TubePrompter Learn

Turn reference-video color and texture into usable AI prompts

To extract color and texture from a reference video, describe what you can see: where each color belongs, how each surface looks, and what changes as the shot moves. Then turn those observations into a short prompt you can compare against a generated result.

Visual style workflowPublished 8 min read

1. Separate four things that can look like one style

A warm, rough-looking image might contain a brown object, a textured material, orange illumination, or grain added across the whole image. Keep those observations separate. Otherwise, a request for a warm texture leaves it unclear whether the model should change the object, the light, or the entire frame.

In this workflow, color-and-texture extraction means making a descriptive prompt card. It does not recover a source LUT, identify a material with certainty, or measure the original lighting setup. When you cannot tell how a look was made, describe its visible result.

Palette

Observe: A rust-red mug is the only strong color; the wall and tabletop are gray.

Write: A muted rust-red mug on a charcoal tabletop against a pale gray wall.

Surface

Observe: The mug has small speckles and a mostly dull finish.

Write: Matte ceramic with fine dark speckles and a slightly uneven rim.

Light

Observe: The left edge is brighter; the shadow side still shows detail.

Write: Soft light from frame left, with readable detail on the shaded side.

Image treatment

Observe: Fine grain appears across both the mug and the background.

Write: Subtle image grain across the frame; the ceramic surface remains clearly defined.

Google DeepMind's Veo prompt guide treats visual style and lighting as separate creative choices. Use that distinction here: describe the surface first, then the light falling on it.

2. Inspect the beginning, middle, and end of one shot

Choose a short, continuous shot where the subject is visible. Pause near its beginning, middle, and end. If there is a cut to a new setup, make a separate card using the scene-by-scene workflow. A color that appears only during a transition is a poor anchor for the whole scene.

  1. Assign color roles. Name the dominant background, the subject color, and one accent. Attach each to an object or area instead of listing disconnected color names.
  2. Describe surfaces in plain language. Start with matte or glossy, smooth or rough, and any distinctive weave, speckle, scratch, or edge. If the material is uncertain, write what it looks like rather than insisting on a material name.
  3. Track what changes. Does a bright patch slide across a surface while the object turns? Record that separately from a painted stripe that stays attached to it. Mark ambiguous details as uncertain.
  4. Choose two invariants. For example: the mug stays rust red, and its fine speckles remain visible. These become your acceptance criteria for the next generation.

Check another frame before calling an isolated block or smear a texture. Do not build a detailed surface instruction around a feature you cannot see consistently. If a camera move changes the view, record that move separately with the camera movement guide.

3. Build a reusable color and texture card

Keep observation notes separate from the final prompt. A note can contain uncertainty; the prompt should contain only the visual decisions you want in the new scene. Replace the reference subject or setting with your own idea while carrying over selected color relationships and surface qualities.

Subject and setting: [your original scene]
Palette: [color + object or region for each role]
Surfaces: [object + visible texture + finish]
Lighting: [direction + visible effect on the surfaces]
Image treatment: [optional grain or other frame-wide look]
Motion: [one subject action and one camera behavior]
Keep consistent: [two visible properties to inspect]

Runway's text-to-video prompting guide recommends starting with the important visual and motion details, then refining as needed. Treat the card as a planning aid: your final prompt can be a few clear sentences, and optional fields can stay empty.

4. Write a prompt with something specific to check

These are original illustrative prompts, not reports of tested generations. They show three different decisions to inspect. Adapt them to your generator and compare actual outputs; the wording does not guarantee a matching result.

Matte ceramic: preserve surface detail

Close view of a muted rust-red ceramic mug on a charcoal tabletop against a pale gray wall. Fine dark speckles and a slightly uneven rim remain visible. Soft light comes from frame left; the shaded side retains detail. The mug stays still as the camera slowly moves closer. Keep the finish matte throughout the shot.

Compare and revise: Watch whether the speckles remain attached to the mug as its apparent size changes. If the finish becomes glossy, simplify the light description and repeat the matte finish instruction before adding more style terms.

Enamel and fabric: give each surface its own description

A pale cream enamel kettle rests on a folded indigo linen cloth beside a neutral gray wall. The kettle has a smooth glossy finish and a small rounded highlight; the cloth has a coarse woven texture and irregular folds. A hand lifts the kettle vertically. The cloth stays in place. Keep the cream kettle distinct from the indigo cloth.

Compare and revise: Look for the cloth becoming smooth or the kettle acquiring fabric texture. Attach each texture phrase directly to its object, then shorten any shared style sentence that blurs the distinction.

Colored light: separate the object from its illumination

A clear glass table lamp stands on a pale stone shelf in a quiet room. A warm amber bulb lights the glass from within. The wall remains muted gray, with a small warm glow immediately behind the lamp. Locked camera. The bulb gradually brightens while the stone shelf keeps its pale neutral color and fine surface variation.

Compare and revise: If the entire room turns orange, localize the warm color to the bulb and nearby glow. Judge the glass, shelf, and wall separately instead of accepting a single overall mood match.

For an advertisement, add the product action only after its surfaces are described. The product ad templates can help you choose a shot idea; keep this card focused on how the product should look.

5. Compare the result before adding more adjectives

Review the generated clip at the same three stages: beginning, middle, and end. Score the two invariants as matched, partly matched, or missing. Keep the subject, action, and generation settings as steady as your tool allows while changing one description at a time. This is an iteration log, not a controlled performance benchmark.

  • Color spreads to the wrong object: repeat the object-color pairing and remove any broad instruction that makes the whole scene the accent color.
  • A surface changes during motion: simplify the movement for the next attempt and explicitly state which finish or pattern should persist.
  • Everything looks gritty: remove frame-wide grain while checking the material texture. Reintroduce image treatment only after the object reads clearly.
  • The palette matches but the shot feels wrong: compare framing, light direction, and motion separately. A color description cannot specify the entire shot.

Start with a YouTube reference in TubePrompter

Sign in, paste a supported YouTube video URL, and run the analysis. Review its color palette, lighting mood, and visual style descriptions against the reference, then use the card above to write your own prompt. Available analysis detail depends on your plan. Treat precise-sounding technical values as AI interpretations to check, not measurements.

Follow the YouTube prompt extraction walkthrough if this is your first analysis. The card and comparison steps in this article are a manual editing workflow you can use alongside the output.

Analyze a YouTube reference

Color and texture prompting FAQ

What is color-and-texture extraction prompting for AI video?

It is a way to turn visible color relationships and surface details in a reference clip into written directions for a new video. Record which objects carry each color, how their surfaces look, and which details should remain stable during motion. Here, extraction means visual description, not recovering the original production settings.

Should I put hex color codes in a video prompt?

You can keep codes in your own design notes, but do not assume that a video model will reproduce them exactly. Start with a named color, its role, and its location: for example, a muted rust-red mug against a charcoal background. Check the generated result before treating it as a color match.

Can TubePrompter identify the original material or color grade?

TubePrompter can provide AI-generated descriptions of a supported YouTube reference, including color and visual style. Treat material labels and technical settings as interpretations to review. They do not establish the original material, camera settings, or grading process.