How to Make ASMR Food Videos with AI

The glass fruit format is one of the most repeated shapes on short-form video: a familiar food rendered in a translucent material, poured onto a slice of toast and spread with a metal spoon, with every clink and scrape audible.

It is worth understanding why it works, because the reason explains every rule about how to make a good one.

Why the format holds attention

The appeal is a mismatch your brain cannot resolve. You know what a blueberry sounds like against a spoon, and glass sounds nothing like that. The audio keeps contradicting the shape on screen, and your attention stays on it trying to reconcile the two.

That is also why the audio cannot be replaced with music. The sound is not accompaniment, it is the content. This is why the ASMR Food template has background music disabled entirely rather than defaulted off.

Two other properties make it unusually well suited to short-form. It is completely narration-free, so it needs no script, no voice and no translation, and it travels across every language. And it is repeatable without going stale, because the format is fixed and only the subject changes, which is exactly the shape a series needs.

Picking a subject

The subject line has two parts: the food and the material. Both are required and neither can be vague.

Blue glass blueberries on toast works. Something satisfying does not, because there is nothing to render.

Choose translucent materials. Glass, crystal, amber resin, hard candy, jade, quartz and ice all transmit light, and the light play is half of what makes the shot watchable. Opaque materials like stone or metal lose it entirely.

Match the colour to the real food. Red crystal strawberries read instantly, because the colour confirms what the object is while the material contradicts it. A blue strawberry just reads as a mistake and the tension disappears.

Prefer small foods. Berries, grapes, citrus segments and cherry tomatoes shard and scatter convincingly under a spoon. A whole melon does not, because the format is built around spreading something across a slice of bread.

Subjects that consistently render well:

• Blue glass blueberries on toast
• Strawberries made of red crystal
• Mini kiwis carved from green jade
• Tiny orange slices made of amber resin
• Small watermelon made of pink quartz
• Grapes made of deep purple glass
• Cherry tomatoes in red hard candy

What the format already handles

You do not describe the camera, the lighting, the hands, the tabletop or the sound. All of that is written into the format brief and restated for every scene: a tabletop shot, a hand pouring from a small glass container, the hand leaving frame, then a metal spoon spreading the material across the bread. Diffused studio lighting, photorealistic, no camera movement, tactile ASMR audio.

Trying to specify those yourself works against the brief rather than adding to it. The single most common mistake with any template is writing a full prompt into a field that wants a subject.

If you do want to write complete prompts rather than subjects, the free-form animated format exists for that, and the prompt guide covers the five layers a full prompt needs.

Length and scene count

One scene is a complete video here. The pour, the spread and the settling all happen inside a single clip, so there is nothing left hanging at the end of five seconds.

Extra scenes give you variations rather than continuation. Three scenes is three different glass foods, which works as a small set, but past two or three the idea has landed and the extra seconds are cost without return.

Scenes run five to ten seconds each and you pick both numbers. Length is scene count times seconds, and so is the price.

Running it as a series

The reason to treat this as a series rather than a one-off is that the format does the recognition work for you. A viewer who has seen one of your videos recognises the second one in a frame, and that is worth more than any individual video performing well.

Keep the format identical and vary only the subject. Do not alternate between this and three other formats on the same channel in the first month, because you will not be able to tell what is working.

Build a subject list before you start. Fruit alone gives you twenty videos, and vegetables, sweets and small savoury items extend it a long way past that. A format with an obvious next subject is a format you can still be posting in six months.

When a clip does not land

Generative video misses sometimes. The usual failures here are the material reading as opaque, the food losing its recognisable shape, or the spoon motion looking wrong.

All three are usually fixed by making the subject line more specific about the material rather than about anything else. Deep blue translucent glass gives a different result from glass.

You can also regenerate a single scene in the editor, which buys exactly the seconds that scene originally used and leaves the rest of the video untouched. That is cheaper than regenerating the whole video, and it is the normal way to work with a format where one clip in five needs a second attempt.