Global-AI-Group-Inc/content-generation-skills / skills/mira-generate
mira-generate
VerifiedGenerate images and video on Mira AI through its MCP server: connect, pick a model, attach the user's AI Blogger, location or product, write the prompt, pass a playbook skill, wait for the result, extend or upscale a clip. Use it whenever the user wants to "generate", "make a video", "make an image", "animate this photo", "create an ad", "сгенерируй", "сделай видео", "сделай картинку" and a Mira MCP server is connected or can be. Chain with mira-video-prompting and mira-image-prompting for what to write, and with the mira-* playbooks for the genre. NOT for prompting theory on its own, and not for voice, music, sound effects, dubbing or captions (mira-audio).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 20.3 KB |
| references/ai-blogger.md | markdown | 14.1 KB |
| references/direct-presets.md | markdown | 3.6 KB |
| references/film-studio.md | markdown | 14.3 KB |
| references/guides.md | markdown | 9.3 KB |
| references/locks-and-references.md | markdown | 12.2 KB |
| references/marketing-studio.md | markdown | 5.6 KB |
| references/media-ops.md | markdown | 9.3 KB |
| references/miracle.md | markdown | 9.8 KB |
| references/playbooks.md | markdown | 1.8 KB |
| references/social-publishing.md | markdown | 4.9 KB |
references/media-ops.md
View on GitHub# Clip operations
Tools that take a clip the account already has and return a new generation. Each takes exactly
one source: `sourceGenerationId` for a Mira generation (`list_generations`) or `libraryItemId`
for a file sent through `create_upload` (kind `video`, or `audio` for a track). A source over the
tool's limit is refused (`source_too_long`) before a credit is spent. Every call returns a
generation id; `wait_for_generation` brings the result. Quote with `estimate_cost` and the
source's length in `sourceSeconds`; never from memory.
Sound (voice, music, effects, dubbing, captions, stems, transcripts) is in mira-audio.
`motion_control`, `recast_video`, `lipsync_video` and `avatar_video` also take a `bloggerId`: the
character frame comes from the user's AI Blogger, built from its canon for the clip's ratio, in
place of `characterImageUrl`, and speech defaults to the blogger's pinned voice. The blogger's own
tools (`blogger_motion`, `blogger_speak`) are in ai-blogger.md.
| Tool | Longest source | `estimate_cost` kind |
|---|---|---|
| `modify_video` | 15 s | `modify` |
| `reframe_video` | 15 s | `reframe` |
| `remove_background` | 30 s | `rembg` |
| `motion_control` | 30 s (`orientation: "video"`), 10 s (`"image"`) | `motion_control` |
| `recast_video` | 3 to 10 s | `recast` |
| `lipsync_video` | 60 s | `lipsync` (model `lipsync-2`, `lipsync-2-pro` or `kling`) |
| `avatar_video` | the voice track: 2 to 300 s | `avatar` (model `pro` or `std`, `sourceSeconds` = the audio length) |
| `swap_object` | 10 s (at least 3 s) | `swap` |
| `draw_edit_video` | 4 to 30 s | `draw_edit` (model = resolution `480p`, `720p` or `1080p`) |
## modify_video
The same framing, subjects and motion with one thing changed. `effect` is `lighting`, `weather`,
`time_of_day`, `backdrop` (the environment behind the subjects) or `restyle` (free-form).
`prompt` describes only the change, in English. `strength`: `adhere` keeps closest to the
source, `flex` (default) balances, `reimagine` allows the biggest change. `referenceImageUrl`
optionally shows the target look.
```
modify_video {"sourceGenerationId": "<clip>", "effect": "time_of_day", "prompt": "blue hour after sunset, street lamps on, wet asphalt reflecting them", "strength": "flex"}
```
## reframe_video
A new aspect ratio (`9:16`, `16:9`, `4:3`, `3:4`, `1:1`, `21:9`) with the newly exposed canvas
generated and the subject kept centred: a vertical cut of a landscape clip without letterboxing.
`prompt` optionally says what the filled areas should continue.
```
reframe_video {"sourceGenerationId": "<clip>", "aspectRatio": "9:16", "prompt": "the street continues above and below, same shopfronts"}
```
## remove_background
A clean key without a green screen. The result is the clip on black plus, in `files[]`, a luma
matte MP4 (role `matte`) and, when available, a ProRes 4444 MOV with alpha (role `alpha`).
`subjectPrompt` says what to keep when the frame has several candidates. In After Effects,
`adobe_import_generation` places the matte as the clip's track matte.
```
remove_background {"sourceGenerationId": "<clip>", "subjectPrompt": "the dancer in the red jacket"}
```
## motion_control and recast_video
Two ways to put a new character into a performance:
- `motion_control`: the character in `characterImageUrl` performs the body motion of the clip.
`orientation: "video"` keeps the clip's framing (up to 30 s), `"image"` the photo's (up to
10 s). `keepSound` keeps the clip's sound (default true). `prompt` optionally dresses the scene.
- `recast_video`: the person in the clip is swapped for the character in `characterImageUrl`,
while the performance, the camera, the background and the sound stay; the result is 1080p
(Kling 3.0 Omni) and `resolution` is ignored.
```
motion_control {"sourceGenerationId": "<dance clip>", "characterImageUrl": "<https image of the character>", "orientation": "video"}
recast_video {"libraryItemId": "<uploaded clip>", "characterImageUrl": "<https image of the character>"}
```
The character image is a Mira image or an upload (`upload_reference_image`), or pass `bloggerId`
for the user's AI Blogger. For a blogger, `blogger_motion` (modes `motion` and `recast`) adds
presets and a spoken line on top (ai-blogger.md). Several people at once, a clip up to 30 s, or a
person together with the place or a product: Miracle (miracle.md). For "this character repeats that motion"
inside a fresh render, `generate_video` with `kling-motion` is the alternative
(mira-video-prompting).
## swap_object
One element of a clip replaced from photos while everything else, the sound included, stays:
`target` is `outfit`, `product`, `location` or `text`; `imageUrls` holds 1 to 4 clean photos of
the new object (uploads or Mira images); `prompt` names what to replace and with what. Any clip
works, with or without a person in it; 3 to 10 s. Longer clips or several objects at once: Miracle
(miracle.md).
```
swap_object {"sourceGenerationId": "<clip>", "target": "product", "imageUrls": ["<https photo of the new bottle>"], "prompt": "replace the energy drink can in her hand with the green glass bottle from the photo, label facing the camera"}
```
## draw_edit_video
Draw to Edit: the user marks on ONE frame of the clip WHERE to change (circles, arrows,
scribbles, short notes) and says WHAT to change; Seedance 2.5 edit changes only that area across
the whole clip and keeps people, motion, camera, timing and the clip's sound. The marks only point
at the area - they never appear in the result. Pass the marked frame as `frameImageUrl`, or the
drawing alone as `overlayImageUrl` (a transparent PNG the size of the frame) and Mira lays it over
the frame at `frameSeconds`; both must be uploaded to Mira first. You cannot draw: ask the user
for the marked frame, or describe the area in words in `prompt` on a clean frame. `prompt` starts
with a verb (replace, remove, add, change); `referenceImageUrl` optionally shows the new look;
`resolution` `480p`, `720p` (default) or `1080p`. Uploaded clips with real people may be refused by
the model (`moderation_blocked`, refunded).
```
draw_edit_video {"sourceGenerationId": "<clip>", "frameSeconds": 2.4, "frameImageUrl": "<https frame with a red circle around the cup>", "prompt": "Replace the paper cup in the circled area with a glass of orange juice"}
```
## lipsync_video
The mouth of the speaker in a clip re-animated to a new track: an audio generation
(`audioGenerationId`, from `generate_audio`) or an uploaded track (`audioLibraryItemId`).
`model` `lipsync-2` (default), `lipsync-2-pro` for close-ups, or `kling` - the cheapest, which
lays the track once from the start and trims to the shorter of the two, so it takes only
`syncMode` `cut_off`. `syncMode` when the lengths differ: `cut_off` (default), `loop`, `bounce`,
`silence`, `remap`. Writing the speech and the voice is in mira-audio.
## avatar_video
Kling AI Avatar - talking portrait (works with a `bloggerId` too). The name is the model's: it is
not the user's character, which is an AI Blogger. A still portrait that speaks or sings a voice
track (Kling AI Avatar v2): lips, face, head and shoulders move with the audio, and the clip is
exactly as long as the track. Needs no clip at all - the source is the AUDIO:
`audioGenerationId` (a `generate_audio` kind `speech` result) or `audioLibraryItemId` (a track
from `create_upload` kind `audio`), 2 to 300 s, plus `characterImageUrl` with one clearly visible
face, front or three-quarter, or a `bloggerId`. `mode` `pro` (default, sharper) or `std` (half
the price); `prompt` optionally steers expression and mood. The price runs per second of audio,
so a long track adds up: quote `estimate_cost {"kind": "avatar", "model": "pro", "sourceSeconds":
<audio length>}` first. Only animate people the user has the right to animate.
```
avatar_video {"characterImageUrl": "<https portrait>", "audioGenerationId": "<speech generation>", "mode": "pro"}
```
When a clip of the person already exists and only the words change, `lipsync_video` is the
cheaper road. For the user's AI Blogger, `blogger_speak` runs the whole chain in one call: the
script, the blogger's voice and the portrait (ai-blogger.md).
## list_effects
Ready-made presets where the prompt is already written and the user brings a photo or a clip:
cakeify, figurine, age progression, weather change, explosion and more. Pass the id as `effect`
on `generate_image` or `generate_video`; your own `prompt` becomes a short refinement appended
to the effect's text, and the effect picks its model unless you name one. `needs` says what to
attach; `options` are set through `effectOptions`. An effect is not a style and not a playbook:
use one only when the user asks for that named effect.
## marketing_brief
A plan for a short product clip, returned in the same call: a script, a shot list, a ready
English prompt for `generate_video` (references cited as `@Image1`...), a caption, hashtags and
voiceover text. `format` is `showcase`, `review`, `unboxing` or `how_to`; describe the product
with `productUrl` and/or `productName` and `productDescription`; `referenceImageUrls` (up to 4)
become the visual references; `durationSeconds` 5 to 15; `language` for the script and caption
(the prompt is always English); `platform` as a hint. Then run `generate_video` with the prompt
and the same references, and voice the text with `voiceover_video` on the finished clip
(mira-audio). Priced with `estimate_cost` kind `marketing`.
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: 323k | 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: 323k | 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: 323k | Updated | Claude CodeCodex | Verified | Install |