YouMind
Sign in

One Prompt to Viral Video: 360° Orbit Effects Without Deployment

@nima2qisi2
SIMPLIFIED CHINESEOct 14, 2026
111K
236
44
12
447

TL;DR

This guide explains how to create viral 360° orbit videos from single images using AI. It provides specific prompts for web-based tools like Doubao and a step-by-step workflow for local ComfyUI deployment, highlighting applications in travel, beauty, and product marketing.

The 360° orbit effect has been trending recently. Uploading a single image to generate this effect is surprisingly simple, with almost no barrier to entry—no subscription required. One image and one prompt are enough.

Think of this visual effect as AI generating continuous footage from different camera angles based on spatial cues in the photo.

What can you do with this?

Many people's first instinct is to use it for "beauty" videos, but there are many other hidden gems and commercial potentials. Travel photos, food shots, beauty visuals, commercial photography, and product showcases all have viable directions.

You can take a casual travel or food photo and generate this effect. It’s very straightforward—one image, one prompt.

Beauty content suits this orbital angle well. The camera can slowly move from front to side, observing the relationship between eyeshadow, lip makeup, facial contours, and highlights. You can also orbit around a perfume bottle or cream jar, revealing the curvature of the glass and metal cap as the perspective changes.

Brand model photos, fashion editorials, and product scene images can be used to generate new dynamic assets.

Dmiao豆苗啊 - inline image
Dmiao豆苗啊 - inline image

How exactly? Very simple—any video generation model like Doubao, Qwen, Yuanbao, etc., works.

Method 01: Web-based — Recommended for Beginners

All examples in this article were generated using this method.

https://x.com/nima2qisi2/status/2109604241363058694

https://x.com/nima2qisi2/status/2109255101466489280

Image-to-Video — Send your desired image and this prompt to Doubao:

Dmiao豆苗啊 - inline image
Dmiao豆苗啊 - inline image

Short vs. Long Prompt Usage in Doubao

General Short Prompt:

text
1Use my uploaded image as the first frame and appearance reference. Generate a 6-second continuous one-take video: The subject and environment in the frame are frozen at the moment of the original image. The camera moves to the left of the frame, orbiting continuously along a horizontal circular track centered on the subject. The lens always faces the subject, maintaining roughly the same height and distance, presenting clear foreground, subject, and background parallax. Maintain the subject's appearance, pose, proportions, and original lighting; the subject does not rotate itself, the background does not rearrange, no cuts, no zooms, no new objects added. No subtitles, no voiceover.

Complete 10-Second Prompt:

text
1Use the reference image as the first frame, identity, clothing, pose, indoor space, and composition reference. 10-second continuous ONE-TAKE video. Core effect: 360° BULLET TIME + ABSOLUTE TIME FREEZE + LEFTWARD CAMERA ORBIT + ONLY EYE GAZE MOVES. The entire world is frozen at the moment of the reference image. The adult female model maintains the original sitting posture from the reference image: body leaning against the sofa, knees bent in the foreground, hands naturally stacked on knees. Head angle, shoulders, arms, fingers, legs, and feet positions are strictly locked. The shape, length, folds, and occlusion relationships of the black dress remain completely consistent. The character absolutely cannot turn with the camera. Only her eyeballs and pupils are allowed very slight movement, slowly tracking the camera orbiting around her. NO blinking. NO head turning. NO mouth changes. NO facial expression changes. ━━━━━━━━━━━━━━━━━━ 【Camera Movement】 The camera starts from the current frontal position in the reference image. Then it moves parallel to the left along a horizontal circular track centered on the person, orbiting the person fully 360°. It is NOT the person rotating. It is NOT the background rotating. It is NOT an in-place pan camera. It MUST be: physical camera dolly + true 3D circular orbit + camera traveling through physical space. Complete full 360° orbit within 10 seconds. Average orbit speed approx: 36 degrees per second. Camera distance from the person remains basically constant, camera height remains basically constant, lens always faces the upper body / center of the face area. Trajectory: Front → left-front 45° → left profile 90° → left-back 135° → back 180° → right-back 225° → right profile 270° → right-front 315° → front 360°. Finally, the lens accurately returns near the starting position, forming a complete closed loop. ━━━━━━━━━━━━━━━━━━ 【0–2.5s|0° → 90°】 The camera starts moving quickly but smoothly to the left from the front. Gradually seeing the left profile of the person. Foreground legs and background body form obvious parallax. Sofa, pillows, carpet, and person produce real 3D spatial displacement. Model body completely frozen. Only eyes slowly follow the camera. ━━━━━━━━━━━━━━━━━━ 【2.5–5s|90° → 180°】 The camera passes the left side of the person, continuing into the back side. Must maintain stable complete 3D human structure. The person cannot re-pose because the front is not visible. Keep: Same sitting posture, same leg position, same hand posture, same head angle, same hairstyle. When the camera reaches behind the person, naturally display the back space invisible in the reference image, but must reasonably reconstruct continuity based on existing clothing, hairstyle, and body structure. Do not suddenly change hair length. Do not change black clothing structure. ━━━━━━━━━━━━━━━━━━ 【5–7.5s|180° → 270°】 The camera continues orbiting to the right from behind the person. The entire indoor space must remain fixed. White sofa, black-and-white pillows, black chimney, wall decorations, carpet, right-side black bag all remain static. Cannot rearrange with camera movement. Form very obvious: foreground / subject / background parallax. Person still completely frozen. Only pupils continue tracking the camera. ━━━━━━━━━━━━━━━━━━ 【7.5–10s|270° → 360°】 The camera passes the right side of the person, continuing back to the front. Slightly decelerate in the last 1 second. Finally, the lens approaches the shooting position of the first frame again. Person's pose remains completely consistent with the first frame. When the camera finally arrives at the front, the model's pupils look directly into the lens center. No blinking. End of screen. ━━━━━━━━━━━━━━━━━━ 【Time Freeze Rules|Highest Priority】 ABSOLUTE FREEZE. Only the pupils and eye gaze can move. Everything else must remain completely frozen. NO movement of: head neck shoulders arms hands fingers torso legs feet hair dress fabric facial muscles eyebrows mouth lips chest breathing background pillows bag carpet wall decoration. The woman must behave like a perfectly frozen 3D sculpture. Only her eyes remain alive. ━━━━━━━━━━━━━━━━━━ 【Eye Mechanism】 Eye movement must be very restrained and natural. Do not let eyes jump instantly. Do not stare dead-center constantly. Eyes slowly track continuously based on camera position: camera moves left → pupils gradually follow camera. camera reaches rear position → eye movement becomes minimal because camera is outside direct field of view. camera comes back into peripheral vision → pupils gradually reconnect with camera. camera returns front → eyes look directly into lens. Head remains absolutely static. ━━━━━━━━━━━━━━━━━━ 【Lens Quality】 photorealistic live-action photography, high-speed bullet-time photography, realistic smartphone / full-frame camera look, 24–35mm equivalent lens, natural indoor daylight, real skin texture, accurate anatomy, stable facial identity, physically correct perspective, strong spatial parallax, consistent room geometry, continuous lighting, no artificial cinematic relighting. Maintain original reference image: soft natural light, black-and-white minimalist interior, real skin texture, light gray-white environment, phone private photography feel. Do not suddenly add cinematic lighting. Do not change exposure. Do not change color temperature.

Portrait Optional Add-on:

text
1Body frozen, only eye gaze slightly tracks. Only allow eyeballs and pupils to move very slightly and naturally. When the camera is within the person's field of view, eyes slowly track the lens; when the camera enters behind, do not require eyes to continue tracking. When the camera re-enters vision, eyes naturally reconnect. Head absolutely static, no blinking, no mouth shape or expression changes.

Method 02: Local Deployment — For Advanced Users

Those familiar with local workflows can try ComfyUI.

  1. Open ComfyUI, load "MiniMax H3 I2V" from the template library. Download FL2VA model, text encoder, and VAE as prompted by the template. [ComfyUI Native Workflow]
  1. Download 360° Orbit LoRA

http://huggingface.co/pablodawson/MiniMax-H3-360-Orbit-LoRA

Place in ComfyUI/models/loras/, load before the model enters the sampler. Prioritize the author-recommended FL2VA pruned int8 ConvINT version for the base model to ensure compatibility with the LoRA.

  1. Connect the same image to both first-frame and last-frame inputs. Fill in the original prompt from the author's project page, reproduce with recommended settings first.
  1. After generation, check: Is the subject frozen? Does the camera orbit? Are sides and back distorted? Can the end frame connect back to the start?

Workflow: ComfyUI → Import Template → Load Models → Enable & Upload First Frame Image → Fill in Orbit Prompt → Set Dimensions and Frames → Run → Check Final Cut.

It is recommended to keep default template parameters first to verify functionality, then adjust resolution and duration. Save images, prompts, model versions, parameters, and seeds for easy review and reuse.

Thanks for watching.

I am Dmiao, sharing AI reusable, low-barrier tricks from a designer's perspective.

If this helped you, please bookmark and share.

Remix in YouMind

Turn one viral article into a full content workflow

Collect the source, decode the pattern, create assets, draft the story, and distribute from one AI workspace.

Explore YouMind
For creators

Turn your Markdown into a clean 𝕏 article

When you publish your own long-form writing, images, tables, and code blocks make 𝕏 formatting painful. YouMind turns a full Markdown draft into a clean, ready-to-post 𝕏 article.

Try Markdown to 𝕏

More patterns to decode

Recent viral articles

Explore more viral articles