Example inputhover

fps: 25
prompt: "An elderly couple, an old man and an old woman, sitting side by side on a bench on a grassy hilltop at night, seen from behind, their backs to the camera, both looking up at the sky watching a spectacular meteor shower. Bright shooting stars streak across the starry night sky, leaving glowing trails. Her head leans gently on his shoulder. Wide cinematic shot from behind, deep blue starry sky, warm peaceful romantic atmosphere."
duration: 8
resolution: "1080p"
aspect_ratio: "16:9"
generate_audio: true

Ltx v2.5 | Text to Video | Pro

Video·ltx-v2.5·by LTX

Lightricks LTX 2.5 Text-to-Video creates videos with synchronized audio from text prompts, with 720p or 1080p output for cinematic clip production.

Try it now →

API reference

Runtime (p50): 1m
Estimated price: From $0.12

Call the API

prediction.sh

curl -X POST \
  -H "Authorization: Bearer $EACHLABS_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "ltx-2-5-text-to-video-pro",
    "version": "0.0.1",
    "input": {
        "fps": 25,
        "prompt": "An elderly couple, an old man and an old woman, sitting side by side on a bench on a grassy hilltop at night, seen from behind, their backs to the camera, both looking up at the sky watching a spectacular meteor shower. Bright shooting stars streak across the starry night sky, leaving glowing trails. Her head leans gently on his shoulder. Wide cinematic shot from behind, deep blue starry sky, warm peaceful romantic atmosphere.",
        "duration": 8,
        "resolution": "1080p",
        "aspect_ratio": "16:9",
        "generate_audio": true
    },
    "webhook_url": ""
}' \
  https://api.eachlabs.ai/v1/prediction/

Documentation8 sections

Overview
Ltx v2.5 | Text to Video | Pro Overview

Ltx v2.5 | Text to Video | Pro is the fidelity-focused LTX text-to-video model available through each::labs for generating cinematic clips from natural-language prompts. It is designed to solve a common production problem: turning written scene descriptions into video with synchronized audio while preserving motion quality and prompt fidelity. The model’s main differentiator is its Pro positioning, which emphasizes higher-quality output and production-ready rendering over maximum resolution claims, with support reported for up to 1080p in the Pro API variant.

Within the LTX family, LTX-2.5 introduces native multi-shot generation, improved prompt understanding, and a diffusion video decoder that aims to reduce artifacts in demanding scenes. For creators and teams building short-form video, Ltx v2.5 | Text to Video | Pro is best understood as a tool for polished, prompt-driven clips that need consistent characters, movement, and audio in a single generation.
Capabilities
Capabilities
- Generates video from text prompts with synchronized audio.
- Supports native multi-shot scenes that keep characters, environment, lighting, and voice consistent across cuts.
- Uses a diffusion video decoder designed to improve faces, textures, motion, and on-screen text.
- Handles prompt-rich scenes with stronger instruction following through a custom text encoder and prompt enhancer.
- Supports automatic duration in the LTX-2.5 family for action-aware clip length selection.
- Fits cinematic short-form production where continuity matters across a single generated sequence.
- Supports 16:9 and 9:16 output formats in API variants documented for the family.
- Produces audio-visual content in one pass rather than requiring separate sound design for basic output generation.
Use cases
Use Cases for Ltx v2.5 | Text to Video | Pro

Creators can use Ltx v2.5 | Text to Video | Pro to turn a written scene into a polished clip for reels or pitch decks. A prompt like “A lone climber reaches a snowy ridge at sunrise, cinematic wide shot, slow camera drift, wind ambience” uses synchronized audio and motion continuity for a believable teaser.

Marketers can generate product concept videos with stable lighting and framing. A prompt like “Luxury skincare bottle on a mirrored pedestal, soft studio light, rotating hero shot, clean premium mood” benefits from the model’s prompt adherence and cleaner texture handling.

Designers can prototype motion ideas before committing to a full edit. A prompt like “Abstract glass panels assembling into a logo mark, minimal background, precise motion, elegant reflections” leverages multi-shot consistency and improved visual fidelity.

Developers building video workflows with Ltx v2.5 | Text to Video | Pro API can use it for prompt-driven scene generation in 16:9 or 9:16 outputs. A prompt like “A futuristic dashboard room with holographic UI, subtle camera push, calm mechanical ambience” works well for rapid concept iteration.
Tips & tricks
Tips and Tricks

For better results with Ltx v2.5 | Text to Video | Pro, write prompts that specify the subject, action, scene, camera movement, lighting, and visual style in one compact sentence. LTX-2.5 includes a custom Gemma 4 12B text encoder and a prompt enhancer, so concise prompts can still expand into richer motion guidance, but clear structure helps the model stay on target. When the scene has a natural ending, let the action resolve cleanly instead of packing too many events into one clip.

Useful prompt style examples include: “A documentary-style close-up of a glassmaker shaping molten glass, slow push-in, warm studio lighting, realistic reflections.” “A neon-lit courier rides through a rainy city street, handheld camera, reflective pavement, synced urban ambience.” “A product reveal shot of a matte-black speaker rotating on a pedestal, clean background, soft rim light, premium commercial look.” For workflows that support it, experiment with automatic duration when the action length is unclear.
Technical spec
Technical Specifications
- Model family: LTX-2.5, an open-weights audio-video foundation model with synchronized video and audio generation.
- Primary modality: text-to-video; the family also supports image-to-video and audio-to-video in published documentation.
- Resolution: Pro API reporting indicates up to 1080p; the broader LTX-2.5 release also mentions native 4K HDR support in the family, but the Pro variant is described as stopping at 1080p.
- Aspect ratios: 16:9 and 9:16 are documented for the API variants.
- Clip duration: Pro is reported at 10 seconds; the family also supports automatic duration selection in some workflows.
- Output characteristics: synchronized audio, improved motion, and better handling of textures and on-screen text through a diffusion video decoder.
- Processing time: LTX reports a 10-second 720p clip in 6.8 seconds on 2× GB200 GPUs for the base family release.
- Workflow notes: frame and dimension constraints are documented in ComfyUI-oriented usage, with multi-shot and high-frame-rate workflows supported in the family.
Things to be aware of
Things to Be Aware Of

Ltx v2.5 | Text to Video | Pro performs best when prompts are specific, because vague scene descriptions can lead to weaker camera choices or less stable subject behavior. Multi-shot generation is useful, but complex scene changes still need careful wording to avoid confusion about who or what should remain consistent. High-motion scenes may benefit from the diffusion decoder, yet extreme choreography or crowded action can still produce artifacts.

Users also need to match generation settings to the intended output format and clip length. Overloading a prompt with too many objects, style tags, or actions often reduces clarity rather than improving it. The model is strongest for short cinematic clips and concept videos, not long-form narrative editing. LTX text-to-video workflows can also be compute-intensive depending on the chosen resolution and deployment path.
Key considerations
Key Considerations

Ltx v2.5 | Text to Video | Pro is a strong fit when prompt adherence, shot consistency, and synchronized audio matter more than raw flexibility. The Pro variant is positioned as the fidelity option, so it makes sense for cinematic clips, concept previews, and social video where continuity across motion and sound is important. It is less suitable if you need the highest possible resolution from the LTX family, since published API comparisons place Pro below the family’s 4K-capable path.

Before using it, plan for short, well-structured prompts and enough detail to describe subjects, motion, lighting, and camera direction. The model is most efficient when the creative brief is clear and the clip length is kept within the supported Pro range. For fast iteration, Ltx v2.5 | Text to Video | Pro API is most useful when you want a production-oriented balance of quality and speed rather than maximum experimental control.
Limitations
Limitations

Ltx v2.5 | Text to Video | Pro is not a general-purpose video editor, and it cannot reliably replace frame-accurate manual post-production. Published API comparisons indicate that Pro is limited to 1080p and 10-second output, so it is not the right choice when you need longer sequences or the family’s highest-resolution workflow. It also cannot guarantee perfect realism, exact object placement, or flawless text rendering in every scene. As with most LTX text-to-video systems, unusual prompts and very dense motion can still expose artifacts.