Global-AI-Group-Inc/content-generation-skills / skills/mira-ugc
mira-ugc
VerifiedCreator-style clips and photos on Mira AI: phone-shot, handheld, someone talking to the camera, a review, an unboxing, a how-to, a testimonial. Use it when the user says UGC, "shot on a phone", "like a customer filmed it", "creator video", "TikTok ad", "influencer style", "отзыв", "распаковка", "как снято на телефон". Pass skill="ugc" to generate_video or generate_image on Mira. Do NOT use for glossy product hero shots (mira-product), brand films (mira-cinematic) or anything drawn (mira-anime, mira-toon3d, mira-pixel).
Mira UGC
UGC sells because it does not look like an ad. The camera is a phone, the light is a window, the person is a customer, and the product is in their hand because they use it. Everything below protects that impression. Polish is the enemy here.
When to use
A real person shows or talks about a product the way a customer would: a review, an unboxing, a short how-to, a testimonial, a "day with it" clip. Vertical, social, meant to sit between other people's phone videos and not stand out as paid.
Doctrine
This is authentic user-generated content: a real customer's casual smartphone clip, never a studio ad. The camera is handheld and slightly imperfect, held by the person or a friend, at arm's length or propped on a table. The setting is an ordinary, well-kept place the person really uses: a kitchen, a bathroom shelf, a car seat, a desk, with painted walls in good repair and a little everyday clutter in the background - a mug, a charger cable, a few books. Light is whatever is there, window or lamp, with no grade and a touch of overexposure where the window is. Skin keeps real texture; nothing is retouched. The person talks or reacts straight into the lens with natural pauses and mid-sentence gestures, and holds the product casually, even awkwardly close, so its label and shape stay readable in every beat. One honest benefit, said or shown, not a list. Small flaws are on purpose: one reframe mid-take, a finger near the lens, a real room sound. Write the final prompt in English. Do not invent brand names, on-screen text or logos the user did not give; the user's own product label stays exactly as it is.
Interview
Ask only what the request leaves open. Options are fixed; each carries the clause the pipeline gets.
- format What kind of clip is it? (default: review)
reviewReview: The creator has used the product and reports one concrete result, holding it up to the lens while speaking.unboxingUnboxing: The clip opens on the sealed package in the creator's hands; the reveal lands at the midpoint and the product is examined up close after it.howtoHow-to: The creator demonstrates one real use of the product step by step, hands in frame, the result shown at the end.testimonialTestimonial: The creator speaks to the camera about a before and an after, the product visible on the table or in hand throughout.
- platform Where will it be posted? (default: reels)
reelsReels: Vertical 9:16, cut for a feed of other phone clips, the face and the product large in frame from the first second.tiktokTikTok: Vertical, fast opening line, casual delivery, the product shown within the first two seconds.storiesStories: Vertical, looser and more intimate, as if replying to a friend, shorter beats.
- hook How does it open? (default: problem)
problemProblem first: The first second shows the annoyance the product removes, then the person picks it up.resultResult first: The first second shows the finished result, then the person explains how they got it.curiosityCuriosity: The first second shows the product half hidden or mid-use, and the person answers the question the viewer just asked.
Hint
What is the product, who is showing it, and what would that person honestly say about it in one line.
Model picks
For speech to camera pick minimax: it holds spoken lines and legible packaging text. For a quick
silent clip omni is cheapest and cuts on its own. For a 15 or 30 second story with the same
creator across shots, seedance-2-5. For a photo, gpt-image-2 for a plain phone-photo look,
gpt-image-2.6-flare when the packaging text must be readable at the exact story ratio,
nano-banana-pro for a multilingual design, muse-image (1 credit) for a cheap draft or a shot
in a real, recognisable place.
Checklist
- The product is visible and readable in every beat, held or placed, not floating.
- One benefit. If you wrote a list, cut it to the one the person would actually say.
- Nothing in the prompt asks for studio light, a grade, a lens or "cinematic" anything.
- The person is a customer, not a model. Say what they are doing, not how attractive they are.
- The room is ordinary and looked after: its life shows in the things in it, not in damage.
Posting it
To post the result to the user's own Instagram or Threads, follow mira-generate, references/social-publishing.md: make the media in the placement's ratio, keep the hook in the first 125 characters of the caption, and pair a "comment WORD" call with a keyword auto-reply that sends the link by DM. Check first, publish only after the user's yes.
Inside AI Marketing Studio
In a campaign (filmProjectId with projectType: "marketing"), take a UGC or TikTok style from
mkt_list_styles, keep brandMode off so the clip looks native to the feed, and cast talent from
the brand library by elementId instead of bloggerId. The first beat is a catalog hookId or your
own hookText. Test two to four hooks with mkt_create_variants (dry run first) before adapting the
winner to other placements. Details: mira-generate, references/marketing-studio.md.
Do not
Do not write "professional", "high quality" or "4K": those words push the render toward an ad. Do not put a headline or text on screen unless the user gave the exact words. Do not describe the room as minimal, spotless or styled like a showroom; UGC lives in real rooms. Do not age the room either: no peeling, patched, stained, cracked or worn walls and floors unless the user asked for a run-down place. Video models draw those words literally, and a living room comes back as an abandoned flat.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |