Real outputs from the public examples of this model on Replicate, with the prompt and settings that produced them. They were not generated live on this page.
Prompt
A woman with long brown hair and light skin smiles at another woman with long blonde hair. The woman with brown hair wears a black jacket and has a small, barely noticeable mole on her right cheek. The camera angle is a close-up, focused on the woman with brown hair's face. The lighting is warm and natural, likely from the setting sun, casting a soft glow on the scene. The scene appears to be real-life footage.
A man in a blue plaid shirt turns his head to look at a woman walking past in a red top. His expression is curious as he glances away from his companion in a light blue sleeveless top. The scene takes place on a busy city street with buildings and pedestrians visible in the background. The natural lighting suggests it's daytime. The camera captures the scene from street level, with the woman in red slightly out of focus in the foreground.
A chef prepares food in a professional kitchen. He has olive skin and dark, close-cropped hair, wearing traditional white chef's attire. Steam rises from the pan he's working with. The kitchen's stainless steel surfaces reflect the bright overhead lighting. The camera follows his hands as he cooks. The scene appears to be real-life footage.
Lightricks was founded in 2013 by Zeev Farbman (CEO), Yaron Inger, Itai Tsiddon, Nir Pochter and Amit Goldstein, spinning out of Hebrew University of Jerusalem's computer-graphics research group. The company is best known for the consumer photo and video apps Facetune, Videoleap, Photoleap and Filtertune, which together have hundreds of millions of installs. Lightricks raised more than $300M from Insight Partners, Goldman Sachs and others, at a $1.8B valuation. In November 2024 Lightricks open-sourced LTX-Video, a 2B-parameter real-time video diffusion model optimized for speed on consumer hardware. LTX-Video positioned the company as one of the few Western players shipping open-weight video AI, and is now embedded in the Videoleap and LTX Studio products.
Diffusion Transformer (DiT) at 2B parameters with heavy spatial-temporal compression
LTX-Video is a 2B-parameter Diffusion Transformer designed for real-time-ish inference: at launch it generated 5-second 768x512 videos in under 4 seconds on a single H100, and 30-frame clips faster than real time. The architecture compresses video aggressively into a 1:192 spatio-temporal latent (vs the 1:48 typical of HunyuanVideo) using a heavy 3D causal VAE; the DiT denoiser then operates on this very compact latent grid. Position information uses 3D RoPE, and the model is trained with Flow Matching using v-prediction. Text conditioning combines a CLIP-style encoder and a Llama-class LLM encoder. LTX-Video supports text-to-video and image-to-video, with a 'mid-conditioning' mode that allows inserting an intermediate frame for guided motion. The model is published under an open licence designed to enable both research and commercial use.
curl https://railwail.com/api/v1/videos/generations \
-H "Authorization: Bearer $RAILWAIL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "ltx-video-lightricks",
"prompt": "A slow drone shot over a misty pine forest at sunrise"
}'
# Response: {"job_id": "...", "status": "queued", ...}
# Poll until status is completed, failed or cancelled:
curl https://railwail.com/api/v1/jobs/JOB_ID \
-H "Authorization: Bearer $RAILWAIL_API_KEY"
import os
import time
import requests
API = "https://railwail.com/api/v1"
headers = {"Authorization": f"Bearer {os.environ['RAILWAIL_API_KEY']}"}
job = requests.post(
f"{API}/videos/generations",
headers=headers,
json={
"model": "ltx-video-lightricks",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
},
).json()
while True:
status = requests.get(f"{API}/jobs/{job['job_id']}", headers=headers).json()
if status["status"] in ("completed", "failed", "cancelled"):
break
time.sleep(5)
print(status["status"], status.get("output_url"))
const API = "https://railwail.com/api/v1";
const headers = {
Authorization: `Bearer ${process.env.RAILWAIL_API_KEY}`,
"Content-Type": "application/json",
};
const job = await fetch(`${API}/videos/generations`, {
method: "POST",
headers,
body: JSON.stringify({
model: "ltx-video-lightricks",
prompt: "A slow drone shot over a misty pine forest at sunrise"
}),
}).then((r) => r.json());
let status;
do {
await new Promise((r) => setTimeout(r, 5000));
status = await fetch(`${API}/jobs/${job.job_id}`, { headers }).then((r) => r.json());
} while (!["completed", "failed", "cancelled"].includes(status.status));
console.log(status.status, status.output_url);
Tencent's HunyuanVideo, a 13B open-weights text-to-video diffusion transformer. Produces high-motion, photorealistic clips with smooth temporal consistency and was one of the first open models to rival closed systems on motion quality.
