AI ASMR video: 10 glass and fruit cutting prompts and how to score them
An AI ASMR video is a short clip, often a knife slicing a glass-like fruit, where the model makes the picture and the sound together. Veo 3.1 makes 4-, 6- or 8-second clips (Google Cloud). Write the sound into the prompt, check that each cut and its sound land together, and budget for several tries per keeper.
Key facts
| Veo 3.1 clip length | 4, 6 or 8 seconds (this page prices 8)[3] |
|---|---|
| Veo native audio | Sound effects, ambient noise and dialogue generated with the video[1] |
| Veo 3.1 on Runway API | 40 credits/s with audio, 15 for Veo 3.1 Fast with audio (1 credit = $0.01)[2] |
| Kling 3.0 Standard on Runway API | 13 credits/s with audio, 9 without[2] |
| YouTube disclosure rule | Realistic AI scenes that did not happen must be disclosed[5] |
In AI ASMR the sound is half the product, and a cut that looks right but sounds wrong gets scrolled past. A prompt gets you a take. A score sheet tells you which take to post.
This page gives you ten prompts, a four-part score sheet and the cost of a usable clip. Prices come from Runway's public rate page as of 2026-10-26.
What is an AI ASMR video?
An AI ASMR video is a short generated clip built around a satisfying sound, usually a knife slicing a glass-like fruit or object.
The trend took off in June 2025, when Veo 3 made picture and sound in one step. Fast Company reported that one creator started posting on 10 June and reached nearly 200,000 followers within days. Google says Veo generates sound effects, ambient noise and dialogue natively. ByteDance describes Seedance 2.0 as using audio-video joint generation. Kling 3.0 is priced with an audio option on Runway.
How do you make an AI ASMR video?
You make an AI ASMR video by writing one object, one action, one camera angle and the exact sounds into a single prompt on a model that makes audio.
Use this order:
- Object. Name the material and the shape: "a translucent glass strawberry".
- Surface. A wooden board or marble slab gives the knife something to tap.
- Camera. Close-up, fixed, shallow depth of field. Moving cameras make cuts harder to read.
- Action. One or two cuts, in order. Tech Tactician's Veo 3 guide reported a second cut landing at an unexpected spot in one result, so fewer cuts is safer.
- Sound. List each sound in the order it should happen. End with "no music, no voice".
PiAPI describes a different route: make a still, animate it with Kling, then add sound with a separate video-to-audio model. That gives you control of the sound, and the sync check is yours to run.
What are good AI glass cutting ASMR prompts?
A good glass cutting prompt names the glass, the blade, the cut order and the sound of each step.
These are original templates. Replace the bracketed words.
- Close-up of a translucent glass [strawberry] on a wooden board. A steel knife slices it in half in one slow motion. Sounds: a soft glass tap, then one crisp crack. No music, no voice.
- Macro shot, fixed camera. A clear glass [apple] on marble. The knife presses in, the glass resists, then splits. Sounds: slow squeak of pressure, one sharp snap.
- A glass [lemon] with visible internal ridges. Two cuts, left to right. Sounds: two clean cracks, 1 second apart.
- A glass [mango] cut into cubes on a wooden board. Sounds: knife on board, small clinks as cubes fall.
- A hand slices a glass [peach] in a bright kitchen. Sounds: room tone, one clean crack, a soft rock of the halves on the board.
What are good AI fruit cutting ASMR prompts?
A good fruit cutting prompt describes the real fruit's texture and the sound it should make.
- Close-up of a ripe [watermelon] on a board. A large knife cuts straight down. Sounds: dense crunch, then a wet release as the halves part.
- A fixed shot of a [pomegranate] cut open with a single slice. Sounds: knife through skin, a short rustle of seeds.
- A [kiwi] sliced into thin rounds, three slices. Sounds: three soft, even slices on a board.
- A frozen [orange] slice, thick ice on the rind. Sounds: a dry crunch, then a faint crackle.
- Loop prompt: a clear glass [pear] whole at the start and whole again at the end, with one cut in the middle that heals. Sounds: one crack in the middle, silence at both ends.
Prompt 10 is the loop test described below.
How do you score an AI ASMR clip?
You score an AI ASMR clip on four checks: sound sync, cut physics, seam loop and cost. Give each of the first three 0, 1 or 2 points.
| Check | 0 points | 1 point | 2 points |
|---|---|---|---|
| Sound sync | Sound is early, late or missing | Sound is close but you can hear drift | Each sound lands on its cut |
| Cut physics | Blade passes through without a visible split, or the object changes shape | Split is visible but the pieces move oddly | Blade, split and pieces behave like one event |
| Seam loop | Visible jump or audio click at the seam | Small jump | Clean seam, no click |
Keep a take only at 5 or 6 points. A take at 3 or under is a retry.
How to check sound sync
Play the clip twice. First with sound only, tapping each cut you hear. Then with picture only, tapping each cut you see. The two sets of taps should line up. Anything off by more than a few frames feels wrong on a headset.
How to check cut physics
Scrub through the cut frame by frame. Look for the blade passing through material with no split, pieces that grow or shrink, and a second cut that does not match the first. Tech Tactician reported that last fault in a Veo 3 result.
How to run a seam loop test
Loop the clip five times on repeat with headphones on. Watch for a jump in the object or the knife at the seam, and listen for a click or a sound that cuts off. A clip that ends in silence loops best. Veo's Google Cloud docs list a first and last frame mode: give it the same image twice to aim for a matching start and end, then run the seam test.
How much does an AI ASMR clip cost?
An 8-second ASMR clip costs $3.20 on Veo 3.1 with audio, $1.20 on Veo 3.1 Fast and $1.04 on Kling 3.0 Standard, at Runway API list prices as of 2026-10-26.
Runway's pricing page sets one credit at $0.01. Veo 3.1 supports 4-, 6- and 8-second clips. The table uses 8 seconds for all three so the sums match.
| Model on Runway | Credits per second (with audio) | 8-second clip |
|---|---|---|
| Veo 3.1 | 40 | $3.20 |
| Veo 3.1 Fast | 15 | $1.20 |
| Kling 3.0 Standard | 13 | $1.04 |
Cost per usable clip is the clip price divided by your keep rate. Two example keep rates:
| Model | If 1 in 3 is usable | If 1 in 5 is usable |
|---|---|---|
| Veo 3.1 | $9.60 | $16.00 |
| Veo 3.1 Fast | $3.60 | $6.00 |
| Kling 3.0 Standard | $3.12 | $5.20 |
Run a batch of five on the cheapest model first, score each take, and use your real keep rate in the sum. For wider price comparisons see Google Veo 3 pricing and Kling AI pricing.
What can go wrong with AI ASMR videos?
The main faults are sound that does not match the cut, impossible physics, and rules for realistic content.
- Preview shutdown. The Veo 3.1 preview models in the Gemini API shut down on 22 October 2026. Older prompt guides that point to them may no longer work.
- Labels. YouTube requires disclosure for realistic AI scenes that did not happen. A glass fruit is clearly not real. A realistic kitchen is less clear.
- Real people and brands. Keep hands generic and leave out logos and copyrighted characters. Get permission before recreating an identifiable person's likeness.
- Silent output. If you turn audio off to save credits, the Veo no-audio guide covers what changes. To draft prompts faster, try the AI video prompt generator.
Questions
How do you make an AI glass cutting ASMR video?
Pick a model that makes audio with the video, describe a glass-like object on a board, one knife, a camera angle and the exact sounds. Ask for no music and no voice. Generate several takes and keep the one where the cut and the sound happen together.
Why does the sound in my AI ASMR clip not match the cut?
Joint audio-video generation does not guarantee that every sound matches the visible action. Name the order of events in the prompt, keep the clip to one or two cuts, and reject takes where the sound is early, late or missing.
Do I need to label AI ASMR videos?
On YouTube, realistic scenes that did not happen must be disclosed. A glass strawberry is obviously impossible; a realistic kitchen scene is the case to label.
How much does one usable AI ASMR clip cost?
At Runway API list prices as of 2026-10-26, one 8-second clip costs $3.20 on Veo 3.1 with audio, $1.20 on Veo 3.1 Fast and $1.04 on Kling 3.0 Standard. Divide by your keep rate to get the cost of a usable clip.
Sources
- Veo
- API pricing
- Veo 3.1 Generate
- Generate videos with Veo
- Disclosing use of altered or synthetic content
- Seedance 2.0
- How To Make Veo 3 ASMR Fruit Cutting Videos (Prompts Included)
- Glass Fruit Cutting ASMR AI Prompt (How I Made a Video Without Google Veo 3)
- The rise of AI-generated ASMR: exploring TikTok's new sensation
Related guides
- Google Veo 3 pricing: what one clip costs by model and routeVeo 3.1 costs $0.03 to $0.60 per second on Google's cloud API. See the price of one 8-second clip for each model, with or without audio, and by route.
- Veo 3 not generating audio? 7 causes and how to fix themVeo 3.1 makes sound, so a silent clip has a cause: the audio toggle, the prompt, or an edit mode with no audio. Seven documented causes and the fix for each.
- AI video prompt generator: fill-in templates for Veo, Kling and SeedanceA no-signup AI video prompt generator as fill-in templates for Veo, Kling and Seedance: each maker's own syntax, one idea built four ways, checked Nov 2026.
- Kling AI pricing explained: what one usable clip costs in 2026Kling AI pricing turned into dollars per 5s and 10s clip, by resolution and audio, on every plan, plus what retries add. Checked 26 Oct 2026.
More on Viral AI videos
- How to make Italian brainrot AI videos: character, prompts, voice and costHow to make an AI Italian brainrot video: original character, image and video prompts, Italian voice and a 15-second cost sum from $4.31 at list prices.
- How to make AI cat videos: 5 formats with prompts and costsHow to make AI cat videos from one cat photo: five formats (vlog, soap opera, singing, skit, dance) with draft prompts, voice options and list-price costs.
- AI dancing video: how to make a photo dance, and which tool to useHow to make an AI dancing video from one photo: what Viggle, Kling, DomoAI and ElevenLabs document on identity, limbs, length and music, and a rights checklist.
- AI cat dance videos: how they're made, with promptsHow AI cat dance videos are made from one cat photo: three prompt recipes, tool limits, 8-second costs from list prices, and which songs risk a takedown.
- How to make an AI commercial from a product photoHow to make an AI commercial from a product photo: storyboard, prompts, a label-drift scorecard and the cost of a 15-second ad on Kling, Runway and Veo.
- AI Halloween trend videos: how to make the zombie, Thriller and costume clipsThe AI Halloween trend is mostly photo filters. Here are the three video versions, a prompt for each, and 10-second costs from Runway's list prices.
