Prompts
Ten copy-ready TikTok AI video prompts for muvi.video, written 9:16 vertical with a 1-2 second hook, sub-15-second pacing, and caption-friendly framing.
To demonstrate hook retention and dynamic physical kinematics on mobile screens, here is an example generation:
“9:16 vertical full-bleed format, rapid energetic camera zoom into a charismatic creator demonstrating a viral reaction, vibrant studio rim lighting, expressive face, 24fps motion blur, broadcast crisp 1080p”
Notice how the subject remains razor-sharp during swift motion, and the upper and lower bounds remain completely clear for TikTok's native UI elements and auto-captions.
A TikTok clip lives or dies in the first one to two seconds. The format rewards a vertical 9:16 frame, a hook that lands before the viewer can scroll, motion that reads on a small screen, and a payoff inside roughly fifteen seconds. None of that comes from typing "make a TikTok video" into a generator. It comes from seven deliberate decisions specified before the model renders a frame: hook, vertical framing, subject and action, motion energy, on-screen text safe zones, audio, and pacing.
On muvi.video, you have access to specialized models tuned for each vertical need:
This guide breaks down the seven components of a high-converting TikTok prompt and provides ten copy-ready vertical prompts you can run today.
A TikTok prompt is a layered specification, not a caption. Veo 3.1 and modern video models respond to explicit instructions about framing, motion, and timing. The seven components below map to the decisions that separate a clip that holds attention from one that gets scrolled past.
1. The Hook (first 1-2 seconds) The opening frame is the only frame guaranteed a view. Specify what happens immediately: a face turning to camera, an object dropping into frame, a fast push-in, a reveal mid-action. "Open on a close-up of hands cracking an egg into a hot pan, sizzle audible from frame one" front-loads motion and sound. "Slow build, then something happens" wastes the only seconds that matter. State the hook action in the first sentence.
2. Vertical Framing (9:16) Aspect ratio is a first-sentence decision. Both Veo 3.1 and Seedance 2.0 support native 9:16, which means the model composes for vertical rather than cropping a horizontal frame. "9:16 vertical, subject centered with headroom above for caption space, full-bleed background" sets the composition logic before anything else. A subject framed for 16:9 and then cropped to vertical loses the edges the prompt was built around, so commit to vertical upfront.
3. Subject and Action TikTok favors a single clear subject doing one legible thing. "A barista pulling an espresso shot, steam rising, focused expression" is readable on a phone held at arm's length. Crowded multi-subject scenes lose definition at small size. Name the subject, name the single action, and let the action carry the clip.
4. Motion Energy Vertical short-form rewards visible motion. Specify camera move and subject move together: "handheld push-in toward subject's face, subject leans into the lens" reads as energetic; "static shot of a person standing" reads as flat. Common TikTok-friendly moves: fast push-in, whip into the subject, orbit around a product, locked-off frame with strong subject movement, and a snap from wide to close on the beat.
5. On-Screen Text and Caption Safe Zones TikTok overlays its own UI on the right and bottom of the frame, and creators stack captions over the top third. Keep the subject and any key action in the central vertical band. "Leave the top fifth and bottom fifth clear of critical action for captions and platform UI" instructs the model to protect those zones. Do not ask the model to render burned-in text; add captions in post where they stay editable.
6. Audio Veo 3.1 generates native ambient and dialogue audio alongside the video, which is the lever for talking-head and dialogue shorts. "She looks at camera and says one short line, room tone underneath, no music" gives a usable sync take. For clips you plan to pair with a trending sound in the TikTok editor, specify "ambient only, no dialogue, no score" so the native audio does not fight the track you add later.
7. Pacing and Payoff TikTok pacing is dense. A clip should reach its payoff well inside fifteen seconds, often inside eight. Specify the arc: "hook in the first second, action through the middle, clear payoff frame at the end." For a single beat, a Veo 3.1 8-second clip is usually enough. For a longer continuous vertical take, Seedance 2.0 runs to 15 seconds. Know your model's duration ceiling before writing the arc, the model cannot stretch a beat past its limit.
Not every TikTok clip needs the same model. The tree below maps the most common vertical use cases to the model that handles them best. Choose before writing the prompt, the wrong model costs coins and produces output that prompt revision alone cannot fix.
Veo 3.1, Best for: hook-driven shorts, talking-head and dialogue clips, native-audio beats
Veo 3.1 generates clips of 4, 6, or 8 seconds at 720p, 1080p or 4K. It supports both 16:9 and 9:16, so it composes natively for vertical TikTok framing. It produces native ambient and dialogue audio alongside the video, which makes it the go-to for talking-head shorts, lip-sync dialogue, and any beat where sound and picture need to land together. Veo 3.1 does not support deterministic seed control, so each run with the same prompt may arrange the frame differently. For a recurring character or set across multiple clips, keep character, wardrobe, and setting descriptions consistent across prompts, or use the Veo 3.1 image-to-video mode to anchor the frame with a reference still. Veo 3.1 and Veo 3.1 Fast are both available, Veo 3.1 for final delivery, Fast for cheaper iteration. For current per-generation coin costs, see the pricing page. For full capabilities, see the Veo 3.1 model page.
Use Veo 3.1 when: your clip is 8 seconds or shorter, you need native audio or spoken dialogue, or you want a hook-driven vertical beat at 1080p.
Seedance 2.0, Best for: longer continuous vertical takes (up to 15 seconds)
Seedance 2.0 generates clips at any integer duration from 4 to 15 seconds, across six aspect ratios including 9:16. It outputs at 480p, 720p, or 1080p. It also supports omni-reference for carrying a subject across the take. When a vertical TikTok needs a single sustained motion arc longer than Veo 3.1's 8-second ceiling, a walk-through, a long product orbit, a continuous transformation, Seedance is the route. Its 1080p ceiling sits below Veo 3.1's 4K, so reserve it for clips where duration matters more than maximum sharpness.
Use Seedance 2.0 when: your vertical clip needs to run longer than 8 seconds in a single take, you want a specific integer duration between 9 and 15 seconds, or you need omni-reference to hold a subject across the take.
Kling 3.0, Best for: specific stylized motion aesthetics
Kling 3.0 Pro (3 to 15 seconds) handles stylized motion that other models render differently. If your TikTok reference is tied to a specific motion texture rather than a technical spec, test a single clip on Kling before committing to a series.
Decision summary:
Each prompt below includes a recommended model and is written 9:16 vertical with a front-loaded hook. Swap subject, product, or location to fit your project. Keep the structural vocabulary intact, it carries the format instruction load. Add captions in post so they stay editable.
Prompt 1, Talking-Head Hook Recommended model: Veo 3.1 (9:16, 8s, native dialogue audio)
9:16 vertical. Open on a tight close-up of a woman in her late 20s looking straight into the lens, she leans in fast on the first frame and says one short line, then pauses. Soft window light from camera left, plain warm background. Subject centered with headroom above for captions, top fifth and bottom fifth clear of critical action. Room tone underneath, no music. 1080p.
Prompt 2, Product Drop Hook Recommended model: Veo 3.1 (9:16, 6s, ambient only)
9:16 vertical. A sneaker drops into frame from above and lands on a clean concrete surface in the first frame, dust puffs out on impact. Fast push-in toward the shoe over the next few seconds. Hard top light, crisp shadow. Product centered in the middle band, edges clear for platform UI. Impact thud and ambient room tone, no dialogue, no score. 1080p.
Prompt 3, Recipe First-Person Recommended model: Veo 3.1 (9:16, 8s, native ambient audio)
9:16 vertical, overhead first-person angle. Open on hands cracking an egg into a hot pan, sizzle audible from frame one. Hands continue, adding a pinch of seasoning, steam rising. Bright even kitchen light. Action kept in the central vertical band, top and bottom fifths clear for captions. Ambient sizzle and kitchen tone, no music. 1080p.
Prompt 4, Before-and-After Reveal Recommended model: Veo 3.1 (9:16, 8s)
9:16 vertical. Open on a cluttered desk in low energy, then a fast whip-pan reveals the same desk clean and styled. Single hook beat at the start, snap transition in the middle, settled payoff frame at the end. Natural daylight from a window right. Subject centered, headroom above for captions. Ambient room tone only, no score. 1080p.
Prompt 5, Outfit Orbit Recommended model: Veo 3.1 (9:16, 8s)
9:16 vertical. A person in a styled outfit stands against a plain colored wall and the camera orbits around them once, smoothly, starting on a strong front pose. Even soft light, saturated wall color for contrast. Full body framed with margin top and bottom for captions and platform UI. Ambient only, no dialogue, no music so a trending sound can be added later. 1080p.
Prompt 6, Pet Reaction Recommended model: Veo 3.1 (9:16, 6s, native ambient audio)
9:16 vertical. Open on a dog's face filling the frame, ears perking up on the first frame as it hears an off-screen sound. Quick tilt of the head, then it looks directly into the lens. Soft indoor light, blurred living-room background. Subject centered, edges clear for captions. Ambient home tone, no music. 1080p.
Prompt 7, Street-Style Walk Recommended model: Seedance 2.0 (9:16, 12s, longer continuous take)
9:16 vertical. A person in a bold jacket walks toward the camera down a city sidewalk in a single continuous take, confident pace, looking ahead then to the lens. Handheld follow, slight bounce for energy. Golden late-afternoon light, busy but soft-focus background. Subject in the central band, headroom for captions. Twelve seconds, 720p, ambient street tone.
Prompt 8, Satisfying Process Loop Recommended model: Seedance 2.0 (9:16, 10s, continuous take)
9:16 vertical. A continuous overhead shot of a hand smoothing a strip of fresh paint across a surface in one even pass, color fully covering the old layer. Slow steady motion, no cuts, framed so the action starts and ends near the same point for a clean loop. Even bright light, high color saturation. Central band protected, edges clear. Ten seconds, 720p, ambient only.
Prompt 9, Get-Ready Mirror Beat Recommended model: Veo 3.1 (9:16, 8s, native ambient audio)
9:16 vertical, shot as if filming into a bathroom mirror. Open on a person snapping the phone up into position, then they apply one quick step of a routine and glance at the lens. Warm vanity light around the mirror. Subject centered, top fifth clear for captions, bottom fifth clear for platform UI. Ambient room tone, no music. 1080p.
Prompt 10, Quick Tip Direct-Address Recommended model: Veo 3.1 (9:16, 8s, native dialogue audio)
9:16 vertical. Open on a man at a desk turning to the camera mid-gesture and beginning to speak one concise line, hands moving to emphasize the point. Clean office background, soft key light from camera right. Subject framed chest-up and centered, headroom above for captions. Room tone underneath the dialogue, no score. 1080p.
Most TikTok prompts fail at the specification layer, not the idea layer. The clip looks generic or gets scrolled because the prompt left the format-critical decisions vague.
Mistake 1: Burying the hook "A calm scene that slowly builds into a reveal" spends the only guaranteed seconds on nothing. The first one to two seconds decide whether the clip is watched. Fix: state the hook action in the first sentence, an entrance, a drop, a fast push-in, a face turning to camera.
Mistake 2: Composing for 16:9 and cropping to vertical Asking for a horizontal frame and planning to crop later sacrifices the edges the prompt was built around and reframes the subject in ways the prompt did not intend. Fix: specify "9:16 vertical" in the first sentence so the model composes for vertical natively.
Mistake 3: Ignoring caption and UI safe zones TikTok stacks captions over the top third and its own UI on the right and bottom. A subject framed edge to edge gets covered. Fix: keep the subject and key action in the central vertical band and instruct the model to leave the top and bottom fifths clear.
Mistake 4: Asking the model to burn in text Rendered text inside the clip is hard to read, often misspelled, and cannot be edited. Fix: leave caption space in the frame and add the text in post where it stays editable and legible.
Mistake 5: Requesting a duration the model cannot deliver A single Veo 3.1 clip caps at 8 seconds. Asking for a 12-second continuous Veo take wastes the instruction. Fix: for a single take longer than 8 seconds, use Seedance 2.0, which runs to 15 seconds; or cut two Veo 3.1 clips together if you need 4K.
A TikTok clip rarely ships from the first generation. The loop below reduces coin spend and tightens the path from idea to publish-ready vertical clip by separating the hook test from the style-lock pass.
Use Veo 3.1 Fast to confirm the opening one to two seconds land. At this stage, skip fine style detail and check only whether the hook reads, the framing is vertical, and the subject sits in the safe band. Fast costs fewer coins than Veo 3.1, so you can compare several hook reads cheaply.
Once the hook works, switch to Veo 3.1 and add the full vocabulary: lighting, motion energy, audio direction, and pacing. Because Veo 3.1 does not support deterministic seed control, use consistent, specific prompt language across takes to keep framing stable. For a recurring subject, use the Veo 3.1 image-to-video mode with a reference still.
For a continuous vertical clip that needs to exceed 8 seconds, move to Seedance 2.0 after confirming the beat in a Veo 3.1 draft. Seedance supports any integer duration from 4 to 15 seconds and omni-reference for holding a subject across the take. Resolution caps at 1080p, so reserve it for cases where duration matters more than sharpness, or use Veo 3.1 clips if you need 4K.
Final delivery should use Veo 3.1 at the target resolution. Render native 9:16, do not crop a 16:9 output to vertical. Add captions and any trending sound in the TikTok editor after export, so the text stays editable and the native audio does not fight the track you add.
Consistency across clips: Veo 3.1 does NOT support deterministic seed control, each generation with the same prompt may vary. For a recurring character or set across a TikTok series, rely on consistent character, wardrobe, and setting descriptions, and anchor frames with reference stills via the Veo 3.1 image-to-video mode.
We suggest Veo 3.1 in 9:16 at 1080p for TikTok on muvi.video; it composes natively for vertical, generates native ambient and dialogue audio, and handles hook-driven shorts and talking-head clips up to 8 seconds. For a single vertical take longer than 8 seconds, Seedance 2.0 runs from 4 to 15 seconds in 9:16 at up to 1080p. For stylized motion, test Kling 3.0. Pick the model by duration and audio needs first, then refine with the seven prompt components.
Yes. Both Veo 3.1 and Seedance 2.0 support native 9:16, which means the model composes for vertical framing rather than cropping a horizontal output. Include "9:16 vertical, subject centered, headroom above for captions" early in your prompt to set the composition logic before other instructions. Cropping a 16:9 output to vertical in post sacrifices resolution and reframes the shot in ways the original prompt did not intend.
TikTok rewards a payoff well inside fifteen seconds, often inside eight. A single Veo 3.1 clip generates at 4, 6, or 8 seconds at 720p, 1080p or 4K, which covers most single-beat hooks. For a continuous vertical take longer than 8 seconds, Seedance 2.0 supports any integer duration from 4 to 15 seconds. For a longer edited piece, cut several short Veo 3.1 clips together rather than generating one long clip, this keeps every frame at full resolution.
State the hook action in the first sentence of the prompt: an entrance, an object dropping into frame, a fast push-in, or a face turning to the lens. The opening frame is the only frame guaranteed a view, so front-load the motion and, on Veo 3.1, the audio. Avoid slow builds, they spend the seconds that decide whether the clip is watched.
Yes, on Veo 3.1. Veo 3.1 generates native dialogue and ambient audio alongside the video, so a talking-head or direct-address clip can land sound and picture together. Specify one short line and add "room tone underneath, no music" for a clean sync take. If you plan to pair the clip with a trending sound in the TikTok editor, specify "ambient only, no dialogue, no score" so the native audio does not fight the track you add.
Add them later. Models render in-frame text unreliably and the result cannot be edited. Instead, leave the top fifth and bottom fifth of the 9:16 frame clear of critical action, keep the subject in the central band, and add captions in the TikTok editor or your post tool after export, where they stay editable and legible.
Veo 3.1 generations are coin-based across all plans, and the included monthly coin allotment differs by tier. Pro and higher plans remove the watermark. For current per-generation coin costs and monthly allotments by plan, see the pricing page.
Apply these prompts in muvi.video Studio. Select Google Veo 3.1 for hook-driven vertical shorts and talking-head clips with native audio. For viral dance kinematics and takes up to 15 seconds, switch to ByteDance Seedance 2.0. For legible apparel and product typography, deploy Alibaba Wan 3.0 Prime. Iterate from first hook test to publish-ready 9:16 clip in a single workspace with drafts to 1080p finals (up to 4K on Google Veo 3.1).
Free starter coins · 14 video models · Up to 4K on Google Veo 3.1
Ten copy-ready TikTok AI video prompts for muvi.video, written 9:16 vertical with a 1-2 second hook, sub-15-second pacing, and caption-friendly framing.
