Best PromptsMulti-Shot Character Consistency
Keep one character looking like the same person across separately generated shots using a locked description block and fixed style spine.
Generate a sequence of shots featuring the same character. **Step 1: write the character lock.** This exact text goes into every shot prompt, word for word, before anything else. Never paraphrase it. CHARACTER LOCK: "[NAME], a [AGE] year old [DESCRIPTION: build, height impression], with [HAIR: exact length, color, texture, and how it is worn], [FACE: skin tone, exactly two or three distinguishing features such as a scar above the left eyebrow, deep-set brown eyes, a wide jaw], wearing [WARDROBE: every visible garment with its color, material, and fit, plus any accessory that appears in more than one shot]." Keep the lock under 60 words. A longer description is honored less consistently, not more, and the details at the end are the first to drift. **Step 2: write the style spine.** Also identical in every shot: STYLE SPINE: "[FILM LOOK: e.g. shot on 35mm anamorphic, shallow depth of field, filmic grain]. [PALETTE AND GRADE: e.g. warm practicals against cool shadows, lifted blacks]. [LIGHTING LOGIC: e.g. motivated by a single window source from camera left]." **Step 3: write each shot** as: CHARACTER LOCK + STYLE SPINE + this shot only. Shot [N]: - Shot size and angle: [WIDE / MEDIUM / CLOSE-UP, EYE LEVEL / LOW / HIGH] - Location and time: [WHERE, WHEN] - Action, one only: [WHAT THE CHARACTER DOES, in countable terms: "picks up the cup, drinks, sets it down"] - Camera: [ONE MOVE OR STATIC] - Audio: [SFX, AMBIENCE, OR ONE LINE OF DIALOGUE] - Duration: [SECONDS] Rules across the sequence: - One action and one camera move per shot. Two of either is two shots. - Vary shot size between consecutive shots, and keep screen direction consistent so the character does not appear to reverse between cuts. - Change only the shot block between prompts. Every change to the lock or the spine is a chance for the character to become a different person. - Anything that must recur (a prop, a wound, a stain, a rolled sleeve) belongs in the lock, not in a shot description. Produce [N] shots covering: [THE BEATS OF THE SCENE].
How to use
Character drift between shots is the main reason AI sequences do not cut together, and it comes almost entirely from rewriting the description each time. Treat the lock as fixed text you paste rather than prose you compose. Keeping it short is counterintuitive and correct: a sixty-word lock is honored more reliably than a two-hundred-word one, so spend the words on two or three distinctive features rather than a full portrait. Where your tool offers a reference image or a cameo feature, use it in addition to the lock, not instead of it.
More video prompts
[Attach your first frame image, and the last frame image if your tool supports it, then use this prompt:] Animate from the first image to the last image. What changes between them: [THE ONE TRANSFORMATION, e.g. "the closed box opens and the product rises out of it" / "the empty street fills with morning light" / "the character turns fro
First and Last Frame Control
Direct a video model between two fixed images so the motion, timing, and transition land exactly where you need them for an edit.
Write and shot-list an explainer video. What we are explaining: [PRODUCT, FEATURE, OR CONCEPT] Who is watching and what they already know: [AUDIENCE + BASELINE] The one thing they should understand afterward: [THE SINGLE TAKEAWAY] What they should do next: [THE ACTION] Length: [60 / 90 / 120 seconds] Visual style: [SCREEN RECORDING / MOT
Explainer Video
Turn a product or concept into a 60 to 90 second explainer with a script timed to shots, so the visuals carry what the words cannot.
A repeatable pipeline for a faceless channel. Run each step, check the output, then paste it into the next. **Step 1: The script.** ``` Write a [60 / 90 / 180] second script about [TOPIC] for a [NICHE] channel. Audience: [WHO WATCHES AND WHY] Channel voice: [e.g. "calm and factual" / "fast and irreverent"] Format: [top-list / single st
Faceless Channel Pipeline
A four-step chain from topic to finished faceless video: script, shot list, generation prompts, and the metadata to publish it.