ASMR video from a scene description

What it is

This n8n workflow turns a short line about an ASMR moment, such as a knife going through a crisp apple, into a video with its own sound. Eight nodes run in one line. A form takes your words, and an agent on OpenAI’s gpt-4.1-mini expands them into a Veo 3 prompt in nine labelled parts, three or four layers of sound among them. Kie.ai renders it with Veo 3 Fast. Five minutes later the workflow asks for the clip’s address.

1. A form takes a line about an ASMR moment
2. An OpenAI agent writes a sound-led Veo 3 prompt
3. Kie.ai renders it with Veo 3 Fast
4. Five minutes later the clip’s address arrives

Who it suits

Channels that post ASMR shorts and want a steady run of satisfying clips without filming anything. Food, soap, slime and fabric all fit. The prompt keeps people out of the picture apart from hands or a mouth, with no music and no speech. Whispered talk or role play needs a rewritten system message first.

Where to use it

Any n8n that can reach OpenAI and Kie.ai will run it. The finished clip stays at an address on Kie.ai.

• n8n Cloud: import the file into a new workflow there
• Your own n8n server: the same file, the same steps
• Shorts, Reels and TikTok: set the frame in the request too, as Good to know explains, then post the clip
• A sheet or a chat: add a node after Result to keep the address or send it on

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Its author puts setup at about 8 minutes.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add an OpenAI credential and select it in OpenAI Chat Model.
3. Add an HTTP Header Auth credential named for the header Authorization, with Bearer, a space and your Kie.ai key as its value.
4. Select that credential in Generate Video and in Get Result.
5. Save, send a scene through the form’s test address, then press Publish.

How to use it

Open the form from On form submission and describe the moment in Describe your ASMR scene. A few words are enough. The field’s own hint is brushing a velvet cloth. The form confirms at once, and the run then waits five minutes before it asks for the clip. Open it under Executions after that and copy Video URL from Result.

Example

You type a knife slicing through a cold, crisp apple. The agent answers with each part on a line of its own:

Subject: …
Setting: …
Action: …
Style: …
Sound: …
Camera Motion: …
Composition: …
Ambiance: …
Aspect Ratio: 9:16

Before sending it, Generate Video deletes the line breaks and any double quotes. Kie.ai receives this:

{
  "prompt": "Subject: …Setting: …Action: …",
  "model": "veo3_fast"
}

Five minutes on, Get Result asks about the task, and Result keeps the first address in resultUrls.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. A credential is a key n8n keeps for you and hands to the node you pick it in. The Header Auth kind adds one header to every request it is picked in, here the key Kie.ai reads. Send a scene from the form’s test address first and open each node to see what it got.

For experts. One model writes and one renders. Prompt Agent has no output parser, so its whole answer goes to Kie.ai as plain text. Both Kie.ai calls use the host kieai.erweima.ai. Kie.ai’s Veo documentation gives the same paths on api.kie.ai, so move the two nodes there if the old host stops answering.

The line breaks are removed with nothing in their place, so the parts run together, as in close-up.Setting:. Write .replace(/\n/g, " ") in Generate Video for a space between them.

Good to know

The status loop waits out errors, not the render.

• Retries: Get Result loops through 30 Seconds only when its request fails. A clip still rendering after five minutes goes on to Result, which finds no address. Lengthen 5 Minutes, or add an If node that sends the run back to the wait until data.successFlag is 1
• The frame: the prompt asks for 9:16, but the request names no aspectRatio, so Kie.ai picks the shape
• The original copy: its Generate Video removed line breaks with /n/g, which deletes every letter n in the prompt. This copy has the backslash back
• Billing: OpenAI bills the prompt, and Kie.ai bills the render
• Your scene: it goes to OpenAI and Kie.ai, and the finished clip sits at Kie.ai’s address
• Open form: once the workflow is published, anyone with the form’s address can start a run on your keys, so turn on the form trigger’s basic authentication or keep the address to yourself
• Credentials: the JSON holds none, and nothing from its author’s account