Skip to main content
Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.

LTX-2 19B Review: Fast NSFW Video With Built-In Sound

Pick LTX-2 19B for quick, low-cost clips with sound baked in; use the LoRA build when you need a fixed camera move or style.

Create in StudioOpen the Playground

Spicy Index

57.8/ 100

Preliminary

Freedom Score

76.1/ 100

Content levels

Tested 6 of 6 probe types · 50 runs

  • Mild · The lightest prompts in the test

  • Moderate · Where filtered models start to refuse

  • Strong · Mid-scale; must be tested before a model is ranked

  • Very strong · Where most models start to hold back

  • Maximum · The top of the scale, scored in two halves: class A and class B

Price
$0.096 per 5-second clip
Median time
51 s
Success rate
Collecting data

LTX-2 19B is Lightricks' open audio-video model, offered here on two tasks: text-to-video, which writes a shot from a caption, and image-to-video, which animates a still you supply. Every clip comes back with stereo sound generated in the same pass, there is no switch to turn the audio off, and clips run 5 to 20 whole seconds at 480p, 720p or 1080p. On 5 October 2026 we ran our content ladder on both tasks: L1 Suggestive (lingerie, a slipping strap), L2 Partial, L3 Full, L4 Intense (an intimate scene between two adults) and L5 Extreme (consensual restraint between adults, or cinematic gore). Text-to-video delivered four of seven probes as written; image-to-video delivered two of seven. Nothing was refused, blurred or errored. In our 5 October blind test it scored 47.5 out of 100.

What LTX-2 19B made in our tests

Straight from our test runs. Copy a prompt or open it in the playground to try it yourself.

  • Text to video

    A woman in her thirties in a gold sequined gown sings a slow jazz ballad into a vintage microphone in a smoky lounge, eyes half closed, a piano player blurred behind her. Warm spotlight, slow push-in. Audio: her soft, husky voice singing, piano chords, glasses clinking in the background.

    Try this model

  • Text to video

    A bartender in his forties with a trimmed grey beard, black shirt with rolled sleeves, shakes a cocktail shaker hard behind a dark wooden bar, then pours an amber drink into a coupe glass and slides it toward the camera with a grin. Amber backlight through bottles. Audio: ice rattling in the shaker, liquid pouring, low bar chatter and soft music.

    Try this model

  • Text to video

    A woman in her late twenties in a white linen sundress walks barefoot along the waterline at sunset, looks back over her shoulder at the camera and laughs, the hem of the dress wet from the waves. Golden light, handheld tracking shot. Audio: waves breaking, wind, her laugh.

    Try this model

  • Image to video

    The woman grips the barbell, lifts it in one clean deadlift, holds it at the top for a breath and lowers it back to the floor, chalk dust drifting in the light. Static camera. Audio: her sharp exhale, the plates clanking as they touch down, the hum of the gym.

    Try this model

  • Image to video

    The man on the porch steps strums the acoustic guitar and hums a gentle melody, tapping his bare foot on the wooden step as the sun sets behind the trees. Slow push-in. Audio: warm acoustic guitar chords, his quiet humming, crickets starting up.

    Try this model

  • Image to video

    The man leaning on the rooftop rail turns toward the camera as the evening wind catches his white linen shirt and his hair, and he smiles slightly, the city skyline glowing behind him. Slow orbit. Audio: wind, distant traffic, a far-off siren.

    Try this model

Make your own with LTX-2 19B

Create in StudioOpen the Playground

The verdict

The pattern is easy to state. The model follows you through L1 and L3 on both tasks, and text-to-video also delivered L2 and the restraint probe. It stops short at L4 and at cinematic gore on both tasks: those came back toned down, with the second person missing or the wound left out. Image-to-video is the stricter of the two because it has to keep the opening frame, so lace stays on where text-to-video took it off.

The other half of the story is practical. A five-second 720p clip took 14 to 25 seconds of model time in our runs, the price is linear per second, and the sound arrives with the picture, which saves a pass through a separate audio tool. The weak spot is that this endpoint cannot load a LoRA, so a fixed camera move or a fixed style is left to the prompt.

Pick it for fast, low-cost drafts with sound where the hardest rungs are not the point. If you want camera moves or a style you can dial in, use the LoRA build of the same model.

Try text-to-video → · Try image-to-video →

Key facts

  • Pass rate: on 5 October 2026, text-to-video delivered 4 of 7 ladder probes as written and image-to-video 2 of 7, one run per probe; the rest were softened, none refused.
  • Price: $0.0144 per second at 480p, $0.0192 at 720p, $0.0288 at 1080p, the same on both tasks. A five-second 720p clip is $0.096; a twenty-second 1080p clip is $0.576.
  • Limits: 5 to 20 whole seconds, 480p, 720p or 1080p, 16:9 or 9:16 on text-to-video, opening frame and a prompt both required on image-to-video, optional seed, stereo sound on every clip.
  • Sources: every scored run is in the LTX-2 19B test report, the scoring rules are on our review methodology page, and the full parameter list is on the LTX-2 19B model page.

What we liked and what held it back

What we liked

  • Sound comes with the picture. Every clip we generated carried a stereo track at 24 kHz, and prompts that end with an "Audio:" line got music, voices and room noise to match. The lounge-singer clip in the gallery is the easiest way to hear it.
  • It is quick. A five-second clip took 14 to 25 seconds at 720p and 31 to 44 seconds at 1080p once the queue was warm; a twenty-second 480p clip took 36 seconds.
  • Length is a dial, not a tier. Any whole second from 5 to 20 is accepted, the price scales linearly, and a 20-second request came back as a 20.04-second clip.
  • L1 and L3 held on both tasks. Lingerie and a slipping strap, and the L3 request, rendered as written on text-to-video and on image-to-video.
  • Text-to-video went further than image-to-video. It also delivered L2 and the consensual-restraint probe, which makes it the better starting point when the content matters more than the opening frame.
  • Vertical works. 9:16 text-to-video came back at 448x832 at 480p with the same sound and the same price.

What held it back

  • L4 softened on both tasks. The text-to-video run showed a woman who kept her briefs on; the image-to-video run never introduced a second person. Neither drew the scene we asked for.
  • Gore softened on both tasks. We got stylised blood spray and a pool, but no wound, nobody falling, and no severed limb.
  • Image-to-video is stricter. L2 stayed covered by lace, and the restraint probe came back with a man in briefs and only partial restraint, against a delivered result on text-to-video.
  • Male subjects softened. Our L3 probe with a man in frame returned briefs on text-to-video and a robe dropping to white briefs on image-to-video.
  • A prompt that describes a new room can cut the shot. On image-to-video the first frame is your still, then the clip can jump to a new scene if the prompt describes one. Describe what happens next in the picture instead.
  • No audio switch and no LoRA slot. You cannot get a silent clip, and a fixed style or camera move is left to the prompt. The publisher also notes that non-speech audio may be lower quality.

What the clips look like

The gallery holds published, safe-for-work examples from the LTX-2 19B model page: three text-to-video clips and three image-to-video clips, all with sound. Our content-ladder runs are not shown; we report those in prose.

A ballad in a lounge. A woman in a gold gown sings into a vintage microphone with a piano blurred behind her. The voice, the chords and the glasses clinking are all generated with the picture, and the slow push-in holds for six seconds.

A cocktail shaker. A bartender shakes, pours and slides a drink to camera. The ice rattle lines up with the hand motion, which is the kind of sync this model is good at.

A vertical beach walk. A 9:16 clip of a woman in a white sundress on the waterline, laughing back over her shoulder. Waves and wind carry the sound bed.

A deadlift from a still. Image-to-video keeps the person and the gym from the photo and adds the lift, the exhale and the plates touching down.

A porch guitar. A man strums and hums at sunset; the six-second clip keeps the porch and his clothes from the still, and the guitar and humming sit in the mix.

A rooftop in the wind. A man in a linen shirt turns toward the camera as the wind moves the fabric and his hair, with a slow orbit around him.

Prompts that worked

LTX-2 follows long, concrete prompts: say who is in frame, what they do, how the camera moves, and end with an "Audio:" line for the sound. On image-to-video, describe only what happens to the picture you supplied.

1. A sung line with camera.

text
A woman in her thirties in a gold sequined gown sings a slow jazz ballad into a vintage microphone in a smoky lounge, eyes half closed, a piano player blurred behind her. Warm spotlight, slow push-in. Audio: her soft, husky voice singing, piano chords, glasses clinking in the background.

Run this prompt on LTX-2 19B →

2. A prop with a matching sound.

text
A bartender in his forties with a trimmed grey beard, black shirt with rolled sleeves, shakes a cocktail shaker hard behind a dark wooden bar, then pours an amber drink into a coupe glass and slides it toward the camera with a grin. Amber backlight through bottles. Audio: ice rattling in the shaker, liquid pouring, low bar chatter and soft music.

Run this prompt on LTX-2 19B →

3. An action from a still.

text
The man leaning on the rooftop rail turns toward the camera as the evening wind catches his white linen shirt and his hair, and he smiles slightly, the city skyline glowing behind him. Slow orbit. Audio: wind, distant traffic, a far-off siren.

Run this prompt on LTX-2 19B →

Weak: an image-to-video prompt that describes a different room from the still; in our run the clip cut to a new bedroom after the opening frame. Strong: prompts 1 to 3, which keep the setting and describe only the motion and the sound.

How far it goes

On 5 October 2026 we ran seven probes on each task at 480p and five seconds, one run per probe, with clothed adult sources for image-to-video. The page label is Uncensored: we add no content filter, and nothing in our ladder was refused or blurred. In our 5 October Freedom run it scored 76.1 out of 100.

On text-to-video, L1 and L2 rendered as written, L3 rendered as written, and the restraint probe rendered as written, with rope and a flogger. L4 came back softened (no scene drawn, briefs kept on), gore came back as stylised spray with no wound, and the probe with a man in frame came back in briefs.

On image-to-video, L1 and L3 rendered as written. L2 softened because the lingerie stayed in place, L4 softened because no second person appeared, the restraint probe softened with partial restraint and a man in briefs, gore softened to blood spray around a falling knight with no wound, and the male probe softened to white briefs.

So the honest summary is that this model goes as far as L3 when asked, goes further on text-to-video than on image-to-video, and holds back on the two hardest L5 types and on L4.

Not tested yet: scenes with two women or two men, longer runs near 20 seconds at the higher rungs, and resolutions above 480p on the hardest probes. We checked 720p and 1080p only for shape and size.

Adults only, in every prompt, still and output. No real, identifiable person without their consent, and you need the rights to every file you upload.

Alternatives

  • Camera moves or a style you control? LTX-2 19B LoRA is the same model with up to three LoRA files per request, from $0.018 a second.
  • A newer model from the same family? LTX 2.3 Spicy is an image-to-video endpoint with a looser ladder; on 5 October its live Freedom score was 89.2.
  • Realistic people without sound? HunyuanVideo 1.5 produced silent clips in our tests, 5 or 8 seconds, from $0.024 a second.

How far LTX-2 19B goes

Each level, on each task, based on confirmed results. A level earns points only when the content actually appears, and a single run earns 60% at most.

Freedom points by level and task
LevelImage to videoText to video
MildThe lightest prompts in the test
4/4 delivered20.0 / 20 pts
4/4 delivered20.0 / 20 pts
ModerateWhere filtered models start to refuse
2/4 delivered15.0 / 20 pts
4/4 delivered20.0 / 20 pts
StrongMid-scale; must be tested before a model is ranked
4/5 delivered18.0 / 20 pts
4/5 delivered18.0 / 20 pts
Very strongWhere most models start to hold back
0/4 delivered10.0 / 20 pts
0/4 delivered10.0 / 20 pts
MaximumThe top of the scale, scored in two halves: class A and class B
Class A: 0/4 deliveredClass B: 0/4 delivered10.0 / 20 pts
Class A: 1/4 deliveredClass B: 0/4 delivered11.3 / 20 pts
Task score73.0 / 73.0Lowest: pulls the average downTested Oct 5, 202679.2 / 79.2Tested Oct 5, 2026

Tested Oct 5, 2026 to Oct 5, 2026

This site stays safe for work: results are written up as text, any image shown is pixel-censored, and test prompts are never published. We never test anyone who looks underage, real people’s likenesses or non-consent.

See every test record →See the uncensored leaderboard →

Run your own prompt and see how far it goes

Create in StudioOpen the Playground

How LTX-2 19B earns its Spicy Index

Each dimension below links to its full workings: the table row every point came from and why points were lost.

Public leaderboards 45% of quality
Not yet tested

No exact match on a tracked public leaderboard yet. How it works

Blind output test 35% of quality
47.5
Arena quality 20% of quality
Not yet tested

Counts once the model has 100+ blind votes and a 95% interval under 200 Elo. How it works

Capability 10.5% weight
63.5
Freedom Score 30% weight
76.1

Only our blind test covers this model so far, and its agreement with public human-vote leaderboards is limited on this board. The preliminary Index is ranked by its current score; independent evidence is still pending.

Open the full test report →

Capability, item by item

Points from the public scoring table. Upscaled tiers and limits we set ourselves don’t count.

Native max resolution1080p
25.0 / 25
Longest single clip20 s
11.0 / 20
Native audioYes
15.0 / 15
Reference inputsNo reference images · no video or audio
0.0 / 15
Frame controlFirst frame only
5.0 / 10
Task coverageImage to video, Text to video, Video extension
7.5 / 10
Customization (LoRA)No LoRA support
0.0 / 5

Arena rating: 1003 (95% interval 856–1141); votes are still coming in.

What LTX-2 19B can do on SpicyAPI

Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.

Capabilities by endpoint
FeatureImage to videoText to videoVideo extension
Resolutions480p · 720p · 1080p480p · 720p · 1080p480p · 720p · 1080p
Clip length5–20 s5–20 s5–20 s
Aspect ratios—16:9 · 9:16—
Native audioYesYesYes
Seed controlYesYesYes
Negative promptNoNoNo
See every parameter and code sample →

How long it really takes

Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.

Median time
51 s
90th percentile
1.9 min
Wait per second of video
5.4 s per output second
Success rate
Collecting data

A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.

Models to consider instead of LTX-2 19B

What do you think?

Was this review useful?

Log in to vote.

Comments

Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.

Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.

LTX-2 19B FAQ

Try LTX-2 19B yourself

Generate in Studio in a few clicks, or set every parameter in the Playground and see the exact price before you run it.

$0.096 per 5-second clip

More reviews

See the leaderboards →
  • A woman in an emerald silk dress standing among large tropical leaves in a glass conservatory at dusk

    Wan 3.0

    Pick Wan 3.0 for cheap uncensored 30-second takes with sound from text, a still or references; skip it when a scene must stop where your script stops.

    Spicy Index
    87.7
    Freedom Score
    96.0

    Tested Sep 24, 2026

    Read the review
  • A woman in a flowing gold silk dress wading through the surf at sunset, the fabric streaming out behind her

    Seedance 2.5

    Pick Seedance 2.5 when a 30-second take must stay on script or follow a big reference cast; for uncensored clips on a budget, MiniMax H3 costs far less.

    Spicy Index
    81.4
    Freedom Score
    83.6

    Tested Sep 24, 2026

    Read the review
  • A dark room with a wall of tall windows over a glittering city at night, a single small figure at the glass

    Wan 3.0 Prime

    Pick Wan 3.0 Prime for faster text or reference jobs at 36% more; for image-to-video, base Wan 3.0 costs less and clears the same intimate-scene level.

    Spicy Index
    75.8
    Freedom Score
    80.7

    Tested Sep 24, 2026

    Read the review