How to Make AI Videos: A Beginner’s Guide, Plus the Master Prompt Behind Viral Bottle Videos
A practical guide for creators who want to start making AI videos today
Spend five minutes on YouTube Shorts, Instagram Reels or TikTok and you will probably meet one of them: the camera glides toward a glass bottle, slips through its neck, and opens into a complete miniature world, such as a misty valley, a stormy sea or a tiny orchard, before pulling back out to the bottle on a kitchen table. These “inside the bottle” videos are short, hypnotic and very easy to rewatch, and many of them are collecting huge view counts.
The good news is that you do not need a camera, a studio or editing experience to make one. You need an idea, an AI video tool and a well-written prompt. This guide walks you through the whole process, from choosing a tool to the master prompt, and adds a few opinions of my own along the way.
Why Bottle Videos Are Taking Over Social Media
The format works for reasons that have little to do with AI. First, it has a built-in hook. Viewers see a bottle and instantly wonder what is inside, so they keep watching to find out. Second, it is satisfying: one smooth camera move, no cuts, and a calm payoff that often loops well. Third, it is endlessly variable. Change the world inside the glass and you have a brand-new video, whether that is a snowy village, a coral reef, a desert canyon or a futuristic city.
Fourth, it suits vertical video. A tall bottle fits naturally into a 9:16 frame, and short clips of eight to ten seconds are exactly what AI tools produce best. Platforms reward watch time and replays, so a hypnotic clip that people watch twice can travel far. The effect is now common enough that AI platforms publish dedicated guides for it, describing the same idea: a camera that pushes through the bottle neck, finds a landscape inside, and pulls back out.
Choose Your AI Video Tools
The AI video market changes quickly, so treat any ranking as a snapshot. In 2026, reviewers most often recommend Google Veo 3.1 for realism and built-in audio, Runway Gen-4.5 for creative control such as camera moves, and Kling 3.0 for strong motion at a lower price. Pika, Luma Dream Machine, Hailuo and PixVerse are also popular, and some offer free tiers that are ideal for testing. One warning: several 2026 reviews report that OpenAI shut down Sora, so do not build your workflow around it. Always check each tool’s current pricing, watermark rules and commercial-use terms before you publish.
Besides a video generator, three more things help: an AI image tool for creating a starting frame, a simple editor such as CapCut or DaVinci Resolve for trimming and adding music, and a notes app to save every prompt that works. Your prompt library will become your most valuable asset.
Step-by-Step: How to Make an AI Video
Step 1: Pick one clear idea
Decide on a single world and a single mood before you open any tool. “A rainy Tokyo street inside a perfume bottle” is a good idea. “Something cool in a bottle” is not. Specific ideas give the AI less room to guess.
Step 2: Create a starting image
For the best control, generate a still image of your bottle first, then animate it with image-to-video. This locks in the bottle’s shape, the glass and the background, which reduces strange changes during the clip. If your tool only supports text-to-video, skip this step and rely on a detailed prompt.
Step 3: Write a structured prompt
Good prompts describe the shot like a film director: what the camera does, what we see, how it is lit and how it feels. The next section gives you a master template you can reuse.
Step 4: Generate, review and regenerate
Expect to generate the same prompt several times. Watch for warping glass, a camera that passes through the bottle in an unnatural way, or an inner world that looks unrelated to the outside. Change one thing at a time so you learn what each change does.
Step 5: Add sound and polish
Some tools, such as Veo, can generate matching audio. If yours cannot, add soft ambient sound, a gentle whoosh as the camera enters the bottle, and calm background music. Sound is half of what makes these videos feel premium.
Step 6: Export and post
Export in vertical 9:16, keep clips under fifteen seconds, and write a title that creates curiosity, such as “What is hiding inside this bottle?” Post consistently and study which worlds get the most replays.
The Master Prompt for Bottle Videos
Here is a template you can copy and customise. Replace everything in square brackets with your own details.
A single continuous cinematic shot, vertical 9:16. The camera starts outside a [type of bottle, e.g. tall clear glass bottle with a cork] standing on [surface and setting, e.g. a wooden kitchen table beside a window]. It moves slowly forward and pushes smoothly through the neck of the bottle, then opens into a complete miniature world inside: [describe the world, e.g. a misty green valley with a river, tiny pine trees and a small wooden cabin]. [Lighting, e.g. soft golden morning light from the left, matching the light outside the bottle.] [Small motion, e.g. mist drifting, leaves moving, water flowing.] Photorealistic, shallow depth of field, rich detail, calm ambient sound. No cuts, no text, no watermark, no distorted glass.
Here is the same template filled in:
A single continuous cinematic shot, vertical 9:16. The camera starts outside a tall clear glass bottle with a cork stopper, standing on a wooden table beside a rainy window. It glides forward through the neck of the bottle and opens into a miniature misty mountain valley with a winding river, tiny pine trees and a small wooden cabin with a glowing window. Soft grey daylight from the left lights both the room and the valley. Mist drifts across the hills and raindrops ripple the river. Photorealistic, shallow depth of field, calm ambient sound of rain. No cuts, no text, no watermark.
Why does it work? The prompt names exactly one camera move, describes the container and the inner world separately, matches the lighting so the inside and outside feel like one place, and ends with exclusions that prevent common glitches. Once you have it working, try new worlds: a ship sailing through a storm, a galaxy of stars, an underwater city, a snowy village, a desert canyon, or a mountain village from your own region.
Tips for Better Results
Keep the lighting consistent. If the bottle sits in warm side light, light the inner world from the same direction, otherwise the transition looks fake. Use one camera move per clip, because asking for a push-in, a rotation and a zoom together usually confuses the model. Keep clips short, then join two or three in your editor if you want a longer story.
Save the best frame of a good clip and use it as the starting image for the next one, which helps you build a consistent series. Finally, describe small, natural motion such as drifting mist, flickering candlelight or rippling water. Tiny movements make a scene feel alive without asking the AI to do something complicated.
Common Mistakes to Avoid
The biggest mistake is a vague prompt. The second is overloading one clip with too many ideas, such as ten characters, three camera moves and a plot twist. The third is giving up after one result. Professional AI creators often generate many versions before choosing one, and they treat each failure as information about what to change.
Another common error is ignoring the first second. Viewers decide whether to keep watching almost immediately, so the bottle should look beautiful and interesting from the very start.
Publishing and Using AI Video Responsibly
Be honest about what you make. Many platforms now ask creators to label realistic AI-generated content, so check the current rules on YouTube, Instagram and TikTok before posting. Do not create videos that imitate real people without permission or that could mislead viewers about real events. Read each tool’s licence too, because free plans often add watermarks or restrict commercial use.
Create an ultra-realistic, visually satisfying viral YouTube Shorts video, vertical 9:16 aspect ratio, exactly 8 seconds.
REFERENCE STYLE:
Recreate the visual concept of a miniature musical ball-drop staircase machine, similar to satisfying viral physics videos. Preserve the recognizable layout: a metallic stepped track, a small ball released by a hand, red collection cups at the bottom, and a tiny green fantasy creature watching from the right.
SCENE AND COMPOSITION:
A miniature mechanical musical staircase machine sits on a clean, light gray tabletop in a softly lit studio.
Build a straight, symmetrical descending ball-drop track consisting of EXACTLY 10 clearly separated, horizontal silver metallic bars arranged like a miniature staircase. Each bar has small, dark metallic striking points. The staircase descends from the upper background toward the lower foreground, with a clearly defined downward ball path.
At the bottom, arrange five small red translucent plastic collection cups in a compact cluster, matching the satisfying toy-machine aesthetic. Position ONE designated target cup directly beneath the final exit point of the ball track.
On the right side of the staircase, place one cute, tiny, bright green fantasy creature with large expressive eyes and thin blue-purple strands of hair. The creature remains in the same position throughout the video, watching the ball.
BALL AND STARTING ACTION:
Show exactly ONE small, shiny silver metallic ball held between the fingers of ONE realistic human hand at the top of the staircase.
The hand gently releases the ball at the beginning of the video. The fingers open naturally, and the ball immediately begins its journey.
CONTINUOUS BALL PHYSICS:
The ball must visibly contact the first metallic bar and then progress through all 10 bars in sequential downward order.
On every impact, the ball makes a believable small bounce or guided hop toward the NEXT lower bar. Its motion must follow one continuous, physically plausible path under gravity.
Use a compact, clearly readable bouncing rhythm, with brief contact on each successive bar. The ball must visibly touch each of the 10 bars at least once, without skipping any bar.
Maintain consistent ball size, realistic inertia, natural gravity, believable rebounds, and correct spatial continuity.
The ball must never teleport, disappear, duplicate, split, reverse direction unexpectedly, or pass through solid metal.
TIMING AND PACING:
0.0–0.6 seconds: The hand releases the single ball and the ball reaches the first bar.
0.6–5.8 seconds: The ball bounces down the 10 metallic bars in a rhythmic, satisfying sequence.
5.8–7.3 seconds: The ball leaves the last bar, falls directly into the designated red cup, and visibly lands inside.
7.3–8.0 seconds: The ball settles inside the cup while the final resonant sound rings out.
Treat these timestamps as pacing targets, not permission to skip any required physical contact. Prioritize showing the complete journey and final landing.
FINAL CUP LANDING:
The ball must visibly cross the rim of the designated cup, enter its interior, make contact inside, and settle naturally at the bottom.
The final impact produces a satisfying hollow “ting-ding” sound with a short natural resonance. Do not end the video before the ball is visibly inside the cup.
CAMERA:
One fixed camera, slightly elevated front-top angle, macro product-video perspective.
Frame the entire staircase, all five cups, the starting hand, and the green creature clearly. Use a carefully composed vertical 9:16 shot with comfortable margins.
Keep the ball's full route visible at all times. No cuts, no transitions, no zooms, no camera movement, no reframing, and no sudden close-ups.
AUDIO AND SOUND DESIGN:
Generate synchronized audio for the actual visible contacts.
Each metallic bar produces a slightly distinct, clean, resonant musical ping. Create a pleasing rhythmic xylophone-like melody from the 10 impacts.
Include subtle ball movement sounds, natural bouncing, and the final hollow cup resonance.
Audio must match the exact timing of the ball's contact with each bar. No random impacts, no premature sounds, no background music, no dialogue, and no narration.
VISUAL QUALITY:
Photorealistic miniature product cinematography, realistic 3D motion, physically convincing materials, brushed silver metal, subtle reflections, translucent red plastic cups, natural contact shadows, detailed surfaces, crisp focus, soft studio lighting, smooth motion, premium satisfying ASMR Shorts aesthetic.
OBJECT CONSISTENCY:
Exactly one ball, one hand, one green fantasy creature, 10 metallic bars, and five red cups.
All objects remain visually consistent in shape, color, scale, and position throughout the entire shot.
STRICT NEGATIVE CONSTRAINTS:
No additional balls, no extra hands, no missing bars, no floating ball without physical cause, no broken staircase, no ball teleportation, no random bouncing, no impossible physics, no obstructed final landing, no creature movement distracting from the action.
No subtitles, no text, no logos, no watermark.
End on the completed successful ball landing, with the ball resting inside the target cup and the final sound naturally fading.
Also, avoid copying the exact same prompt as everyone else. When thousands of channels post nearly identical clips, viewers get bored quickly. Add your own twist, whether that is a unique world, your own voiceover or a short story told in the caption.
My Take: Where AI Video Is Heading
I believe AI video is changing who gets to be a visual storyteller. A few years ago, a shot like the bottle effect needed a studio, a 3D artist and days of rendering. Now a person with a phone and a good idea can make it in an afternoon. That is exciting, but it also means that production skill is no longer the advantage. Ideas, taste and consistency are.
Trends will also burn out fast. The bottle video is popular today, and something else will replace it next month. So do not build your channel around a single effect. Build a repeatable workflow: find a hook, write a structured prompt, test several versions, add sound and publish. If you learn that process, you can switch to the next trend in a day.
Finally, the creators who last will add something human: a voice, a story, a point of view. AI can produce beautiful pictures, but it cannot decide why people should care. That part is still your job.
Final Thoughts
Making AI videos is easier than it looks, and the bottle format is one of the best places to start because it is simple, visual and repeatable. Choose a tool, write a clear prompt using the master template, generate several versions, add sound and post. Start with one video this week, learn from the results, and improve from there.
Comments
Post a Comment