The Master Guide / 04 The treatment and presentations · Writing the treatment
Storyboard images
Every image in a treatment shows the vision. None of them controls a generation.
Every image in a treatment shows the vision. None of them controls a generation.
Every image in a presentation is a storyboard image or an illustration, never a start frame (decided). The finished shots need not contain these frames. A storyboard is never attached to any generation, in any role (decided): it is read and its content written in words.
How many [best available, not yet validated]:
Section | How many | Generated or sourced |
Cover, vision, close | One hero each (one or two for the vision) | Generated: they show the film |
Section openers | One each | Sourced by default; it sets that section's mood |
Story at a glance | One per scene, the end card last | The lead frame of each scene page |
The film, scene by scene | 1–3 per scene. A second or third only where the turn, a reaction or the product beat needs its own picture, or where a camera idea is shown as start, turn and landing frames. Never one frame per shot | Generated frames lead [proven]; a sourced still may lead a scene whose feeling is the point |
Overflow page after a scene | 2–4 | As the scene |
Casting | The chosen face's casting still per role, plus a performance frame where it helps | Generated only; never a famous face |
Wardrobe and locations | Each wardrobe look on your frame, and 1–3 views per place under its scene's light | Generated; a sourced still for a kind of place or texture |
Product | The hero states, the label moment and the end card | Generated from the client's photos |
Camera, light and colour, and the edit | 1–3 | A sourced still is often clearest for a quality, beside a generated frame of it |
Music and sound | Two or three | Generated or sourced |
Typography | The end card and text moments, in live type | Type is always set live |
Versions | One | Generated or sourced |
Style (non-photoreal) | The style owner frame | Generated |
A treatment usually carries 25–35 frames.
Generated frames [proven] show what only this film has: its people, looks, product and label, the specific moments of its scenes.
Sourced stills come from Frameset, ShotDeck, FILMGRAB and award-level human-made work, never AI-made (see References → Sourced stills on pages).
Named exceptions
fully generated, when the world is one no film has shown;
mostly sourced, when frames can't be made in time (generated frames still cover the cover and the key product moment);
stylised films, in their own style;
motion-graphics-led films, with frames rendered from the code design;
an agency board, redrawn in the film's look, its order and beats kept.
Captions
Story at a glance: a few words and the scene's seconds ("The armchair · 11 s").
Scene pages: under each frame, one "What we see" line of 12 words or fewer, plus the exact words on screen if any. Dialogue stays in the paragraph. At most one plain camera line per scene.
Overflow page: the scene number and time, one or two plain lines and the exact words on screen; 40 words per frame and 120 per page at most.
Department pages: short plain captions about the choice, never lens numbers or trade terms.
Sourced stills: your decision in the still's language ("Our light: one window, deep shadows, faces half lit"); never "like Film X", and no credit line (see References → Sourced stills on pages).
Casting: the character's name and one or two lines on who they are; never "frame study".
Never on any frame: IDs, file names, tool or model names, "start frame", "draft", arrows or camera drawings.
A repeated frame is a designed event, and its caption says so.
Making them
Estimate the frame batch first: about two attempts per image, three with a face, at about 2 Higgsfield credits per image at 2K and 4 at 4K.
Write the treatment's look block once from the department sheets and paste it unchanged into every frame (see Still prompts, storyboard frames and diagrams → Text, sourcing and the look block). Attach the chosen identity close-ups; don't re-describe the faces. No earlier frame is attached to a later one.
Every frame agrees with the sheets and the paragraph: names, wardrobe looks, places, product count, label side and text, light.
Check each frame at full size: hands and touch points are a hard stop; faces, product shape and label are checked too (see The quality bar → The hard stops). A near-miss gets a model edit or a new image; never a pixel fix by hand (decided).
Sample the hex values from the approved frames into the sheets (steps below).
The optional animatic and mood film
Optional still animatic, after the story at a glance, when timing is the point: the storyboard frames held for their scenes' durations, with temp audio or a scratch read.
Optional mood film (decided), only where the feeling is the point (an emotional, music-led or world-led film): 20–40 s cut from other films' footage to temp music or a scratch read. It shows the feeling, never your look or a shot to match. Its clips are award-level human-made work, gathered on the stage's one download list (see References → Which layer each stage studies), and its page carries one plain line saying it is made from other films. Your own video still waits for the anchors approval.
Do it yourself
Sample a hex value from an approved frame (Photoshop, or Digital Color Meter on a Mac)
Open the approved frame at full size.
In Photoshop, take the eyedropper, click on the item's colour, and read the hex value in the colour picker. On a Mac, open Digital Color Meter with its values shown as hexadecimal, and hold the pointer over the item.
Write the hex value into the item's sheet, after its colour name.
Sources
Sean Baker on video lookbooks that play like a trailer (Filmmaker Magazine); Smuggler's treatment team and its mood films (APA job ad); Bianca Poletti on mixing film and non-film imagery (Free the Work).