How to make satisfying AI ASMR videos with OpenArt

AI tools have made it possible to create incredibly realistic ASMR videos without cameras, mics, or a studio. With the right workflow, you can generate hyper-detailed visuals and perfectly synced sound effects in just a few minutes, all from your desk. This guide walks through a complete process inside OpenArt to create three viral ASMR formats: creator-led food videos, top-down slime and texture shots, and comparison shorts that showcase one object made from different materials.

Why use OpenArt for AI ASMR videos?

OpenArt is an all-in-one AI platform that combines image generation, video generation, characters, worlds, and audio. For ASMR creators, the key benefits are:

• A powerful image model (GPT Image 2) for hyper-realistic characters and objects.
• A video model (Seedance 2.0) that can generate visuals and native sound effects from a single prompt.
• Character elements that keep your AI "creator" consistent across multiple clips.
• A unified workflow so you don’t have to jump between different tools.

If you’ve worked with cinematic AI content before, this approach feels similar to workflows used in projects like cinematic AI video creation, but tuned specifically for ASMR visuals and sound.

Core workflow overview

All three ASMR formats use the same core workflow:

1. Generate a base image (for creators or objects) with GPT Image 2.
2. Save important images as character or reference elements inside OpenArt.
3. Write a highly structured prompt that describes style, action, environment, and sound.
4. Convert that prompt into JSON for more precise control (using an external prompt-to-JSON helper).
5. Use Seedance 2.0 to generate vertical 9:16 videos with audio enabled.
6. (Optional) Edit and combine clips in a video editor like CapCut for multi-part formats.

Once you understand this structure, you can reuse it endlessly with new materials, textures, and concepts.

Designing your AI ASMR creator

The first format is a creator-led video where a realistic person eats or interacts with unusual foods and objects. To make this work, you need a consistent AI creator.

Generate a realistic creator image

Inside OpenArt, go to the Image workspace and choose GPT Image 2 as the model. Set the aspect ratio to 3:4, then write a prompt for your creator. Aim for a hyper-realistic, commercial-quality portrait that looks like a professional content creator. Think about:

• Age, gender, and overall vibe (cozy, futuristic, luxury, etc.)
• Lighting style (soft studio light, natural daylight, neon glow)
• Framing (portrait, mid-shot, close-up on face)

Generate a few options until you get a creator that looks like a real person you’d expect to see on social media.

Save the creator as a character element

Once you’re happy with the image, save it as a character element inside OpenArt and give it a clear name. This is crucial: without this step, every new video generation will slightly change the face, and your creator won’t look consistent across clips.

By tagging and reusing the same character element in each video prompt, you anchor the creator’s appearance so they feel like a recurring personality rather than a random AI face.

How to write prompts that sound and look real

For AI ASMR, prompt writing is just as important as model choice. The video model in OpenArt doesn’t just read visual descriptions—it also uses your audio descriptions to synthesize matching sound effects. The more physical and specific your wording, the more convincing the result.

A simple structure for ASMR prompts

A reliable prompt structure for this workflow looks like this:

1. Style and mood – Set the overall aesthetic and emotional tone (e.g., "hyper-realistic commercial close-up, cozy and intimate mood").
2. Narrative summary – One line that explains what’s happening (e.g., "a creator bites into a crystal candy that shatters like glass").
3. Dynamic description – Describe the action in detail: how the object moves, breaks, stretches, or deforms.
4. Static description – Lock in the environment: camera angle, background, lighting, props, and color palette.
5. Audio and quality suffix – Describe the sounds and add quality constraints (e.g., "crisp glass-like crack, layered with soft chewing and a close whisper, stable camera, consistent lighting, realistic physics").

After writing this in plain English, you can use a helper tool (linked from the workflow) to convert it into a structured JSON prompt. Pasting JSON directly into the video generator gives Seedance 2.0 a more precise set of instructions, which noticeably improves realism and consistency.

Format 1: creator-led AI ASMR food videos

The first format is a classic ASMR style: a creator eating or interacting with surreal, visually striking foods. Think crystal candy, molten gold concrete, neon gummy cubes, or Halloween-themed desserts.

Set up the video generator

In OpenArt, switch to the Video workspace and select Seedance 2.0 as the model. Set the aspect ratio to 9:16 for vertical social content and enable audio so the model can synthesize sound effects.

Tag the character element you saved earlier so your creator’s face and style stay consistent. Then paste your JSON prompt and generate.

Examples of creator-led ASMR clips

Using this method, you can create multiple clips with the same creator but radically different materials:

• A fragile crystal treat that shatters with a smooth, glassy crack and crunchy texture sounds.
• A "dense concrete" dessert finished in molten gold, where the bite feels heavy and the gold stretches and sticks with a slow, sticky pull.
• A galactic neon gummy cube that wobbles and glows like "pure space jelly," with squishy, elastic sound design.
• A Halloween eye macaron where the eyeball element deforms before the gelatin releases, creating a slightly unsettling but satisfying squish.

Because the workflow and character anchor are identical for each clip, you can batch-generate a whole series of creator-led ASMR videos that feel like they all belong on the same channel.

Format 2: top-down gloved hands and texture videos

The second format removes the visible creator and focuses on hands, textures, and sound. These are the classic top-down ASMR shots: gloved hands pressing into slime, honey, marbles, circuit boards, and other materials.

Skip the character, focus on materials

For this style, you can stay directly in the Video workspace with Seedance 2.0 and the same 9:16, audio-enabled setup. You don’t need a character element—just detailed prompts.

Describe:

• Camera angle: overhead, macro framing, close-up on hands and materials.
• Materials: slime, obsidian shards, marbles, honey, circuit boards, frozen berries, pencils, etc.
• Physics: how things stretch, crack, float, snap, or sink.
• Sound: sticky pulls, crisp cracks, soft squelches, gentle clinks.

Material-driven examples

Here are a few variations built from the same base style:

• Obsidian and slime – Overhead angle, macro framing that makes the slime feel enormous. The slime shifts between violet and green under softbox light, with rich squelching and subtle stone sounds.
• Marbles and honey – Lighter, less dense materials where marbles float and shift as the honey lifts them. The audio captures gentle clinks and slow, viscous movement.
• Circuit boards and slime – Visually unique shots where boards snap under pressure while slime rises through the gaps, mixing brittle snaps with wet, rising sounds.
• Frozen cranberries – A strong contrast moment: sharp, crisp cracking as frozen berries break, then soft, warm sounds when hands reach into slime in the same shot.
• Architect’s desk – Hands hold a bowl and push it into slime on a desk scattered with pencils. After lifting, the pencils fall into the slime, creating a calm, oddly peaceful sequence.

All of these come from the same workflow; only the materials and physics in the prompt change. The more specific your descriptions, the more the AI can infer realistic behavior and sound.

Format 3: comparison shorts with one object, many materials

The third format is one of the most viral: comparison shorts where a single object is recreated in multiple materials. The object itself can be anything; the magic comes from the contrast between materials and environments.

In this workflow, the example object is a toilet paper roll, reimagined in different materials like slime, honeycomb, rusted metal, and ancient stone. Each material gets its own mini-world.

Step 1: generate still images for each material

Go back to the Image workspace and use GPT Image 2. You’ll generate six images:

• One hook image that introduces the concept (e.g., a slime toilet roll on a dark matte table).
• One image for each material variation (five total).

For each material image, keep the object centered but match the entire environment to the material. For example, a honeycomb roll might sit in a warm, golden kitchen scene, while a rusted metal roll might appear in an industrial or decayed setting.

When you line these images up side by side, the contrast should be striking: same object, completely different worlds.

Step 2: turn each image into a video clip

Back in the Video workspace with Seedance 2.0, upload each image as the first frame for its corresponding video. Use a JSON prompt for each clip that describes how the material behaves and sounds when touched, squeezed, or broken.

For the hook clip, you might pair the slime roll visual with a calm whispered question like, "Which toilet roll is the best? Let’s find out." The contrast between the calm voice and chaotic visuals helps grab attention.

For the rusted metal roll, you can design the most dramatic audio moment: a metallic shriek as rust flakes off, layered over a dust cloud effect. It feels intense because it’s surrounded by quieter, softer clips in the final sequence.

For the ancient stone roll, you can build slow tension as cracks form and then collapse, with deep, weighty sounds that still follow believable physics.

Step 3: edit everything into one comparison short

This format only truly works once all clips are edited together. In a tool like CapCut (or any editor you prefer):

1. Create a new project and place all clips in order: hook, then each material clip.
2. Use hard cuts only—no transitions. The abrupt switch between materials is part of the appeal.
3. Add short on-screen labels for each material at the start of its clip (e.g., "Slime", "Honeycomb", "Rusted Metal"). Keep them brief and readable.
4. Add a soft ambient music track at low volume under everything. It should glue the video together without overpowering the ASMR sound effects.

Export at 1080p, and you’ll have a polished comparison short where each material feels unique, both visually and sonically. Because the concept is so flexible, you can endlessly swap the object and materials for fresh content.

Scaling your AI ASMR content

Once you’ve built a few successful prompts and character or material setups, scaling becomes straightforward:

• Reuse your creator character across new foods and objects.
• Reuse your top-down style with new textures and props.
• Reuse your comparison format with new objects and material sets.

Since everything runs inside one platform, you can quickly iterate, test new ideas, and build a content library. If you’re already exploring AI video tools, this ASMR workflow pairs nicely with other approaches like full story videos or cinematic sequences, such as those described in AI-generated music video workflows.

With structured prompts, consistent characters, and material-focused worlds, you can turn OpenArt into a powerful ASMR studio that runs entirely on AI.