What Is LTX Video 2.3?
LTX Video 2.3 is a diffusion-based video generation model from Lightricks, designed for fast, controllable AI video creation. It runs on dedicated GPU infrastructure and processes each job asynchronously, so you can submit a generation and poll for the result without blocking your workflow.
Synclip integrates LTX 2.3 directly into the Video Creator workspace, the AI Canvas node editor, and the public REST API — no separate account or API key needed. Credits are deducted only on success; failed jobs are automatically refunded.
The model supports three generation modes: text-to-video (no image required), image-to-video using a first frame, a last frame, or both frames simultaneously. This flexibility sets it apart from most AI video models that require a first frame before a last frame can be used.
- 5 / 10 / 15 second output durations
- First frame, last frame, or both — each optional and independent
- Fast and Standard quality tiers
- Landscape and portrait orientations
- Available via UI, AI Canvas workflow nodes, and REST API
- 50% off during the launch promotion
LTX 2.3 Fast vs LTX 2.3 Standard — Which Should You Use?
Both models share the same underlying architecture and feature set. The difference is generation time and output fidelity.
LTX 2.3 Fast
- Lower credit cost per video
- Faster turnaround — ideal for iterating on prompt and framing
- Good for social media clips, drafts, and storyboards
- Slightly softer textures at longer durations
- Best paired with a 5s or 10s duration for quick loops
LTX 2.3 Standard
- Higher fidelity motion and texture detail
- Better temporal consistency across 15s clips
- Recommended for final-quality deliverables
- More stable with complex first/last frame combinations
- Use when image prompt alignment matters most
During the launch window both models are at 50% off their standard credit price. The discounted price is shown in the model selector and before you submit a generation.
Choosing the Right Duration: 5s, 10s, or 15s
LTX Video 2.3 supports three fixed output durations. Choosing the right one affects credit cost, rendering time, and how much motion the model needs to hallucinate.
Best for: Product reveals, Instagram Reels, short loops, motion graphics transitions
💡 With a first frame provided, 5s gives very tight control — the model has little room to drift from your anchor image.
Best for: Scene demonstrations, explainer clips, talking-head B-roll, social ads
💡 The sweet spot for most use cases. Enough time for a clear narrative beat without requiring a complex prompt.
Best for: Full scenes, cinematic shots, two-character interactions, long product demos
💡 Use LTX 2.3 Standard for 15s clips — Fast may lose temporal consistency in longer outputs.
Image-to-Video Modes: First Frame, Last Frame, or Both
This is the key differentiator of LTX Video 2.3 compared to most AI video models. Frames are completely independent — you can provide a last frame without a first frame, which opens up reverse-motion and arrival-shot techniques that other models do not support.
First Frame Only (i2v_first)
Provide a starting image. The model generates motion forward from that image. Great for product shots, character reveals, and any scene where the opening composition is critical.
- Product on a surface → camera drift or zoom
- Portrait photo → subtle breathing or eye movement
- Illustration → gentle parallax reveal
Last Frame Only (i2v_last)
Provide only an ending image. The model generates the motion that arrives at that frame. Use this to create "approach shots" or to reverse-engineer a transition. No first frame required — this mode is unique to LTX.
- Camera pushes that land on a specific composition
- Arrival sequences (vehicle, person, object entering scene)
- Reverse shots: show the ending and let AI fill the approach
First + Last Frame (i2v_first_last)
Provide both a start and end frame. The model interpolates and fills the motion between them. This is the most controlled mode and works well for smooth transitions, morphs, and scene-to-scene cuts.
- Transitions between two product images
- Before/after reveals
- Smooth scene interpolation for video editing workflows
Text-to-Video (no images)
No images required. Write a prompt and let the model generate the full scene from scratch. Use this for abstract visuals, B-roll variety, or when you want maximum creative latitude from the model.
- Abstract or atmospheric backgrounds
- Nature, landscape, or environment B-roll
- Concept visualization where no reference exists
Step-by-Step: Generate Your First LTX Video
Follow these steps to generate a video using LTX 2.3 in the Synclip Video Creator.
Step 1 — Select a model
Open the Video Creator workspace. In the model selector at the top, choose either LTX 2.3 Fast or LTX 2.3 Standard. Both show a "50% OFF" badge indicating the launch promotion.
Step 2 — Set duration and orientation
Pick 5s, 10s, or 15s depending on your use case. Select landscape (16:9) or portrait (9:16). LTX does not currently support square orientation.
Step 3 — Upload frames (optional)
If you want image-to-video, upload a first frame, a last frame, or both. You can leave both empty for pure text-to-video. Remember: first and last frames are independent — uploading only a last frame is valid.
Step 4 — Write your prompt
Describe the motion and scene. For image-guided generations, focus your prompt on motion verbs and camera behavior rather than restating what is already in the image.
- Do: "Camera slowly zooms out, shallow depth of field, golden hour light"
- Avoid: "A photo of a product on a marble surface" — the model already sees the frame
Step 5 — Generate and poll
Click Generate. LTX processing takes 30–120 seconds depending on duration and quality tier. The job runs asynchronously — you can navigate away and find the result in My Creations when it completes.
Prompt Templates for LTX Video 2.3
These templates are structured for LTX's motion-forward approach. Include camera action, lighting quality, and motion speed for best results.
Product Reveal (First Frame)
- E-commerce product videos
- Social ad creatives
- Brand launch content
Tip: Pair with a clean product image as the first frame. Keep the background simple — LTX will generate foreground motion more consistently on neutral backgrounds.
Cinematic Landscape B-Roll (Text-to-Video)
Tip: For text-to-video landscapes, be specific about camera movement direction ("gliding left to right" vs "zooming in"). Vague camera descriptions produce less consistent results.
Use 10s or 15s for landscapes — 5s is too short for atmospheric motion to develop.
Arrival Shot (Last Frame Only)
- Character introduction sequences
- Interview opener B-roll
- Product close-up reveal from distance
Tip: Set your desired ending composition as the last frame. The prompt guides camera direction; the frame anchors where you land.
Smooth Transition (First + Last Frame)
Tip: Use visually related images — similar color palettes and brightness levels produce smoother interpolations. Avoid dramatic exposure differences between frames.
Abstract Visual (Text-to-Video)
Tip: Abstract prompts give LTX maximum creative latitude. Pair with 10s or 15s for more motion development. Use LTX 2.3 Standard for richer texture quality.
Frequently Asked Questions
What is LTX Video 2.3?
LTX Video 2.3 is an AI video generation model by Lightricks that creates videos from text prompts or images. It supports 5, 10, and 15 second durations, and uniquely allows independent first and last frame control — you can set only a last frame without needing a first frame.
How is LTX Video different from Sora or Veo?
LTX Video 2.3 processes jobs faster and at lower cost than Sora 2 or Veo 3.1 Pro. Its key differentiator is independent frame control: first and last frames are both optional and can be used separately. Sora and Veo require a first frame before a last frame can be set. LTX is optimized for product content, social media, and high-volume iteration workflows.
Do I need to provide an image to use LTX Video?
No. All three frame modes are optional. You can generate a video from text alone (text-to-video), or add a first frame, last frame, or both to guide the output. Each frame is independent — you do not need to set a first frame to use a last frame.
What durations does LTX Video 2.3 support?
LTX Video 2.3 supports exactly three durations: 5 seconds, 10 seconds, and 15 seconds. Other values are not accepted. Choose based on your use case — 5s for tight loops and product reveals, 10s for most social content, 15s for full scenes.
What is the difference between LTX 2.3 Fast and Standard?
Both models produce videos at the same resolutions and durations. LTX 2.3 Fast generates more quickly and at lower cost, making it ideal for drafts and iteration. LTX 2.3 Standard produces higher fidelity output with more temporal consistency, recommended for final deliverables especially at 15 seconds.
How long does LTX video generation take?
Generation time depends on duration and quality tier. Typical ranges: 5s Fast around 30–50 seconds, 10s Standard around 60–90 seconds, 15s Standard up to 120 seconds. Jobs run asynchronously — you can navigate away and find results in My Creations.
Is LTX Video available via API?
Yes. LTX 2.3 and LTX 2.3 Fast are available via the Synclip REST API using model values "ltx23" and "ltx23fast". You can pass first_frame_url, last_frame_url, or both as optional parameters. Duration must be 5, 10, or 15.
What is the 50% OFF promotion?
During the LTX Video 2.3 launch period, all generations with ltx23 and ltx23fast are priced at half the standard credit rate. The discounted price is shown in the model selector and in the "Estimated credits" field before you submit. The promotion applies automatically — no coupon needed.