No sign-up. No install.
AI video prompt
Give the length and what happens. The cut count follows the length instead of being fixed first, and the subject holds.
This is what comes out
Make a prompt for a 15-second wireless-earbud unboxing reel**Content for the generation field**
Wireless-earbud package, [FILL IN: exact product and packaging description], hands open the package in a clear physical sequence, lid lifts, inner tray reveals the charging case, case opens, earbuds become visible, compact tabletop product composition, close-up framing for packaging details, controlled hand movement, stable object placement, soft directional light landing on the box and case surfaces, visible paper and plastic textures, restrained neutral background, shallow but readable depth of field, clean product-video styling, tactile unboxing rhythm, no unsupported branding or product claims.
Use the user's exact product facts wherever supplied. Do not add a colour, logo, material, feature, action, person, setting, or sound that the input does not establish. If the user has not supplied a visual fact, keep it as `[FILL IN: item]` rather than inventing it.
Use structured tables and noun-phrase blocks for production values. Use short narrative sentences only for the change rationale and handoff explanation. The decisive passage is the reveal: give it the longest uninterrupted beat so the product can be read clearly.Full instruction | What the engine reads it as | 3 scenes |
|---|---|
| Left for the engine to decide | 9 things — put them in the request and they go away |
| Set by the instruction | 4/4 |
| Lines you can drop | None |
| Same subject on a second run | No |
The instruction went in as-is; this is what came back, unedited.
One run. Paste it again and the wording will differ.
This is what Mephify is programmed to produce, so it is AI-authored output, published exactly as it came out.
Make yours
Other media documents: AI image promptAI music prompt
Paste it into whichever generator you use. No tool names are baked in.
1. Type
The fields below change with the type you pick — each type needs different material.
2. What happens
We use only the facts you provide — nothing invented, and your text is never stored.
Ultimate prompt
We turn the choices above into an instruction you can paste into your AI.
The same instruction gives different wording on Gemini. Compare on the example page.
Common questions
- Can I use it without an account?
- All you need is the product and the length. Fill the fields and press the button; there is nowhere on this site to create an account.
- Where do I paste what comes out?
- Not using one particular tool? That is fine. Any video generator that takes a text prompt receives it, and no tool name is written into the prompt. Each cut is labelled, so you can paste them one at a time where your service works that way.
- Does what I type stay anywhere?
- The text is not kept. What remains is a count, such as how many characters you typed; the privacy policy has the details.
- Do the cuts fit the length?
- Fifteen seconds in six cuts? That is what the prompt avoids: the cut count follows the length instead of being fixed first. In the unboxing run above, one stand-in split 15 seconds into four scenes and another into three, and each scene carried its own duration. Fix the count in the request when you need an exact number. Read more
- What about aspect ratio and resolution?
- State both in the request. The unboxing prompt left them open, so one stand-in defaulted two of its four parameters while the other took all four from the prompt. Write 9:16 and the resolution you want, and the engine stops choosing.
- What if the recipe or the product is missing?
- Then the engine invents what to show. The recipe run gave nothing but fried rice, and the two stand-ins guessed 29 and 34 things: the ingredients and amounts, the pan, the counter, even the direction of the light. Give the ingredient list or the product description and the scene contents stop being invented.
More ai video prompt examples
From the operator's verification log
What to watch for
Unlike the Korean edition, this one leaves aspect ratio and resolution to the engine — state 9:16 and the resolution in the request. Runway's two 'contradictions' are the slot-versus-render tension every media instruction has (fill the slot, or keep it as a slot) rather than real conflicts. The product still has to be described; otherwise you get someone else's earbuds.
Where not to use it
A stand-in check, not a rendered video. Scene counts (4 versus 3) show the instruction leaves cutting to the engine — fix the count in the request if you need it.