AI video prompt builder

Pick a character, a format, a camera and a vibe, type what happens, and get a structured video prompt you can paste into Seedance, Kling, Veo or any reference-to-video model. Free, no login, nothing leaves your browser.

1. Character

MalfoidTheaOdette+ Your own

2. Format

Phone-POV livestreamDance to a songTutorial / how-toStorytimeReactionWalk-in trendCinematic scene

StyleInfluencerCinematicSakuga AnimeGlam EditorialRetro VHSCozy VlogDark Fantasy3D ToonCameraAutoSelfie / LiveClose-upWideHandheldPush-inChaseLow AngleOverheadHeldVibeChillHypeDreamyEerieMelancholyRomanticEpicComedyFuryLength5 s10 s15 s

3. What happens (one line)Spoken line (optional, goes in quotes)

Your prompt

Copy prompt

References:
@Image1: Malfoid. Keep this face, hair and outfit exactly.

Visual style: Photoreal social-media footage shot on a modern phone camera, natural skin texture with visible pores and fine peach fuzz, true-to-life fabric detail, real-world lighting, slight handheld micro-shake, crisp autofocus and authentic smartphone colour science. Vertical 9:16 frame.

Avoid: No subtitles, captions or on-screen text. No logos, watermarks or interface overlays. The camera itself is never visible: no phone, phone case or hand holding a phone in frame. No slow motion. No streaming-interface overlays, chat bubbles, hearts or viewer counts. No change to any face or outfit.

Malfoid (@Image1): A 19-year-old young woman: very long straight platinum-blonde hair worn loose past the waist with a centre part and soft face-framing strands, pale porcelain skin, grey-green eyes with a heavy-lidded haughty gaze, straight dark-blonde brows, a small straight nose, full pouty lips and a delicate jaw, small silver hoop earrings, slender figure. Wears a black wool school robe with an emerald-green silk-lined hood, a charcoal-grey V-neck knit vest over a crisp white collared shirt, an emerald-and-silver striped tie, a charcoal pleated skirt and sheer black tights. Sneering, proud, secretly soft-hearted. Camera: One continuous take from the character's own phone held at arm's length, front-camera selfie framing, the arm and slight wobble of a real handheld phone, the character talks straight into the lens like a livestream, no cuts. Pacing: a setup beat, a hard pause, then an oversized reaction. Light: bright even daylight, clean saturated colour, no dramatic shadow. Spoken lines are lip-synced, with matching mouth movement. Sound: quiet room tone, a notification ping, fabric rustle.

Timeline (10 s, Phone-POV livestream):
0-3s: Malfoid looks straight into the lens and opens the stream mid-conversation with the chat.
3-7s: Malfoid reads her hate comments out loud and claps back at each one, saying: "Reading my hate mail. Live."
7-10s: The turn: something goes wrong, or the character softens for one second, then recovers their cool.

End with action.

2159 characters · 3 timed beats · works as a starting point for Seedance, Kling, Veo and other text- or reference-to-video models.

Cast's one-paragraph version (what Cast would send without its AI writer)

Photoreal social-media footage shot on a modern phone camera, natural skin texture with visible pores and fine peach fuzz, true-to-life fabric detail, real-world lighting, slight handheld micro-shake, crisp autofocus and authentic smartphone colour science, vertical 9:16 frame. Malfoid (@image1): A 19-year-old young woman: very long straight platinum-blonde hair worn loose past the waist with a centre part and soft face-framing strands, pale porcelain skin, grey-green eyes with a heavy-lidded haughty gaze, straight dark-blonde brows, a small straight nose, full pouty lips and a delicate jaw, small silver hoop earrings, slender figure. Wears a black wool school robe with an emerald-green silk-lined hood, a charcoal-grey V-neck knit vest over a crisp white collared shirt, an emerald-and-silver striped tie, a charcoal pleated skirt and sheer black tights. Sneering, proud, secretly soft-hearted. Keep every face, hairstyle and outfit exactly as in the reference images. Action across 10 seconds in 3 quick, readable beats: reads her hate comments out loud and claps back at each one, saying "Reading my hate mail. Live.", each beat landing with clear body movement and expressive faces while hair and fabric swing and the environment reacts. Camera: one continuous take from the character's own phone held at arm's length, front-camera selfie framing, the arm and slight wobble of a real handheld phone, the character talks straight into the lens like a livestream, no cuts. Pacing: a setup beat, a hard pause, then an oversized reaction. Light: bright even daylight, clean saturated colour, no dramatic shadow. Real-time full-speed motion, no slow motion. No on-screen text, subtitles, captions, logos, watermarks or interface overlays. End with action.

Copy paragraph

Your first video in Cast is free at 5 seconds; longer ones use credits.

Make it in Cast

Why a structured prompt

Video models follow a timeline far better than a paragraph of adjectives. Every video in our prompt library was made from the same skeleton, and this tool writes that skeleton for you:

1. References. One line per reference image (@Image1, @Image2…), and the song as @Audio1 for dances. A character with a reference image keeps its face; one described only in text gets a new face every time.
2. Visual style. One sentence for the look. The tool uses the same style clauses as Cast's own prompt writer.
3. Avoid. No subtitles, logos or interface overlays, no phone in the shot, no slow motion, no change to the face. These are the artifacts we hit most often.
4. Cast, camera, pacing, sound. Who is in it (copied verbatim from their description), where the camera is, how the beats move, and three or four sounds.
5. Timeline. Timestamped beats sized to the clip: about one beat per three seconds. The first beat names the format, the middle carries your action and the quoted line, and the last is a turn.

The formats

• Phone-POV livestream: opens when the character looks straight into the lens and opens the stream mid-conversation with the chat; ends with the turn: something goes wrong, or the character softens for one second, then recovers their cool.
• Dance to a song: opens when the character strikes a pose and hits the first move exactly on the beat; ends with a final pose on the last beat and a knowing look into the lens.
• Tutorial / how-to: opens when the character holds the first item up to the lens and names it; ends with reveals the result to the lens, with a twist at the end.
• Storytime: opens when the character leans toward the lens as if sharing a secret; ends with delivers the punchline and holds the reaction.
• Reaction: opens when the character looks at something just off-screen, unimpressed; ends with an oversized reaction straight into the lens.
• Walk-in trend: opens when the character walks toward the low camera in step with the beat; ends with snaps back into formation and stops right at the lens.
• Cinematic scene: opens when the character is revealed in the environment, already in motion; ends with lands the final beat with a clear, readable pose.

Example output

Malfoid, Phone-POV livestream, Influencer style, 10 seconds, action “reads her hate comments out loud and claps back at each one”, line “Reading my hate mail. Live.”:

References:
@Image1: Malfoid. Keep this face, hair and outfit exactly.

Visual style: Photoreal social-media footage shot on a modern phone camera, natural skin texture with visible pores and fine peach fuzz, true-to-life fabric detail, real-world lighting, slight handheld micro-shake, crisp autofocus and authentic smartphone colour science. Vertical 9:16 frame.

Avoid: No subtitles, captions or on-screen text. No logos, watermarks or interface overlays. The camera itself is never visible: no phone, phone case or hand holding a phone in frame. No slow motion. No streaming-interface overlays, chat bubbles, hearts or viewer counts. No change to any face or outfit.

Malfoid (@Image1): A 19-year-old young woman: very long straight platinum-blonde hair worn loose past the waist with a centre part and soft face-framing strands, pale porcelain skin, grey-green eyes with a heavy-lidded haughty gaze, straight dark-blonde brows, a small straight nose, full pouty lips and a delicate jaw, small silver hoop earrings, slender figure. Wears a black wool school robe with an emerald-green silk-lined hood, a charcoal-grey V-neck knit vest over a crisp white collared shirt, an emerald-and-silver striped tie, a charcoal pleated skirt and sheer black tights. Sneering, proud, secretly soft-hearted. Camera: One continuous take from the character's own phone held at arm's length, front-camera selfie framing, the arm and slight wobble of a real handheld phone, the character talks straight into the lens like a livestream, no cuts. Pacing: a setup beat, a hard pause, then an oversized reaction. Light: bright even daylight, clean saturated colour, no dramatic shadow. Spoken lines are lip-synced, with matching mouth movement. Sound: quiet room tone, a notification ping, fabric rustle.

Timeline (10 s, Phone-POV livestream):
0-3s: Malfoid looks straight into the lens and opens the stream mid-conversation with the chat.
3-7s: Malfoid reads her hate comments out loud and claps back at each one, saying: "Reading my hate mail. Live."
7-10s: The turn: something goes wrong, or the character softens for one second, then recovers their cool.

End with action.

The full 30-second version of this idea, with the exact prompt that made it, is Malfoid reads her hate mail live.

Tips that matter more than wording

• Leave out real people's names, brands and film or book franchises. Video models' copyright filters block them, often after the full render. See why Seedance blocks videos.
• A 5-second clip holds one action and a reaction; 10 seconds holds three beats; 15 seconds about five.
• Put spoken words in quotes. Models with native audio lip-sync quoted lines.
• For dances, attach the song as an audio reference and delete any “no background music” line.
• More in the Seedance prompting field guide and how to write an AI video prompt.

What this tool is and is not

It is a deterministic template: the same picks always give the same prompt, and no AI model is called. Inside the Cast app, an AI writer expands your line into a denser prompt and renders it with Seedance; your first 5-second video there is free with no sign-up. The tool does not generate video by itself.

Updated 2026-10-17