Skip to content

moonlight-lupin/agent-skills / creative/clips-studio

clips-studio

Verified

Use when the user wants to "make a video/clip", "generate a video from text", "animate this image/photo", "create a marketing reel / social video / teaser", "do a 3D / parallax move", "pan/zoom/orbit a shot", or mentions fal.ai, Kling, Veo or Seedance for video. Staged fal.ai workflow — brainstorm, draft cheaply, produce the final at quality. Three modes: text-to-video, animate (still to motion), camera-move (push-in/pan/orbit). Not for still images (image-studio), slide decks, or charts. Honesty discipline: never depict real identifiable subjects via text-to-video.

FileKindSize
README.mdmarkdown1.9 KB
SKILL.mdskill md23.5 KB
examples/example-run.mdmarkdown8.1 KB
references/fal-video-models.mdmarkdown11.9 KB
references/motion-honesty-guide.mdmarkdown11.4 KB
references/video-prompting-techniques.mdmarkdown3.4 KB
scripts/falvid.pyscript34.5 KB
tests/__init__.pyscript0 B
tests/test_falvid.pyscript31.3 KB

examples/example-run.md

View on GitHub
# Worked example — animate a space, a building push-in, a product clip and text-to-video

A fictional, illustrative run (no real people/places). It shows all three modes plus the
image→video combo. The work folder holds a clean photo of an empty lounge (`lounge.jpg`) and a
building shot (`building.jpg`).

## Stage 1 — brainstorm (chat, no egress)

Agreed briefs (saved to `_workings/brief_demo.md`):

- **Lounge (`animate`):** the room is **empty**, so populate first, then animate. Motion: *"two
  people on the sofa mid-conversation, one gesturing; a third walks past in the background; soft
  afternoon daylight, gentle and natural."* 5s, 16:9, **no audio**.
- **Building (`camera`):** *"slow, gentle dolly push-in toward the entrance; smooth and steady."* 5s,
  no audio. Gentle by design (see `motion-honesty-guide.md`) — a push-in invents almost no geometry.

Confirmed the start images and that egress to fal.ai is OK for these (non-confidential marketing
imagery).

## Stage 2 — populate the empty lounge (in image-studio), then draft cheaply

**Populate** the empty room with people using `image-studio` (it edits the still; this skill does the
video):

```bash
# in image-studio (run from the clips-studio skill directory, or cd to image-studio first)
python ../image-studio/scripts/falgen.py edit \
  --prompt "Add two people sitting on the sofa mid-conversation and one walking past — natural, relaxed, believable. Keep the room, furniture, finishes and daylight exactly as they are." \
  --image _workings/lounge.jpg \
  --model fal-ai/nano-banana-pro/edit --arg resolution=2K \
  --out-dir _workings --name lounge_populated
```

Review `_workings/lounge_populated.png`. Then **draft the motion cheaply** here (Kling 2.5 Turbo Pro,
~$0.35 for 5s):

```bash
python scripts/falvid.py animate \
  --prompt "Two people on the sofa mid-conversation, one gesturing; a third walks past in the background; soft afternoon daylight, gentle natural motion." \
  --image _workings/lounge_populated.png \
  --duration 5 --aspect 16:9 \
  --out-dir _workings --name vid_lounge_draft \
  --run-log _workings/run-log_demo.md
```

Watch the draft for warping faces/hands and unnatural sliding. If the walking figure morphs, drop to
two people and re-draft — cheaper than fixing it on the quality model.

For the **building**, draft the camera move (Seedance default, ~$0.26). Check the source orientation
first and pass `--aspect` to match it:

```bash
python scripts/falvid.py camera --move push-in \
  --image _workings/building.jpg --duration 5 --aspect 16:9 \
  --out-dir _workings --name vid_building_draft
```

Check it doesn't invent a new entrance or windows. If it drifts, add `--static` or shorten the move.

## Stage 3 — produce the finals (confirm cost first)

Motion approved. Put the spend to the user: *"Lounge final as a 5s Veo 3.1 clip at 1080p, no audio,
~$1.00; building stays on Seedance, ~$0.26 — go ahead?"* On the go-ahead:

```bash
# Lounge — quality people motion on Veo 3.1
python scripts/falvid.py animate \
  --prompt "Two people on the sofa mid-conversation, one gesturing; a third walks past; soft afternoon daylight, gentle natural motion." \
  --image _workings/lounge_populated.png \
  --model fal-ai/veo3.1/image-to-video \
  --duration 5 --resolution 1080p --aspect 16:9 \
  --out-dir . --name vid_lounge_final \
  --run-log _workings/run-log_demo.md

# Building — keep Seedance, the camera default
python scripts/falvid.py camera --move push-in \
  --image _workings/building.jpg --duration 5 --resolution 720p --aspect 16:9 \
  --out-dir . --name vid_building_final \
  --run-log _workings/run-log_demo.md
```

Outputs: `./vid_lounge_final.mp4` and `./vid_building_final.mp4`. Then:

```bash
python scripts/falvid.py costs    # session breakdown — tell the user the total
```

Hand over the file paths and the estimated total, and remind the user: these are **illustrative
drafts** for review — not published material, and not a literal record of a real place. If publishing,
suggest an "illustrative" note in the post.

---

## Combo example — a vertical photo for a 16:9 hero (image-studio → video)

The user has a **portrait phone shot** of a building (`building_portrait.jpg`, ~3:4) but wants a
**16:9** clip for a website hero. The target frame is wider than the source — so run the **Stage-1
aspect & orientation check** first.

**1. Flag the mismatch and offer the honest routes.** *"Your photo is vertical but the hero is 16:9.
Best is a landscape photo — can you supply one? If we must use this vertical shot, I can (2) pillarbox
it (bars, nothing invented) or (3) generatively extend it to 16:9 — which invents the surroundings.
Which?"* (Force-cropping to 16:9 is **off the table** — it would slice the top and bottom off the
building.)

**2. User picks (3). Ground it before inventing.** *"To make the extension reflect the real street,
drop in any reference images — a street-view, the neighbouring buildings, a wider shot — so I extend
to what's actually there, not a guess."* Use those to steer the outpaint.

**3. Outpaint the still to 16:9** (cheap image step, in `image-studio`):

```bash
python ../image-studio/scripts/falgen.py edit \
  --prompt "Extend this photo sideways to fill a 16:9 frame. Keep the building EXACTLY as it is; continue the street, sky and the real neighbouring buildings (see references) naturally to left and right. Photorealistic, consistent dusk light." \
  --image _workings/building_portrait.jpg \
  --model fal-ai/nano-banana-pro/edit --aspect 16:9 --arg resolution=2K \
  --out-dir _workings --name building_16x9_extended
```

**4. CONFIRM the extended still with the user — before any video spend.** Show the extended PNG, take
corrections (*"there's a building on the left, not empty space — fill it in"*, *"trim the extra sky"*),
re-edit on the cheap image model until they approve. **Only then** animate — never animate an
unconfirmed invented frame.

**5. Animate the approved 16:9 frame** (pass `--aspect 16:9` so it isn't re-cropped):

```bash
python scripts/falvid.py camera --move push-in --aspect 16:9 \
  --prompt "Very gentle slow push-in; the building stays exactly as it is; subtle motion only in the sky." \
  --image _workings/building_16x9_extended.png --duration 5 \
  --out-dir . --name building_hero_final --run-log _workings/run-log_demo.md
```

All-in ~$0.41 (≈$0.15 extend + ≈$0.26 clip). Output is **illustrative** — the extended surroundings
are invented and now in motion, so it can never stand as a record of the real street; label it.

> **Aspect lesson:** the video models do **not** reliably inherit the source's framing — Seedance
> defaults to **16:9** and **centre-crops** a portrait/square still (you'll see only the middle band).
> Always pass `--aspect` to match the still: a 3:4 source → `--aspect 3:4`.

---

## Other quick examples (general subjects)

The studio isn't space-only — the same staging works for any clip.

**Product `animate`** — bring a product photo to life (draft cheap, then Veo 3.1 final):

```bash
python scripts/falvid.py animate \
  --prompt "The bottle slowly rotates on the plinth; soft studio light sweeps across the label; the background stays still. Keep the product exactly as it is — do not alter the label, shape or colour." \
  --image _workings/product.png --duration 5 --aspect 1:1 \
  --out-dir _workings --name product_draft
```

Keep object/hand interactions out (a hand reaching for the bottle will fumble — see the hands ceiling
in `motion-honesty-guide.md`). Never change the product's real label/shape/colour.

**Text-to-video `generate`** — an abstract brand-mood clip with **no source image** (cheap Wan draft
→ Veo 3.1 final):

```bash
python scripts/falvid.py generate \
  --prompt "Soft abstract flowing gradient in deep burgundy and warm neutrals, slow elegant motion, premium and calm; no text, no logos, no recognisable place or person." \
  --duration 5 --aspect 16:9 \
  --out-dir _workings --name brandmood_draft
```

`generate` invents the whole scene, so keep it **abstract/fictional** — never use it to depict a real
place, a real product or a real person (use `animate` on a real photo for those).
canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall