Przejdź do treści
Biblioteka skilli
Projektowanie i UX · Budowanie i kod

storyboard

/storyboard

Rysuje storyboard sceny z postaci referencyjnej w FLUX 3 Klein, a potem animatik z dźwiękiem roboczym, zanim wydasz na wideo.

Licencja
Apache-2.0
Ostatnia zmiana repozytorium
05.11.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść25 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
3
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 05.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

Co robi i kiedy

Rysuje storyboard sceny z postaci referencyjnej w FLUX 3 Klein, a potem animatik z dźwiękiem roboczym, zanim wydasz na wideo.

Oryginalny opis (EN)

Use when a dw MCP server is connected and the user wants a storyboard - a set of boards (stills) for a scene drawn on FLUX 3 Klein from a character reference, a coverage set of one moment from six camera positions, boards used as the first frames of H3, LTX-2.5 or Kandinsky 6 shots, or a timed animatic of the boards with a scratch voice and music bed before any video is paid for.

Jak działa

  1. Before anything
  2. 1. Draw the boards (Popcorn shape)
  3. 2. Coverage of one moment (Shot Board shape)
  4. 3. A board as a start frame
  5. 4. Animatic
  6. Run and judge
  7. Not in scope
  8. Sources
Sekcje z nagłówków SKILL.md, w oryginale · 100 linii, 756 słów
Podgląd SKILL.md
---
name: storyboard
description: Use when a dw MCP server is connected and the user wants a storyboard - a set of boards (stills) for a scene drawn on FLUX 3 Klein from a character reference, a coverage set of one moment from six camera positions, boards used as the first frames of H3, LTX-2.5 or Kandinsky 6 shots, or a timed animatic of the boards with a scratch voice and music bed before any video is paid for.
---

# Storyboards on a dw server

A board is a still that says what a shot will look like before a video model
is paid for. This skill draws the boards on FLUX 3 Klein, builds a coverage
set of one moment, hands boards to a video family as start frames, and cuts
them into a timed animatic. It does not pick a video family's template or
size: that family's skill does, and its rules win. Camera and look words come
from `cinematography`; the shot list's fields from `script-to-video`.

## Before anything

1. `get_server_info`: Klein needs a CUDA card (`flux3-klein`'s rules apply:
   16 steps, no guidance, sides in multiples of 16, gated license).
2. Boards must be on the server: `keep_output(name, asset_name="boards/<scene>-NN")`
   after each draw. Leave the extension off - the engine adds the file's own
   (Klein writes `.jpg`) - and cite `asset:boards/<scene>-NN.jpg` later.
3. All boards of one animatic share one size: keep the same first reference,
   whose aspect ratio sets the board's, for every board of the scene.

## 1. Draw the boards (Popcorn shape)

Higgsfield's Popcorn (the concept only; no code is borrowed) draws a whole
scene's boards from reference images in one pass. Here that is
`templates/flux3/multi-reference`: `images` is the character and location
references in order, and the `instruction` names each one by position, "image
one" the character, "image two" the location. It draws up to eight boards
per scene. Call the template once per board and keep the same references and
`seed` so the cast and place hold; the instruction changes the camera and the
action only, in `cinematography`'s `shot_size`, `move` and `light` words.

- One idea per board: the single state a frame can show, as in the shot list.
- Past eight boards, split the scene. A scene that needs more has several
  beats and wants a second reference set.
- A board is a draw, not a promise. Look at it (`get_output_image`) against
  the reference before it goes anywhere else.
- To change one board and hold the rest, `templates/flux3/image-edit` with
  an instruction naming only the change.

## 2. Coverage of one moment (Shot Board shape)

Shot Board (MIT) names six cuts that cover any moment: establishing, medium,
face close-up, over-the-shoulder, insert and low angle. Pick the hero board
first - the one that best shows the moment - then redraw the other five from
it with `templates/flux3/multi-reference` or `image-edit`, the hero as
"image one". `references/coverage-set.md` has the six, what each frames, and
the instruction pattern. The eight-board cap holds across the set.

## 3. A board as a start frame

A board that passes review becomes a shot's first frame. The video family's
skill picks the template and size; the board must match that size, so resize
it first rather than letting the pipeline crop a face off.

| Family | Template | Board goes in |
| --- | --- | --- |
| H3 | `templates/minimax/image-to-video` | `image` |
| LTX-2.5 | `templates/ltx2/image-to-video` | `image` |
| Kandinsky 6 | `templates/kandinsky6/image-to-video` | `image` |

Write that shot's prompt to say what happens next, not what the board
already shows. The board fixes the first frame only; later frames may drift
from it, so check the last frame (`get_output_frames`) against the board.
After the shots exist, `templates/face-drift-check` measures whether the
character's face held.

## 4. Animatic

Boards, a scratch voice and a music bed cut into one timed clip, cheaply, to
judge pacing and order before video. `references/animatic.md` has the recipe:
`templates/generate-speech` for the voice, `templates/minimax/music` for the
bed, then `templates/animatic` with the boards, one duration per board and
the soundtrack. Change a duration and re-run: it costs seconds, not GPU
hours.

## Run and judge

`validate_workflow`, quote `plan.estimate`, `run_workflow`, `wait_for_job`,
`get_job`. A template's `subfolder` sorts what it writes: `final` for the
deliverable (the animatic), `intermediate` for scratch. Boards you keep as
assets are the deliverable of the drawing pass; set `"role": "intermediate"`
on a board that is only a step toward a hero. Read a project's shot list with
`get_prompt("<project>/shots")` (`list_prompts` finds the project). Judge boards
with `get_output_image` and the animatic with `get_output_frames`.

## Not in scope

- Video generation, the cut and the score of the final film: the family
  skills, `script-to-video` and `series-episodes`.
- Layout by box from a short request (`flux3-klein`'s `layout-boxes` is by
  hand).
- Sound timed to the picture: the animatic holds boards, it does not align
  them to audio.

## Sources

`dw:flux3-klein`, `dw:cinematography`, `dw:script-to-video`; Higgsfield
Popcorn (concept only) and Shot Board (MIT) for the two shapes;
`workflows/templates/animatic.json`; `docs/TASKS.md` (`stills_to_video`).
Pliki w paczce (3)
  • SKILL.md 5 KB
  • references/animatic.md 2 KB
  • references/coverage-set.md 2 KB

Wersje

RepozytoriumWynikLicencja
Ta strona · dkackman/diffusers-workflow90Apache-2.0
buriedsignals/splash100MIT
veevia/design-agency100MIT
citypaul/.dotfiles100MIT
getcrew44/crew44100MIT
matheussilvano/videomaker96MIT
savvides/idstack95MIT
openjiuwen-ai/jiuwenswarm82Apache-2.0

Migawka bazy z 13.11.2026 · wersja paczki 78251461b83f (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 90storyboard-animatic
    Zamienia scenę w storyboard z kadrami kluczowymi i animatik w ruchu w NovaField, z zatwierdzaniem plansz i cięciem do czasu.
  • 100flux3-klein-stills
    Generuje kadry kluczowe i obrazy referencyjne modelem FLUX Klein o otwartych wagach przez ComfyUI.
  • 96flux-klein-9b-prompt-writer
    Pisze prompty do generowania obrazów w modelu FLUX.2 klein 9B.
  • 90flux3-klein
    Generuje obraz modelem FLUX 3 Klein przez serwer MCP: od krótkiego opisu, przez edycję zdjęcia, po łączenie referencji i układ elementów.
  • 95ttapi-flux-image
    Generuje, edytuje i łączy obrazy modelami FLUX przez TTAPI, w tym Kontext i Klein, także z wielu zdjęć referencyjnych.
  • 100cinematography
    Podpowiada zamknięte słownictwo kamery, światła i gradacji dla generowania wideo w modelach H3, LTX-2.5, Kandinsky 6 i FLUX 3 Klein.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.