determine sound: what is being crushed, what it is made of, and how slowly it happens.

• The action. A press crushing something, a knife through something, liquid pouring.
• The material. Glass, candy, foam, ice. The material is the sound.
• The pace. Slow and close is the convention. Quick cutting kills the effect.
• The length. Forty seconds is a working size for a feed clip.
• The shape. Vertical, since this format lives on phones.

What do you type into Vizard Agent?

Describe the sound as much as the picture. Vizard Agent generates both, and naming the material and the moment of failure — the crack, the shatter, the give — is what tells it where the payoff of the whole clip sits.

Prompt

Variants worth knowing:

• No music at all. Standard for the genre; the sound design carries it.
• A loop. Ask for the last frame to match the first if it is going somewhere that repeats.
• Higher resolution. Say so and Vizard Agent upscales the finished cut rather than generating larger throughout.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real ASMR clip. Note how much of it happens after the video exists: most of the run is listening to the result and fixing what the ear notices rather than what the eye does.

1. Loads the all-generated film playbook and reads the generation options.
2. Generates a reference still of the setup and looks at it before shooting anything.
3. Generates the five crush shots and checks the clips.
4. Extracts review frames from every shot and reviews them one by one.
5. Measures the native audio and builds a rough cut, then listens through it for sound and sync.
6. Generates a hum bed and a squelch layer, mixes and renders.
7. Checks the waveform, raises the room tone, and tests an upscale against the original.
8. Measures the quiet stretch after the shatter and inspects what is happening there.
9. Builds a grinding layer to fill it, tracks down the room tone that went missing in the re-mix, and finalises the loudness.

Step 8 is the craft. Vizard Agent noticed a hole in the sound after the loudest moment, measured exactly how long it ran, and generated a layer to fill it — which is the difference between a clip people watch twice and one they scroll past.

What does the result look like?

From the run this page is written from, probed on the delivered file: 720x1280, H.264, 24fps, 40.31 seconds, AAC audio. Vertical, forty seconds, upscaled from the generated shots, with a layered sound bed — room tone, hum, the crush itself and a grind underneath the quiet passage.

Ask for a longer version and Vizard Agent generates more shots against the same reference still rather than stretching the ones it has.

Forty seconds is five held shots. Each one runs about eight seconds, which is long enough for the sound to develop and short enough that the generated picture does not start to wander.

Expect tens of minutes rather than minutes. This category was not separately measured; comparable work runs a median of 28 to 38 minutes end to end, and across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

Generated ASMR is a genre with an honesty problem: the sounds are made, not recorded, and viewers of this format are unusually attentive to exactly that. Vizard Agent builds a convincing bed, and convincing is not the same as real.

• The sound is designed, not captured. If your audience believes they are hearing a real object, that is worth being straight about.
• Materials behave oddly in generated video. Glass shattering and liquids are where the picture betrays itself.
• Fast cutting destroys the effect. This format needs long held shots, which is also where generated video is weakest.
• Loudness matters more than usual. These are watched on headphones. Vizard Agent finalises the loudness, and it is worth listening on headphones yourself.
• Real recording still wins. A phone and a quiet room will beat generation for anything you can actually film.
• The genre rewards patience. A clip that rushes to the payoff misses the point; the anticipation is most of the effect.

How do you fix a result that came back wrong?

Say what you heard. Vizard Agent keeps the generated shots, each audio layer and the mix graph separately, so raising the room tone or replacing one sound layer is a re-mix rather than a regeneration of any footage.

• "It goes quiet in the middle." A layer built to fill the measured gap.
• "Too loud overall." Loudness re-finalised across the cut.
• "The third shot looks wrong." That shot regenerated, the audio kept.

How does Vizard Agent compare to doing it yourself?

By hand this is a camera, a rig, the actual object, and a microphone good enough to justify the format — or a generation tool that gives you silent clips and leaves the entire sound design to you.

Common questions

Does it generate the sound too? Yes. Vizard Agent builds the layers — room tone, the action, the texture underneath — and mixes them.

Can it match a real recording? For hard materials it gets close. For anything you can film yourself, filming is still better.

How long should an ASMR clip be? Around forty seconds. The format needs held shots, so more length means more generated footage.

Can I get it at higher resolution? Yes. Vizard Agent upscales the finished cut and compares the detail against the original first.

Should it loop? It helps on feeds that repeat. Ask and Vizard Agent matches the ending to the opening.

Can I use my own footage instead? Yes. Upload it and Vizard Agent does the sound design over your material.

Should there be music? Usually not. The genre works on sound design alone, and Vizard Agent builds the layers rather than scoring over them.

Can it make a series of them? Yes. Keep them in one project and Vizard Agent reuses the same setup and reference still, so the clips read as one series rather than as unrelated videos.