Does LTX-2 19B generate audio?

Yes, on every clip. Lightricks built it as one model for picture and sound, and each MP4 carries stereo AAC audio at 24 kHz that follows the prompt. There is no switch to turn the sound off, and the publisher notes that sound without speech can be of lower quality.

How long can a clip be?

Any whole second from 5 to 20. A request for 4 or 21 seconds is rejected at submit. A 5-second request returns a 5.04-second file because frames are seconds times 24 plus one.

Which resolutions and shapes do I get?

480p, 720p and 1080p. On 16:9 text-to-video that is 832 x 448, 1280 x 704 and 1920 x 1024, and 9:16 swaps the sides. Image-to-video follows your photo: a 16:9 photo came back at 832 x 448, 1216 x 704 and 1856 x 1088.

Yes. Nothing filters the prompt, the image or the result. Most of our content probes rendered as written, and the strongest came back toned down by the model. Everyone on screen must be an adult, and rights and consent for what you generate stay with you.

A 5-second 720p clip took 14 to 25 seconds of model time in our runs, 1080p took 31 to 44, and a 20-second 480p clip 36. A 480p 5-second clip ran 7 to 9 seconds warm, and 45 to 118 when cold or queued.

Why does my image-to-video clip cut to a different scene?

The first frame is your photo, but a prompt that describes another place can make the clip cut away to a new shot. Describe what happens next inside the picture.

Is a prompt required on image-to-video?

Yes, both the image and the prompt are required on this model, unlike HunyuanVideo 1.5 where the prompt is optional.

Do I need LoRAs for camera moves or styles?

Only if you want that control. LTX-2 19B LoRA is the same model with up to three LoRA files per request, such as official camera-control LoRAs; this page's endpoints take no LoRA field.

LTX-2 19B API: Uncensored NSFW Video With Sound, 5–20 s

lightricks/ltx-2-19b/text-to-video

LTX-2 19B text-to-video generates a 5 to 20 second clip and its stereo soundtrack from one prompt, at 480p, 720p or 1080p in 16:9 or 9:16. Voices, music and effects follow the prompt, billing is per second, and there is no content filter.

From $0.014/s

Run model View API

Input

Text to video

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": "lightricks/ltx-2-19b/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration_seconds": 6,
    "seed": 530511,
    "prompt": "Night on the open-air rear balcony of the last carriage of a vintage passenger train racing across a dark plain, the steel tracks streaming away behind it. An adult Japanese man in his mid-thirties in a long wool overcoat worn open over a white shirt with the collar undone grips the brass rail as the wind tears at his coat and hair; the rails run away behind him into the dark, lit red by the tail lamp. A level crossing flashes past, its lights strobing across his face, and he turns to the camera with a slow, dangerous smile. Handheld 35mm, slight push-in, cold blue night with red tail-lamp glow, film grain. Audio: the fast clatter of wheels over rail joints, roaring wind, a crossing bell rising and fading as it passes, then one long horn from the locomotive far ahead."
  }
}

Introduction

LTX-2 19B Text to video turns one prompt into a clip with its own soundtrack. LTX-2 is Lightricks' open audio-video foundation model with 19 billion parameters; this endpoint returns 5 to 20 seconds at 480p, 720p or 1080p in 16:9 or 9:16, with stereo sound in the same MP4. The platform adds no content filter.

Choose a mode by starting material

What you have	Mode
a written description	LTX-2 19B Text to video
a photo to animate	LTX-2 19B Image to video
your own LoRA files	LTX-2 19B LoRA Text to video

Key features & highlights

Picture and sound together

Voices, singing, instruments, effects and room tone are generated with the frames, so they line up with the action.

5 to 20 seconds

Any whole second in that range, billed per second.

Fast at 720p

In our tests a 5-second 720p clip took 14 to 25 seconds of model time.

Uncensored

No filter on the prompt or the clip. Our mid-level content probes rendered as written; the strongest came back toned down.

Request parameters

Request parameters for LTX-2 19B · Text to Video
Parameter	Required	Description
`prompt`	Yes	Describe the scene, the action and the camera, then the sound on its own "Audio:" line: voices, music, effects, ambience. at least 1 characters
`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 448), 720p (1280 x 704, the default) or 1080p (1920 x 1024). Each step costs more per second. `480p` · `720p` · `1080p` · Default: `720p`
`duration_seconds`	No	Any whole second from 5 to 20, billed per second. `≥ 5` · `≤ 20` · 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, and the soundtrack is included.

Pricing
Specification	Price
`480p`	$0.0144 / second
`720p`	$0.0192 / second
`1080p`	$0.0288 / second

Use cases

Performance clips

Singers, musicians and dancers with their sound in one file.

Longer takes

Twenty seconds is enough for a scene with a beginning and an end.

Clips for unfiltered platforms

No content filter, sound included, adults only on screen.

Explore key use-case tutorials

Prompt guide & pro tips

Put sound on its own line

End with "Audio:" and list what should be heard.

Speech is stronger than ambience

The publisher notes that sound without speech can be of lower quality; a voice or a song carries better.

Draft at 5 seconds

Find the shot at 5 seconds and 480p, then extend the length or the resolution.

Expect slightly short frames

720p is 1280 x 704 and 1080p is 1920 x 1024; plan crops accordingly.

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": "lightricks/ltx-2-19b/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.

One request, one finished clip with picture and sound

LTX-2 19B is Lightricks' open audio-video model, and it returns every clip as a finished MP4 with a stereo soundtrack generated in the same pass. Text-to-video and image-to-video run 5 to 20 seconds at 480p, 720p or 1080p, and a 5-second 720p clip took 14 to 25 seconds of model time in our runs. Voices, singing, effects and ambience follow your prompt, and there is no switch to turn the sound off. Nothing filters the prompt, the image or the result.

Open the playground

One request, one finished clip with picture and sound

Model: Lightricks LTX-2 19B, an open audio-video model that makes picture and sound together
Clip length: Any whole second from 5 to 20; 4 and 21 are rejected at submit
Resolution: 480p 832 x 448, 720p 1280 x 704, 1080p 1920 x 1024 on 16:9; 9:16 swaps the sides
Audio: Stereo AAC at 24 kHz on every clip, with no way to switch it off
Speed: 5 s at 720p: 14 to 25 seconds; at 1080p: 31 to 44; 20 s at 480p: 36 seconds
Content filter: None; most of our probes rendered as written and the strongest came back toned down

Sound: Sound is part of the clip, not an extra stepEvery clip came back with AAC stereo audio: a lounge singer's ballad, a bartender's shaker, a guitar on porch steps. Add a line starting with “Audio:” and list what should be heard to steer it; without it the model still adds sound, but you choose less of it.
Length: Up to 20 seconds in one generationPick any whole second from 5 to 20. A 20-second 480p clip finished in 36 seconds of model time in our run, long enough for a scene rather than a fragment. Frame counts are seconds times 24 plus one, so a 5-second request is a 5.04-second file.
Limits: What it will not doNo audio switch and no frame-exact sizes: 1080p is 1920 x 1024 and 720p is 1280 x 704. Text-to-video offers only 16:9 and 9:16, and image-to-video needs both a photo and a prompt. The publisher notes that sound without speech can be of lower quality.

Working with LTX-2 19B

Give the sound its own line

End the prompt with “Audio:” and name what should be heard, for example “ice rattling in the shaker, low bar chatter, soft music”. The lounge singer and the bartender in the gallery both prompts carried an Audio line, so describe the action first and the audio second.

Choose length and shape per request

Text-to-video takes 16:9 or 9:16, and the sheet shows three moments of one vertical clip for feeds. Duration is any whole second from 5 to 20, so draft at 5 and extend the idea only once the shot works.

Draft at 480p, deliver at 720p or 1080p
A warm 480p 5-second run finished in 7 to 9 seconds, though cold or queued runs took 45 to 118. 720p is the sweet spot in our tests at 14 to 25 seconds, and 1080p took 31 to 44.
On image-to-video, stay inside the picture
The first frame is your photo and the clip follows its shape. If the prompt describes a different place, the clip can cut away to a new shot: in our test a stage-singer photo with a prompt set in another room cut to that room. Describe what happens next in the picture you sent.
Keep one scene per clip
One subject, one action and one setting, then the sound line. Long clips let a scene breathe, but stacking several events in a single prompt makes the model pick for you.

Who uses LTX-2 19B

Fitness and lifestyle creators
Clips that already sound right
A woman at a barbell in a dim gym: image-to-video starts from the photo, she grips, lifts and holds, and the Audio line of the prompt asks for the sounds of the room. No separate sound pass to line up.
Musicians and music-video makers
Performances with real audio
A man strumming a guitar on porch steps and humming at sunset came back with the picture and the sound together. Describe the song's mood in the Audio line, not a melody.
Fashion and portrait teams
A still that moves in the wind
A man leaning on a rooftop rail turned to the camera as the evening wind caught his white linen shirt and hair. The photo fixes the face and the wardrobe, the prompt only adds motion.
Uncensored platform builders
Uncensored, with the sound included
No filter on the prompt, the image or the result, and the audio comes with it. 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 LTX-2 19B

Choose text or photo
Text-to-video for an idea in 16:9 or 9:16, image-to-video when a face or place must stay fixed.
Write the picture, then the sound
One scene, one action, then a line beginning “Audio:” with what should be heard.
Draft at 480p for 5 seconds
Check the sound and the opening frame before you spend more on length or resolution.
Raise resolution and length
Rerun the winner at 720p or 1080p and stretch the duration up to 20 seconds when the action needs it.
Watch for cut-aways
On image-to-video, keep the prompt inside the photo's scene; a different setting can trigger a new shot.

LTX-2 19B prompts worth copying

A performance with its own audio

For text-to-video at 16:9, 720p, 6 seconds.

An adult woman in her thirties in a long burgundy dress sings softly into a vintage microphone on a small club stage, eyes half closed, a pianist behind her in warm amber light, slow push-in. Audio: her voice singing a slow ballad, soft piano, quiet applause at the end.

A street scene in vertical

For text-to-video at 9:16.

An adult man in a dark rain jacket walks through a night market under paper lanterns, steam rising from food stalls, he stops, glances back at the camera and smiles. Handheld camera, shallow depth of field. Audio: crowd chatter, sizzling pans, light rain on canvas.

Motion from a window portrait

For image-to-video from a portrait of an adult at a window. Stay inside the photo's setting.

The adult woman in the cream sweater lifts her mug, looks out of the window at the rain, then turns to the camera and smiles. The camera holds steady. Audio: rain on the glass, a soft kettle click, her quiet laugh.

All LTX models6

LTX 2.5Text to video · Image to videoFrom $0.09per output second
LTX 2.3 SpicyImage to video · Video extensionSpicy list price: $0.02From $0.019per output second
LTX 2.3 Spicy LoRAImage to video · Video extensionSpicy list price: $0.03From $0.029per output second
LTX-2.3 LoRAText to video · Image to videoFrom $0.024per output second
LTX-2 19BCurrentText to video · Image to video · +1From $0.014per output second
LTX-2 19B LoRAText to video · Image to videoFrom $0.018per output second

Workbench