HeyGen Video 1.0

HeyGen Video 1.0 is HeyGen's video generation model for short clips with synchronized audio. It generates the full frame from a text prompt, a first-frame image, or up to 12 image, video, and audio references, and produces dialogue, ambience, and sound effects in the same pass, with no separate lip-sync step. It outputs 5 to 15 second clips at 480p or 768p, suited to presenter shots, product and training content, and social video.

Complete technical specification for integration
Ready-to-use code snippets for common workflows
Step-by-step tutorials for advanced use cases
← All GuidesWriting long, specified prompts
How to write prompts for HeyGen Video 1.0, from naming the camera and ruling out the model's defaults to staging contact, short lettering and prompt enhancement.
Introduction
HeyGen Video 1.0 builds the whole shot from a description: the subject, the setting, the light and the sound, all in one pass. It follows the brief literally instead of improvising around it. What you write lands, and what you leave out is filled in by the model's defaults.
That makes length an asset rather than a risk. A few hundred to a few thousand characters gives noticeably better output than a one-line idea, and the prompt field takes up to 32,000. The clip below comes from a prompt of about 1,100 characters, and every one of them made a decision the model would otherwise have made for you.
Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue.
This guide covers how a full brief is built, the camera language the model responds to, the clauses that rule out its defaults, how to stage contact between objects, how to get short lettering right, how to fill a longer clip, and the prompt enhancement setting that decides whether your words reach the model unchanged.
The request
A text-to-video call needs only the prompt. Everything else has a working default.
import { createClient } from '@runware/sdk'
const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()
const [result] = await client.run({
model: 'heygen:video@1.0',
positivePrompt: 'Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue.',
width: 1344,
height: 768,
duration: 8
})import asyncio
import os
from runware import Runware
async def main():
async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
results = await client.run({
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue.",
"width": 1344,
"height": 768,
"duration": 8
})
asyncio.run(main())curl https://api.runware.ai/v1 \
-H "Authorization: Bearer $RUNWARE_API_KEY" \
-H "Content-Type: application/json" \
-d '[
{
"taskType": "videoInference",
"taskUUID": "5d1c8e2a-3f47-4b9e-a2c6-7e0f9b3d4a18",
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue.",
"width": 1344,
"height": 768,
"duration": 8
}
]'runware run heygen:video@1.0 \
positivePrompt="Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue." \
width=1344 \
height=768 \
duration=8{
"taskType": "videoInference",
"taskUUID": "5d1c8e2a-3f47-4b9e-a2c6-7e0f9b3d4a18",
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod at chest height. A woman in her early fifties stands at the rail of a passenger ferry crossing a gray northern harbor on an overcast morning. She wears a rust-orange waxed cotton rain jacket zipped to the collar, a charcoal knit beanie with her hair tucked fully under it, and dark jeans. Her face has real texture: fine lines around the eyes, wind-reddened cheeks, no makeup. She rests both forearms on the white-painted steel rail and looks out across the water, then turns her head toward the camera and gives a small, easy smile. Behind her the harbor falls softly out of focus: pale water, a low line of warehouses, a flat gray sky. Muted, natural color under soft, even daylight. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names, printed words or badges anywhere in frame. Audio: the steady low hum of the ferry engine, water slapping against the hull below, wind across the open deck, and two distant gulls. No music, no dialogue.",
"width": 1344,
"height": 768,
"duration": 8
}Response
[
{
"taskType": "videoInference",
"taskUUID": "5d1c8e2a-3f47-4b9e-a2c6-7e0f9b3d4a18",
"videoUUID": "b8e4f1a7-2c93-4d05-9e6b-1f7a3c8d2e40",
"videoURL": "https://vm.runware.ai/video/os/a14d18/ws/2/vi/b8e4f1a7-2c93-4d05-9e6b-1f7a3c8d2e40.mp4"
}
]positivePromptis the only required field and takes 1 to 32,000 characters.widthandheightpick one of twelve fixed pairs, six aspect ratios from 21:9 to 9:16 at 480p or 768p, and default to 1344 × 768. Text-to-video does not acceptresolution.durationis a whole number of seconds from 5 to 15, defaulting to 5.settings.promptEnhancementcontrols the rewriting pass the prompt goes through before generation. See Prompt enhancement.
Sound comes from the same prompt as the picture, so every example here ends with an Audio: clause. The dialogue and sound guide covers directing it. Media in inputs switches the model into another mode, covered in the image-to-video, reference and audio reference guides. Everything below carries into all of them.
Writing the full brief
A one-line prompt still returns a usable clip, but every decision in it belongs to the model: the lens, the time of day, the finishes, the pace and the sound. A full brief makes those decisions itself. Below, the same listing kitchen from one sentence and from a brief of about 1,000 characters, both at the 21:9 size a listing banner uses.
A modern kitchen in a new apartment.
Shot on a full-frame cinema camera with a 35mm lens at T4, locked off on a tripod at counter height, framing the full length of the kitchen island. A newly finished apartment kitchen at ten in the morning. A long island in pale honed marble with a waterfall edge, three brass pendant lights hanging above it, flat-front cabinets in warm white oak, a matte black faucet, and wide-plank oak floors. Full-height windows on the left let in low, clear winter sun that falls across the island in long bright rectangles. Nobody is in the room. The only movement is the slow drift of the window light across the marble as a thin cloud passes and the sun returns. Bright, clean real-estate photography with natural color, straight verticals and no lens distortion. No people, no logos, no brand names and no printed words anywhere in frame. Audio: quiet room tone, the faint hum of the refrigerator, and distant street traffic far below the windows. No music, no dialogue.
The one-line take is a plausible kitchen, but its materials and its light are whatever the model reached for. The full brief gets the kitchen the listing actually has: the honed marble with its waterfall edge, the brass pendants, the white oak, and winter sun falling across the island at a specific hour. That difference is what makes a clip usable for a real property rather than a stock placeholder.
Each clause in the brief answers one question a director would otherwise leave open. Hover a segment to see which:
There is no penalty for detail. A clause the model already agrees with costs nothing, and a clause it would have gotten wrong is a fixed take instead of a reroll. Name the action even when the action is almost nothing: "the only movement is the light" is an instruction, and silence on the point leaves the model to invent something to move.
Naming the camera
Mood words give the model nothing it can act on. "Cinematic", "high quality", "beautiful lighting" and "stunning 4K footage" describe an outcome without saying how to reach it.
The type of camera, the lens and the aperture say how, and a sentence describing the frame and the focus makes sure the result matches them.
The three takes below share one prompt about a yoga instructor holding a pose. Only the opening camera clauses change.
Cinematic, high quality, beautiful lighting, stunning 4K footage. A yoga instructor in her thirties holds warrior II pose on a cork-colored mat in a bright studio with pale wood floors and a white-painted brick wall. She wears a fitted sage-green long-sleeve top and black leggings, with her dark hair pulled into a tight low bun. Her front knee is bent directly over her ankle, both arms extend level with the floor, and her gaze rests over her front fingertips. She breathes slowly and steadily and holds the pose without moving her feet. Soft daylight from tall windows on the right. No logos, brand names or printed words anywhere in frame. Audio: her slow, even breathing and a quiet studio room tone. No music, no dialogue.
Shot on a large-format cinema camera with an 85mm lens at T1.8, locked off on a tripod at waist height. A tight frame from her knees up, with the brick wall and the windows behind her falling into a soft, out-of-focus blur. A yoga instructor in her thirties holds warrior II pose on a cork-colored mat in a bright studio with pale wood floors and a white-painted brick wall. She wears a fitted sage-green long-sleeve top and black leggings, with her dark hair pulled into a tight low bun. Her front knee is bent directly over her ankle, both arms extend level with the floor, and her gaze rests over her front fingertips. She breathes slowly and steadily and holds the pose without moving her feet. Soft daylight from tall windows on the right. No logos, brand names or printed words anywhere in frame. Audio: her slow, even breathing and a quiet studio room tone. No music, no dialogue.
Shot on a full-frame mirrorless camera with a 24mm lens at f/8, locked off on a tripod at waist height. A wide frame that shows her full body and most of the studio, with the floor, the brick wall and the windows all in sharp focus. A yoga instructor in her thirties holds warrior II pose on a cork-colored mat in a bright studio with pale wood floors and a white-painted brick wall. She wears a fitted sage-green long-sleeve top and black leggings, with her dark hair pulled into a tight low bun. Her front knee is bent directly over her ankle, both arms extend level with the floor, and her gaze rests over her front fingertips. She breathes slowly and steadily and holds the pose without moving her feet. Soft daylight from tall windows on the right. No logos, brand names or printed words anywhere in frame. Audio: her slow, even breathing and a quiet studio room tone. No music, no dialogue.
The 85mm take pulls the instructor out of the room, framed tight on her body with the studio softened behind her, the look a class promo wants. The 24mm take keeps her full body and the whole studio in focus, the look a studio tour wants. The mood-word take lands wherever the model decides, because nothing in its prompt did.
Name the support as well as the glass. "Locked off on a tripod" keeps the camera still for the whole clip, and a still camera is the composition the model handles most reliably (see Staging contact).
Ruling out the defaults
Some of the model's defaults are strong enough that describing what you want does not fully override them. The fix is to name the default and exclude it in the same prompt.
Skin and grade
Faces drift toward a retouched, beauty-filtered look. Describing pores and lines gets you part of the way there. Excluding the treatment gets you the rest. Both takes below describe the same weathered shopkeeper, and only the second adds the exclusions.
Shot on a large-format cinema camera with a 50mm lens at T2.8, locked off on a tripod at eye level. A man in his late sixties stands behind the wooden counter of a small neighborhood hardware store with both hands resting flat on the counter. He has a short white beard, deep lines across his forehead and around his eyes, sun spots on his cheeks and a slightly crooked nose. He wears a faded navy canvas work apron over a plaid flannel shirt. Behind him, pegboard walls hold rows of hand tools and small drawers of screws. He looks into the camera, nods once and smiles with his eyes. Warm light from overhead fluorescent tubes mixed with daylight from the shop window. Real skin with visible pores, fine lines and uneven tone. No logos, brand names or printed words anywhere in frame. Audio: quiet shop room tone, the faint buzz of the fluorescent tubes, a car passing outside. No music, no dialogue.
Shot on a large-format cinema camera with a 50mm lens at T2.8, locked off on a tripod at eye level. A man in his late sixties stands behind the wooden counter of a small neighborhood hardware store with both hands resting flat on the counter. He has a short white beard, deep lines across his forehead and around his eyes, sun spots on his cheeks and a slightly crooked nose. He wears a faded navy canvas work apron over a plaid flannel shirt. Behind him, pegboard walls hold rows of hand tools and small drawers of screws. He looks into the camera, nods once and smiles with his eyes. Warm light from overhead fluorescent tubes mixed with daylight from the shop window. Real skin with visible pores, fine lines and uneven tone. Muted, natural color. No retouching, no beauty filter, no glamour lighting, no teal-orange grade. No logos, brand names or printed words anywhere in frame. Audio: quiet shop room tone, the faint buzz of the fluorescent tubes, a car passing outside. No music, no dialogue.
The second take keeps the skin the prompt described, sun spots and deep lines included, under the flat mix of fluorescent and window light a real shop has. That is the difference between a testimonial that reads as a real business owner and one that reads as a model in costume. The exclusion clause is reusable as is:
Muted, natural color. No retouching, no beauty filter, no glamour lighting, no teal-orange grade.Color works the same way as skin. Ask for the grade you want, then rule out the default it would slide back to.
Invented branding
Plain clothing and equipment can pick up invented logos: a blank chest or a machine casing is an empty surface the model may fill with a mark of its own. A clip carrying a made-up brand is unusable for commercial work.
Put the no-logos clause in every commercial prompt, and name the surfaces that must stay blank, like the clothing or the product:
No logos, brand names, printed words or badges anywhere in frame.Staging contact
The model is strongest on short, contained shots: one subject, one place, one action, and a camera that stays put or barely moves. Inside that envelope, objects keep their size and shape and stay where they were put.
The weak point is the moment two things touch. Left vague, a hand lands near its target rather than on it, and an object that is set down can drift after it lands. The fix is to name the contact point and say that it holds.
Shot on a large-format cinema camera with a 40mm lens at T2.8, locked off on a tripod at shelf height. A bright, minimal sneaker store with white walls, a polished concrete floor, and wall shelves of plain, unbranded sneakers. A sales associate in her twenties in a plain black t-shirt places a plain white leather sneaker, with no stripes, swooshes or marks of any kind, on a floating glass display shelf, then steps back out of frame. Soft, even overhead light. No logos, brand names, printed words or badges anywhere in frame. Audio: quiet store room tone and her footsteps on the concrete floor. No music, no dialogue.
Shot on a large-format cinema camera with a 40mm lens at T2.8, locked off on a tripod at shelf height. A bright, minimal sneaker store with white walls, a polished concrete floor, and wall shelves of plain, unbranded sneakers. A sales associate in her twenties in a plain black t-shirt holds a plain white leather sneaker, with no stripes, swooshes or marks of any kind, by the heel in her right hand and lowers it onto a floating glass display shelf. The whole sole meets the glass at once and rests flat on it, toe pointing left. Her fingers let go of the heel, and the sneaker stays exactly where she set it, without sliding or tipping, as she steps back out of frame. Soft, even overhead light. No logos, brand names, printed words or badges anywhere in frame. Audio: quiet store room tone and her footsteps on the concrete floor. No music, no dialogue.
The second prompt says where the hand holds the shoe, how the sole meets the glass, and that it stays. Those three facts are the whole action, so writing them down leaves nothing for the model to improvise at the moment of contact.
Prefer stillness to manipulation. A subject wearing or holding a product is the most reliable way to feature it. Close hand work, like assembling a product or operating a device, is where the model is least reliable, so stage the product in the hand rather than in use.
Lettering in the shot
Lettering renders when it is in capitals, spelled out exactly, and named as the only lettering in frame. Short phrases are the dependable case: the longer the copy, the more characters there are to get wrong.
Shot on a large-format cinema camera with a 35mm lens at T4, locked off on a tripod on the opposite sidewalk. A small white delivery van is parked at the curb of a quiet, tree-lined residential street on a bright spring morning, seen side-on. Its side panel is painted with one line of large dark-green block capitals that reads FRESH DAILY. The only lettering anywhere in frame is the words FRESH DAILY, spelled exactly FRESH DAILY. No other logos, brand names, license plates or printed words anywhere. The van's amber hazard lights blink slowly. Clear morning light, muted natural color. Audio: birdsong, a distant car passing, and the soft tick of the hazard lights. No music, no dialogue.
Shot on a large-format cinema camera with a 35mm lens at T4, locked off on a tripod on the opposite sidewalk. A small white delivery van is parked at the curb of a quiet, tree-lined residential street on a bright spring morning, seen side-on. Its side panel is painted with two lines of dark-green capitals that read FRESH PRODUCE FROM LOCAL FARMS, DELIVERED TO YOUR DOOR EVERY MORNING. The only lettering anywhere in frame is the words FRESH PRODUCE FROM LOCAL FARMS, DELIVERED TO YOUR DOOR EVERY MORNING, spelled exactly as written. No other logos, brand names, license plates or printed words anywhere. The van's amber hazard lights blink slowly. Clear morning light, muted natural color. Audio: birdsong, a distant car passing, and the soft tick of the hazard lights. No music, no dialogue.
Both panels come back as written, the two-word one and the twelve-word tagline across its two lines. The same pattern produced both, and it is worth keeping verbatim:
The only lettering anywhere in frame is the words FRESH DAILY, spelled exactly FRESH DAILY.Stating the phrase twice and declaring it the only lettering does two jobs. It pins the spelling, and it stops the model adding signage, plates or labels of its own elsewhere in the frame. For copy that has to be exact in every frame, like a price or a legal line, check the clip frame by frame or add real type in the edit over a text-free shot.
Filling a longer clip
duration runs from 5 to 15 seconds, and a single contained action fills the default 5 comfortably. For a longer clip, lay out a sequence of simple beats in order and anchor the first and last ones in time, with the camera locked for the whole take.
Shot on a full-frame mirrorless camera with a 35mm lens at f/4, locked off on a tripod across a quiet residential street, one continuous take with no cuts. In the first seconds, a delivery rider in a plain teal jacket and helmet pulls up to the curb on an electric scooter and puts the kickstand down. He takes his helmet off, hangs it on the handlebar, lifts a small cardboard parcel out of the rear box, walks up the short path to a white front door and presses the doorbell. In the final seconds, the door opens, a woman in a gray hoodie takes the parcel, smiles and says: "Thank you!" He nods and walks back toward the scooter. Bright, overcast daylight. No logos, brand names or printed words anywhere in frame, including on the parcel, the scooter and the jacket. Audio: the scooter's electric hum, the kickstand clicking down, footsteps on the path, a doorbell chime, the door opening, her short thank-you, and light street ambience. No music.
The prompt reads like a shot list: the rider arrives, parks, carries the parcel to the door, and the door opens "in the final seconds". Each beat is one simple action, so fifteen seconds plays as a short scene rather than one action stretched thin. For anything that needs cuts, generate each shot as its own clip and edit them together.
Prompt enhancement
Before generation, the prompt passes through a rewriting step that expands it. settings.promptEnhancement controls that step:
turbo, the default, applies a fast rewrite.qualityapplies a more thorough rewrite.disabledsends your prompt exactly as written.
import { createClient } from '@runware/sdk'
const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()
const [result] = await client.run({
model: 'heygen:video@1.0',
positivePrompt: 'Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter\'s small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.',
width: 768,
height: 1344,
duration: 5,
settings: {
promptEnhancement: 'disabled'
}
})import asyncio
import os
from runware import Runware
async def main():
async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
results = await client.run({
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.",
"width": 768,
"height": 1344,
"duration": 5,
"settings": {
"promptEnhancement": "disabled"
}
})
asyncio.run(main())curl https://api.runware.ai/v1 \
-H "Authorization: Bearer $RUNWARE_API_KEY" \
-H "Content-Type: application/json" \
-d '[
{
"taskType": "videoInference",
"taskUUID": "e27a9c14-6b3f-48d1-a5e0-93c2f7b61d05",
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.",
"width": 768,
"height": 1344,
"duration": 5,
"settings": {
"promptEnhancement": "disabled"
}
}
]'runware run heygen:video@1.0 \
positivePrompt="Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue." \
width=768 \
height=1344 \
duration=5 \
settings.promptEnhancement=disabled{
"taskType": "videoInference",
"taskUUID": "e27a9c14-6b3f-48d1-a5e0-93c2f7b61d05",
"model": "heygen:video@1.0",
"positivePrompt": "Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.",
"width": 768,
"height": 1344,
"duration": 5,
"settings": {
"promptEnhancement": "disabled"
}
}Response
[
{
"taskType": "videoInference",
"taskUUID": "e27a9c14-6b3f-48d1-a5e0-93c2f7b61d05",
"videoUUID": "3a9f6e21-d84c-4b17-8f25-c0e6b9a14d73",
"videoURL": "https://vm.runware.ai/video/os/a14d18/ws/2/vi/3a9f6e21-d84c-4b17-8f25-c0e6b9a14d73.mp4"
}
]The rewrite helps a short prompt, since it fills in the decisions a single line leaves open. On a brief that already makes every decision, it can only add choices you did not make. The scooter prompt below ran under the default and with enhancement disabled.
Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.
Shot on a large-format cinema camera with a 65mm lens at T2, locked off on a tripod at knee height. An electric kick scooter in a matte sand color, with a black deck and black handlebar grips, stands on its kickstand on a quiet city sidewalk at dawn, seen in three-quarter view from the front. The sidewalk is still damp from the night and reflects the pale sky. Behind the scooter, a row of closed storefronts with rolled-down shutters sits softly out of focus. The scooter's small round headlight switches on, and a moment later the thin light strip on its stem glows white. Nothing else moves. Cool blue dawn light, with the first warm sun touching the tops of the buildings. Muted, natural color, no teal-orange grade. No people, no logos, no brand names and no printed words anywhere in frame, including on the scooter. Audio: an empty-street room tone, a distant bus, and a soft electronic chirp as the headlight switches on. No music, no dialogue.
The default rewrite turned the kick scooter into a seated moped. With enhancement disabled, the clip keeps the kick scooter the brief describes, and its headlight and stem light switch on as written. Turn it off once your prompt is a full brief, and leave it on while you are still working from a sentence or two.
Where it does best
The model plays to its strengths on a single contained shot: one subject, one place, one action, with a locked or slow camera. Inside that envelope objects keep their size and position and behave physically, and the sound matches the space described because it comes from the same prompt.
Two kinds of content are less reliable, so plan around them rather than prompting harder:
- Soft organic motion such as loose hair and paper. Tuck hair under a hat or into a bun, and keep paper and fabric still where you can.
- Close hand work, like assembling or operating equipment. Show the product held or worn instead.
Tips
-
Write the full brief. A few hundred to a few thousand characters beats a one-liner. There is no penalty for detail, only for decisions left open.
-
Open with the camera and the frame. Type, lens, aperture and support, then the framing and focus you expect: "Shot on a large-format cinema camera with a 50mm lens at T2, locked off on a tripod. A waist-up frame with the background soft." Mood words like "cinematic" change almost nothing.
-
Describe the texture, then exclude the treatment. "No retouching, no beauty filter, no glamour lighting, no teal-orange grade" is what keeps a real face real.
-
Ban logos in every commercial prompt. Plain garments and devices pick up invented marks. Name the surfaces that must stay blank.
-
Name contact points and say they hold. Where the hand grips, how the object lands, and that it stays put.
-
Keep lettering short and spelled out. Capitals, the exact words, and declared the only lettering in frame. Check longer copy frame by frame.
-
Say what should not move. "Nothing else moves" or "the only movement is the light" stops the model from inventing motion to fill the clip.
-
Lay out long clips beat by beat. One simple action per beat, in order, with the first and last anchored in time.
-
Disable prompt enhancement for finished briefs.
settings.promptEnhancement: "disabled"sends your words as written. Leave the default on for short prompts.