What is HunyuanVideo 1.5 good at?

One person doing one thing, shot with a readable camera move. In our runs skin and fabric looked natural and faces performed convincingly on image-to-video. Tencent built it as a compact 8.3-billion-parameter model that runs on consumer GPUs, and here it makes 480p or 720p clips of 5 or 8 seconds.

Is HunyuanVideo 1.5 uncensored?

Yes. Nothing filters the prompt, the photo or the result. In our tests mid-level probes rendered as written, and at the strongest level the model itself held back part of the scene. Everyone on screen must be an adult, and the rights and consent for what you generate stay with you.

No. Every clip is a silent 24 fps MP4. If you need a soundtrack generated with the picture, LTX-2 19B returns stereo audio with every clip.

In our runs a 5-second 480p clip took 63 to 150 seconds of model time, most between 70 and 130, an 8-second 480p clip 77 to 151 seconds, and a 5-second 720p clip 142 to 241 seconds.

Which sizes can I get?

Text-to-video gives 832 x 480, 480 x 832, 1280 x 720 or 720 x 1280, chosen by the ratio (16:9 or 9:16) and the resolution. Image-to-video follows your photo: a 16:9 photo came back at 832 x 464 on 480p and 1264 x 720 on 720p.

Do I need a prompt for image-to-video?

No, a call without a prompt still animated the photo in our tests, but a sentence about what happens next gives you control over the motion. Describe the action, not what the picture already shows.

Why do shots with two people look odd?

Close contact between two people is the weak spot of this model: limbs tend to tangle and anatomy slips. Single subjects held up much better in our tests.

Why did my image request fail?

Usually the image link could not be fetched; an unreachable URL answers with a download error and an HTTP status. Use a public, direct link to the image file.

HunyuanVideo 1.5 API

tencent/hunyuan-video-1.5/text-to-video

HunyuanVideo 1.5 text-to-video turns a written prompt into a 5 or 8 second clip at 480p or 720p, landscape or portrait. A negative prompt keeps artefacts out, the output is a silent 24 fps MP4, and there is no content filter.

From $0.024/s

Run model View API

Input

Text to video

Prompt

Negative prompt

Aspect ratio

Resolutionaffects price

Durationaffects price

Seed

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Example result

Extend

Video upscaling

Video editing

{
  "model": "tencent/hunyuan-video-1.5/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration_seconds": 5,
    "seed": 520501,
    "prompt": "Blue hour at a single-storey desert motel, the sky deep cobalt above a black ridge of mountains. An adult Latina woman in her early thirties leans in the open doorway of her room in a fitted ivory satin slip dress, a cropped denim jacket hanging loose from one arm, a steel ice bucket resting on her hip. Warm tungsten light from the room rims her hair and the satin; a moth circles the bare porch bulb. She tilts her head against the door frame and holds a long, smouldering look at the camera, then lets a slow smile break. Low-angle 35mm anamorphic shot, slow push-in, teal-and-amber grade, fine film grain, no signs or lettering anywhere.",
    "negative_prompt": "blurry, distorted hands, extra fingers, text, lettering, watermark, logo"
  }
}

Introduction

HunyuanVideo 1.5 Text to video turns a written prompt into a short clip. HunyuanVideo 1.5 is Tencent's open 8.3-billion-parameter video model; this endpoint returns 5 or 8 seconds at 480p or 720p in landscape or portrait, with a negative prompt for artefacts. The platform adds no content filter to the prompt or the clip.

Choose a mode by starting material

What you have	Mode
a written description	HunyuanVideo 1.5 Text to video
a photo to animate	HunyuanVideo 1.5 Image to video

Key features & highlights

Natural people at a small size

The model has 8.3 billion parameters, yet in our tests faces, skin, hair and fabric looked photographic and moved without melting.

Four frame sizes

aspect_ratio and resolution combine into 832 x 480, 480 x 832, 1280 x 720 or 720 x 1280, so vertical clips are composed for the screen, not cropped.

Negative prompt

List what you never want to see. It is the quickest fix for stray text, warped hands or blur.

Uncensored

No filter on the prompt or the result. In our tests most content probes rendered as written; the strongest held part of the scene back.

Request parameters

Request parameters for HunyuanVideo 1.5 · Text to Video
Parameter	Required	Description
`prompt`	Yes	Describe who is in frame, what they wear and do, and what the camera does. One clear action per clip works best. at least 1 characters
`negative_prompt`	No	Optional list of things to keep out, such as "blurry, distorted hands, extra limbs, text, watermark". It does not change the price.
`aspect_ratio`	No	16:9 for landscape (the default) or 9:16 for portrait. The shape does not change the price. `16:9` · `9:16` · Default: `16:9`
`resolution`	No	480p (832 x 480) or 720p (1280 x 720, the default). 720p costs twice as much per second; draft at 480p. `480p` · `720p` · Default: `720p`
`duration_seconds`	No	5 or 8 seconds, billed per second. `5` · `8` · Default: `5`
`seed`	No	Keeps runs comparable while you change one thing at a time. It does not promise a frame-identical clip. `≥ 0` · `≤ 2147483647`

Pricing

Billed per second of output at the chosen resolution. The aspect ratio does not change the price.

Pricing
Specification	Price
`480p`	$0.024 / second
`720p`	$0.048 / second

Use cases

Vertical social clips

9:16 at 720 x 1280 for feeds and stories.

Prompt exploration

At 480p a five-second draft is cheap, so you can try many prompts before you upscale the idea.

Unfiltered clips for your platform

No content filter on the prompt or the clip, with adults only on screen.

Explore key use-case tutorials

Prompt guide & pro tips

Write the shot

Subject, wardrobe, action, light, camera: in that order, in two to four sentences.

One action per clip

Five seconds is one gesture or one move. Chains of actions blur together.

Keep contact scenes simple

Two people touching tend to tangle limbs; slow movement and a wider shot help.

Draft small

Run the idea at 480p with a fixed seed, then rerun the winner at 720p.

API quick start

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "tencent/hunyuan-video-1.5/text-to-video",
  "input": {
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "resolution": "720p",
    "aspect_ratio": "16:9",
    "duration_seconds": 5
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

Turn a prompt or one photo into a lifelike five-second clip

HunyuanVideo 1.5 is Tencent's open 8.3-billion-parameter video model, and here it turns a text prompt or a single photo into a 5 or 8 second clip at 480p or 720p. In our runs it was strongest on one person doing one thing: natural skin and fabric, a singer closing her eyes mid-song, a man turning to the camera with a slow smile. Clips are 24 frames per second and silent. Nothing filters the prompt, the photo or the result; at the strongest level of our content probes the model held part of the scene back, and that is the model's own restraint, not a platform filter.

Open the playground

Turn a prompt or one photo into a lifelike five-second clip

Model: Tencent HunyuanVideo 1.5, 8.3 billion parameters, released as open weights in November 2025
Clip length: 5 or 8 seconds at 24 fps; 8 seconds is 193 frames, 5 seconds is 121
Resolution: 480p or 720p; text-to-video gives 832 x 480, 480 x 832, 1280 x 720 or 720 x 1280
Speed: 480p, 5 s: usually 70 to 130 seconds of model time; 720p, 5 s: 142 to 241 seconds
Audio: None; every clip is silent
Content filter: None; the strongest probes came back toned down by the model itself

Best at: One person, one action, one camera moveSingle subjects held up well in our tests: skin, fabric and faces stayed natural, and on image-to-video the face performed convincingly. Write one action per clip and say what the camera does.
Photo to clip: Image-to-video keeps your picture's shapeThe clip starts from your photo and keeps its proportions: a 16:9 photo came back at 832 x 464 on 480p and 1264 x 720 on 720p, a 2:3 photo at 512 x 768 on 480p. The prompt is optional; a call without one still animated the frame.
Limits: What it will not doNo audio track, only 5 or 8 seconds, only 480p and 720p, and no aspect-ratio field on image-to-video because the photo decides. Two people in close contact tend to tangle limbs, so keep to single subjects or a loose, wide-framed slow dance.

Working with HunyuanVideo 1.5

Write the shot, not just the subject

A prompt that names who is in frame, what they wear, what they do over the five seconds and what the camera does gave our best clips. The jazz-club dancer in the gallery came from one sentence about the dress, the turn, the amber light and a slow circling camera; the four-frame sheet is one clip.

Start from a photo when the face matters

Image-to-video fixes the person, the wardrobe and the light, so the prompt only has to describe the next moment. Our singer photo became a clip of her leaning toward the microphone and closing her eyes; the left frame is the photo, the right one is a few seconds in.

Draft at 480p, finish at 720p
480p 5-second clips usually came back in 70 to 130 seconds against 142 to 241 for 720p, so test the idea small and rerun the prompt you like at 720p. Same prompt and seed give a similar clip, never an identical one.
Pick the shape on text-to-video
Text-to-video takes 16:9 or 9:16 and builds the frame from the ratio plus the resolution. Vertical 720 x 1280 suits feeds; the rain-soaked rooftop portrait in the gallery is a 9:16 clip. On image-to-video, crop the photo to the shape you want first.
Use the negative prompt for glitches
On text-to-video, a short list such as “blurry, distorted hands, extra limbs, text, watermark” keeps common artefacts out. Keep it to things you never want and describe the look in the main prompt.

Who uses HunyuanVideo 1.5

Social and short-form teams
Vertical clips for feeds
Generate 9:16 at 720 x 1280, keep one seed per series, and compare versions fairly. The rooftop example, a man pushing back wet hair under neon rain, came from a text prompt alone.
Photographers and portrait artists
Bring a still to life
Send a portrait and describe what the person does next. Our man in a white bathrobe at a window turned to the camera with a slow smile while the city lights came on outside; the left frame is the photo, the right one is the clip.
Product and travel creators
Motion for a scene you already have
A generated photo of a couple on a yellow scooter on a coastal road became five seconds of riding, hair and dress moving in the wind. Cheap 480p drafts show whether the motion works before you commit to 720p.
Uncensored platform builders
A model that does not push back
Nothing filters the prompt, the photo or the result, so an unfiltered platform does not need a workaround. Everyone on screen must be an adult, and in our tests the strongest probes still came back toned down.

From first clip to finished batch with HunyuanVideo 1.5

Choose the entry point
Text-to-video when you only have an idea, image-to-video when a face, outfit or location must stay fixed.
Draft at 480p for 5 seconds
Write one action and one camera move, add a negative prompt on text-to-video, and run a few seeds.
Read the clip, not the thumbnail
Check hands, faces and any second person. If limbs tangle, simplify the scene to one subject.
Rerun the winner at 720p
Keep the prompt and seed, switch the resolution, and use 8 seconds only if the action needs the time.
Batch the series
Fix the style sentence, change only the subject line, and send the set. Add sound afterwards, since the clips are silent.

HunyuanVideo 1.5 prompts worth copying

A slow push-in portrait

For text-to-video at 9:16. One person, one small action and one camera move.

An adult woman in her thirties in a camel wool coat stands at a rain-streaked bus stop at dusk, tucks a strand of hair behind her ear and looks up toward the camera with a faint smile. Warm shop lights blur behind her, slow push-in, shallow depth of field, cinematic, photorealistic.

Bring a portrait to life

For image-to-video from a portrait. Describe the next moment, not the photo.

The adult man in the denim jacket lifts his coffee cup, glances out of the café window, then turns back to the camera and laughs quietly. Soft afternoon light, the camera holds steady.

A slow dance, kept simple

For text-to-video at 16:9. Keep the pose loose and the framing wide so limbs stay readable.

Two adults in their thirties slow dance on a small rooftop terrace at golden hour, she in a flowing green dress, he in a white shirt with rolled sleeves, a string of lights swaying overhead. Handheld camera drifting around them, warm and intimate, photorealistic.

More HunyuanVideo 1.5 questions

Use it as your lifelike-people workhorse

Pick text-to-video for ideas and image-to-video for faces, write one action and one camera move, draft at 480p, and rerun the winner at 720p.

Keep scenes to one subject, add a negative prompt for glitches, and add sound afterwards because the clips are silent.

Workbench

Input

Output

Introduction

Choose a mode by starting material

Key features & highlights

Natural people at a small size

Four frame sizes

Negative prompt

Uncensored

Request parameters

Pricing

Use cases

Vertical social clips

Prompt exploration

Unfiltered clips for your platform

Prompt guide & pro tips

Write the shot

One action per clip

Keep contact scenes simple

Draft small

API quick start

Related models and next tasks

Turn a prompt or one photo into a lifelike five-second clip

Working with HunyuanVideo 1.5

Write the shot, not just the subject

Start from a photo when the face matters

Draft at 480p, finish at 720p

Pick the shape on text-to-video

Use the negative prompt for glitches

Who uses HunyuanVideo 1.5

Vertical clips for feeds

Bring a still to life

Motion for a scene you already have

A model that does not push back

From first clip to finished batch with HunyuanVideo 1.5

Choose the entry point

Draft at 480p for 5 seconds

Read the clip, not the thumbnail

Rerun the winner at 720p

Batch the series

HunyuanVideo 1.5 prompts worth copying

A slow push-in portrait

Bring a portrait to life

A slow dance, kept simple

More HunyuanVideo 1.5 questions

What is HunyuanVideo 1.5 good at?

Is HunyuanVideo 1.5 uncensored?

Does it make sound?

How fast is it?

Which sizes can I get?

Do I need a prompt for image-to-video?

Why do shots with two people look odd?

Why did my image request fail?