LTX Video 2.0 Pro transforms static images into 4K video with synchronized audio through fal's API, generating production-ready content in seconds instead of hours.
Generating Video from Static Images
LTX Video 2.0 Pro generates 4K video with synchronized audio from a single input image at $0.06 to $0.24 per second depending on resolution. The model's transformer-based architecture achieves a 1:192 compression ratio through spatiotemporal downscaling, enabling full self-attention across frames while maintaining temporal coherence.1
This guide covers fal API integration from authentication through production deployment, including pricing, response handling, and optimization strategies.
Pricing and Model Variants
fal offers two LTX-2 image-to-video endpoints with different cost and capability tradeoffs:
| Model | Endpoint | Cost/Second | Duration | Use Case |
|---|---|---|---|---|
| Pro | fal-ai/ltx-2/image-to-video | $0.06 (1080p), $0.12 (1440p), $0.24 (2160p) | 6-10s | Production quality with audio |
| Fast | fal-ai/ltx-2/image-to-video/fast | $0.04 (1080p), $0.08 (1440p), $0.16 (2160p) | 6-20s | Rapid iteration, prototyping |
A 6-second 1080p video costs $0.36 with the Pro model. The Fast model supports durations up to 20 seconds, but clips longer than 10 seconds require 25 FPS and 1080p resolution.
falMODEL APIs
The fastest, cheapest and most reliable way to run genAI models. 1 API, 100s of models
API Setup
Store your API key as an environment variable:
export FAL_KEY="your-api-key-here"
Install the client library:
npm install --save @fal-ai/client # JavaScript
pip install fal-client # Python
Request Parameters
The endpoint fal-ai/ltx-2/image-to-video accepts two required parameters: image_url (publicly accessible URL or base64 data URI supporting PNG, JPEG, WebP, AVIF, HEIF) and prompt (text describing motion and scene dynamics).
| Parameter | Options | Default |
|---|---|---|
duration | 6, 8, 10 | 6 |
resolution | 1080p, 1440p, 2160p | 1080p |
aspect_ratio | 16:9 | 16:9 |
fps | 25, 50 | 25 |
generate_audio | true, false | true |
import { fal } from "@fal-ai/client";
const result = await fal.subscribe("fal-ai/ltx-2/image-to-video", {
input: {
image_url: "https://your-domain.com/image.jpg",
prompt:
"The camera slowly dollies in as gentle ripples move across the water surface",
duration: 6,
resolution: "1080p",
},
logs: true,
onQueueUpdate: (update) => {
if (update.status === "IN_PROGRESS") {
update.logs.map((log) => log.message).forEach(console.log);
}
},
});
Response Schema
The API returns video metadata:
{
"video": {
"url": "https://v3.fal.media/files/...",
"content_type": "video/mp4",
"file_name": "ltxv-2-i2v-output.mp4"
}
}
The requestId returned by fal.subscribe enables status polling and debugging. Video URLs are temporary; persist outputs to your own storage for production use.
Prompt Structure
Effective prompts describe camera movement and subject motion rather than static scene descriptions. The model responds well to cinematographic language:
- Camera movement: "dollies in," "pans left," "tracking shot," "static wide shot"
- Subject motion: "walks slowly," "hair blowing in wind," "ripples across surface"
- Temporal pacing: "gradually," "sudden," "continuous"
Ineffective: "A woman in a denim jacket on a city street at night"
Effective: "The camera slowly dollies in toward her face as people blur past, city lights flicker and reflections shift across her denim jacket"
Avoid prompts that redescribe the input image. Focus on what should change.
Async and Webhook Integration
For high-volume applications, use the queue API with webhooks instead of blocking:
const { request_id } = await fal.queue.submit("fal-ai/ltx-2/image-to-video", {
input: { image_url, prompt },
webhookUrl: "https://your-domain.com/webhook",
});
The webhook receives the complete response payload when generation finishes. Use request_id to correlate webhook responses with original requests.
Error Handling
Common failure modes:
- Invalid image URL: Image not publicly accessible or returns non-image content type
- Unsupported format: Format not in PNG, JPEG, WebP, AVIF, HEIF
- Rate limiting: Too many concurrent requests from your API key
Constraints
The API currently supports only 16:9 aspect ratio. Input images are cropped to fit so plan compositions accordingly. When generate_audio is set to false, the output video contains no audio track.
Cost Optimization
Resolution dominates cost. A 10-second 4K video costs $2.40, while the same duration at 1080p costs $0.60. For preview workflows, generate at 1080p first, then regenerate approved outputs at higher resolutions.
Cache generated videos with their input parameters (image hash + prompt) to avoid regenerating identical content.
Production Checklist
Before deploying:
- Validate image URLs return 200 status before submitting to avoid wasted API calls
- Implement webhook handlers for async queue results
- Store videos in your own CDN; fal URLs are temporary
- Log
requestIdfor debugging failed generations - Set up cost alerts based on expected usage volume
The fal platform scales automatically. The same integration handles both development testing and production traffic without code changes.
Next Steps
The retake-video endpoint enables selective editing of specific video segments at $0.10/second. For different quality and motion characteristics, evaluate Kling Video v2.6 or Pixverse. Complete API schema and additional examples are available in the fal documentation.
Recently Added
References
-
HaCohen, Y., et al. "LTX-Video: Realtime Video Latent Diffusion." arXiv preprint arXiv:2501.00103, 2024. https://arxiv.org/abs/2501.00103 ↩
