---
title: LTX-2.3 | Runware Docs
url: https://runware.ai/docs/models/lightricks-ltx-2-3
description: High-fidelity multimodal video generation with native audio
---
# LTX-2.3

![LTX-2.3](https://assets.runware.ai/covers/lightricks-ltx-2-3.jpg)

LTX-2.3 is a multimodal video generation model that produces synchronized video and audio from text or images. It supports text-to-video and image-to-video workflows with native dialogue and ambient sound generation, emphasizing temporal stability, strong motion coherence, and production-ready output quality for professional creative pipelines. It can also extend an existing clip with new footage at its start or end.

- **Status**: live
- **ID**: `lightricks:ltx@2.3`
- Try in Playground
- **Creator**: Lightricks
- **Release Date**: March 5, 2026
- **Zero Data Retention**: Yes
- **Capabilities**: Text to Video, Image to Video, Video to Video, Audio to Video, Extend, Checkpoint, Audio

## Pricing

$0.04 from, per second · ~57s per request

**Rates** (per second)

- 720p: `$0.04`
- 1080p: `$0.08`
- 2K: `$0.16`
- Extend · 1080p, per second added or used as context: `$0.10`

**6 example generations**

- [first-frame](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-salt-pan-percussion-music-video) · 720p · 5s: `$0.20` (~41s)
- [first-last-frame](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-bacterial-cellulose-lab-b-roll) · 720p · 7s: `$0.28` (~46s)
- [audio-driven](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-hands-free-jar-opener-demo) · 1024×1024 · 8s: `$0.64` (~54s)
- [video-extend](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-ocean-liner-engine-room-extension) · 5s: `$0.80` (~1m)
- [multiple-keyframes](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-mushroom-farm-post-punk-music-video) · 1344×768 · 10s: `$0.80` (~1m 9s)
- [text-to-video](https://runware.ai/docs/models/lightricks-ltx-2-3/examples#example-clockwork-chamber-pop-music-video) · 1080p · 6s: `$0.96` (~1m 11s)

times vary, measured end to end · [How pricing works](https://runware.ai/docs/models-api/pricing)

## Compatibility & Validation

Either provide `inputs.video`, or specify `width/height`.

---

Inside `inputs`, `video` cannot be used with `frameImages` or `audio`.

---

When `inputs.video` is provided, `negativePrompt`, `fps`, `seed`, `steps`, `CFGScale`, `lora`, `safety`, and `settings.enhancePrompt` cannot be used.

---

When `inputs.video` and `width` is provided, `positivePrompt`, `duration`, and `settings.extend` cannot be used.

---

When none of `inputs.video` and `width` are provided, `positivePrompt` and `duration` are required.

---

When `inputs.video` is provided, `duration` must be at least `2`.

---

When `inputs.video` is provided, dimensions are limited to `720×720`, `1080×1080`, `720×900`, `1080×1350`, `900×720`, `1350×1080`, `720×1280`, `1080×1920`, `1280×720` or `1920×1080`.

---

When `settings.extend` is provided, `inputs.video` is required.

---

When `settings.reframe` is provided, `inputs.video` and `width` are required.

---

`width` and `height` must be used together.

## Request Parameters

**API Options**

Platform-level options for task execution and delivery.

### [taskType](https://runware.ai/docs/models/lightricks-ltx-2-3#request-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `videoInference`

Identifier for the type of task being performed.

### [taskUUID](https://runware.ai/docs/models/lightricks-ltx-2-3#request-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier for tracking tasks and matching async responses. Must be unique per task.

### [outputType](https://runware.ai/docs/models/lightricks-ltx-2-3#request-outputtype)

- **Type**: `string`
- **Default**: `URL`

Video output type.

**Allowed values**: `URL`

### [outputFormat](https://runware.ai/docs/models/lightricks-ltx-2-3#request-outputformat)

- **Type**: `string`
- **Default**: `MP4`

Specifies the file format of the generated output. The available values depend on the task type and the specific model's capabilities.

- \`MP4\`: Widely supported video container (H.264), recommended for general use.
- \`WEBM\`: Optimized for web delivery.
- \`MOV\`: QuickTime format, common in professional workflows (Apple ecosystem).

**Allowed values**: `MP4` `WEBM` `MOV`

### [outputQuality](https://runware.ai/docs/models/lightricks-ltx-2-3#request-outputquality)

- **Type**: `integer`
- **Min**: `20`
- **Max**: `99`
- **Default**: `95`

Compression quality of the output. Higher values preserve quality but increase file size.

### [webhookURL](https://runware.ai/docs/models/lightricks-ltx-2-3#request-webhookurl)

- **Type**: `string`
- **Format**: `uri`

Webhook URL that receives JSON responses via HTTP POST when generation tasks complete. For batch requests with multiple results, each completed item triggers a separate webhook call as it becomes available.

**Learn more** (1 resource):

- [Webhooks](https://runware.ai/docs/models-api/webhooks) (platform)

### [deliveryMethod](https://runware.ai/docs/models/lightricks-ltx-2-3#request-deliverymethod)

- **Type**: `string`
- **Default**: `async`

Determines how the API delivers task results.

**Allowed values**:

- `sync` Returns results directly in the API response. Over HTTP, a task still running after 120 seconds continues as an async task and comes back with status 'processing'. Poll for its results using getResponse.
- `async` Returns an immediate acknowledgment with the task UUID. Poll for results using getResponse.

**Learn more** (1 resource):

- [Task Polling](https://runware.ai/docs/models-api/task-polling) (platform)

### [uploadEndpoint](https://runware.ai/docs/models/lightricks-ltx-2-3#request-uploadendpoint)

- **Type**: `string`
- **Format**: `uri`

Specifies a URL where the generated content will be automatically uploaded using the HTTP PUT method. The raw binary data of the media file is sent directly as the request body. For secure uploads to cloud storage, use presigned URLs that include temporary authentication credentials.

**Common use cases:**

- **Cloud storage**: Upload directly to S3 buckets, Google Cloud Storage, or Azure Blob Storage using presigned URLs.
- **CDN integration**: Upload to content delivery networks for immediate distribution.

```text
// S3 presigned URL for secure upload
https://your-bucket.s3.amazonaws.com/generated/content.mp4?X-Amz-Signature=abc123&X-Amz-Expires=3600

// Google Cloud Storage presigned URL
https://storage.googleapis.com/your-bucket/content.jpg?X-Goog-Signature=xyz789

// Custom storage endpoint
https://storage.example.com/uploads/generated-image.jpg
```

The content data will be sent as the request body to the specified URL when generation is complete.

### [safety](https://runware.ai/docs/models/lightricks-ltx-2-3#request-safety)

- **Path**: `safety.checkContent`
- **Type**: `object (2 properties)`

Content safety checking configuration for video generation.

#### [checkContent](https://runware.ai/docs/models/lightricks-ltx-2-3#request-safety-checkcontent)

- **Path**: `safety.checkContent`
- **Type**: `boolean`

Enable or disable content safety checking. Increases total generation time and adds a small charge on top of the generation cost.

#### [mode](https://runware.ai/docs/models/lightricks-ltx-2-3#request-safety-mode)

- **Path**: `safety.mode`
- **Type**: `string`
- **Default**: `fast`

Safety checking mode for video generation.

**Allowed values**:

- `fast` Checks key frames.
- `full` Checks all frames.

### [ttl](https://runware.ai/docs/models/lightricks-ltx-2-3#request-ttl)

- **Type**: `integer`
- **Min**: `60`

Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`.

### [includeCost](https://runware.ai/docs/models/lightricks-ltx-2-3#request-includecost)

- **Type**: `boolean`

Include task cost in the response.

### [numberResults](https://runware.ai/docs/models/lightricks-ltx-2-3#request-numberresults)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `4`
- **Default**: `1`

Number of results to generate. Each result uses a different seed, producing variations of the same parameters.

**Inputs**

Input resources for the task (images, audio, etc). These must be nested inside the \`inputs\` object.

### [frameImages](https://runware.ai/docs/models/lightricks-ltx-2-3#request-inputs-frameimages)

- **Path**: `inputs.frameImages`
- **Type**: `array of strings or objects`
- **Min items**: `1`
- **Max items**: `10`

An array of frame-specific image inputs to guide video generation. Each item can be either a plain image input (UUID, URL, Data URI, or Base64) or an object that pairs an image with a target position in the video.

The `frameImages` parameter allows you to constrain specific frames within the video sequence, ensuring that particular visual content appears at designated points. Position can be specified using `frame` (named positions or frame indices) or `timestamp` (seconds), depending on model support. This is different from `referenceImages`, which provide overall visual guidance without constraining specific timeline positions.

When the `frame` parameter is omitted, automatic distribution rules apply:

- **1 image**: Used as the first frame.
- **2 images**: First and last frames.
- **3+ images**: First and last frames, with intermediate images evenly spaced between.

**Examples**:

**Shorthand format:** When you don't need to specify a frame position, you can pass a plain image input directly.

```json
"frameImages": [
  "aac49721-1964-481a-ae78-8a4e29b91402"
]
```

**Object format:** When you need to specify a position, use an object with `image` and either `frame` or `timestamp` (model-dependent).

```json
"frameImages": [
  {
    "image": "aac49721-1964-481a-ae78-8a4e29b91402",
    "frame": "first"
  }
]
```

**First and last frames:** With two images, they automatically become the first and last frames of the video sequence. You can mix shorthand and object formats.

```json
"frameImages": [
  "aac49721-1964-481a-ae78-8a4e29b91402",
  {
    "image": "3ad204c3-a9de-4963-8a1a-c3911e3afafe",
    "frame": "last"
  }
]
```

**Multiple frames:** With three or more images, the first and last are anchored while intermediate frames are evenly distributed across the timeline.

```json
"frameImages": [
  {
    "image": "aac49721-1964-481a-ae78-8a4e29b91402",
    "frame": "first"
  },
  "c00abf5f-6cdb-4642-a01d-1bfff7bc3cf7",
  {
    "image": "3ad204c3-a9de-4963-8a1a-c3911e3afafe",
    "frame": "last"
  }
]
```

**Format 1: string[]**:

- **Type**: `string`

Image input (UUID, URL, Data URI, or Base64).

**Format 2: object[]**:

#### [image](https://runware.ai/docs/models/lightricks-ltx-2-3#request-inputs-frameimages-format-2-image)

- **Path**: `inputs.frameImages.image`
- **Type**: `string`
- **Required**: true

Image input (UUID, URL, Data URI, or Base64).

#### [frame](https://runware.ai/docs/models/lightricks-ltx-2-3#request-inputs-frameimages-format-2-frame)

- **Path**: `inputs.frameImages.frame`
- **Type**: `string | integer`
- **Min**: `-1`

Target frame position for the image. Accepts a named position like `first` or `last`, or a zero-based frame index (-1 for the last frame).

**Allowed values**: `first` `last`

### [video](https://runware.ai/docs/models/lightricks-ltx-2-3#request-inputs-video)

- **Path**: `inputs.video`
- **Type**: `string`
- **Duration (without width)**: `3s to 3600s`
- **Duration (with width)**: `up to 60s`
- **Area (without width)**: `up to 8.29 MP`
- **Size (with width)**: `64 px to 4096 px`
- **File size**: `up to 100 MB`
- **Formats**: `mp4, mov`

Video input (UUID or URL). Source video to extend or reframe. Without `width` and `height` it is extended: it needs at least 73 frames, and the output keeps the whole source and adds `duration` seconds to it, at 1080x1920 when the source is taller than it is wide and 1920x1080 otherwise. With `width` and `height` it is reframed: it takes at most 1800 frames, and the output keeps its length and shows it whole in the new frame, with the area around it generated.

### [audio](https://runware.ai/docs/models/lightricks-ltx-2-3#request-inputs-audio)

- **Path**: `inputs.audio`
- **Type**: `string`
- **Duration**: `up to 30s`

Audio input (UUID or URL). Enables audio-to-video conditioning. The audio is merged with the generated video and trimmed to fit the video duration. Maximum duration 30 seconds.

**Core Parameters**

Primary parameters that define the task output.

### [model](https://runware.ai/docs/models/lightricks-ltx-2-3#request-model)

- **Type**: `string`
- **Required**: true
- **Value**: `lightricks:ltx@2.3`

Send `auto` instead of an AIR identifier to have the model router choose the model for this task. The rest of the request is unchanged, every value you send is kept, and only the parameters you leave out are filled in. See [Model Router](https://runware.ai/docs/models-api/model-router).

**Learn more** (3 resources):

- [Text To Image: Model Selection](https://runware.ai/docs/learn/text-to-image#model-selection) (learn)
- [Image Inpainting: Model Specialized Inpainting Models](https://runware.ai/docs/learn/image-inpainting#model-specialized-inpainting-models) (learn)
- [Image Outpainting: Other Critical Parameters](https://runware.ai/docs/learn/image-outpainting#other-critical-parameters) (learn)

### [positivePrompt](https://runware.ai/docs/models/lightricks-ltx-2-3#request-positiveprompt)

- **Type**: `string`
- **Required**: conditional — Required when inputs.video is not provided
- **Min**: `2`
- **Max**: `10000`

Text prompt describing elements to include in the generated output.

**Learn more** (1 resource):

- [Prompts](https://runware.ai/docs/learn/prompts) (learn)

### [negativePrompt](https://runware.ai/docs/models/lightricks-ltx-2-3#request-negativeprompt)

- **Type**: `string`

Prompt to guide what to exclude from generation. Ignored when guidance is disabled (CFGScale ≤ 1).

**Learn more** (1 resource):

- [Prompts: Negative Prompts](https://runware.ai/docs/learn/prompts#negative-prompts) (learn)

### [width](https://runware.ai/docs/models/lightricks-ltx-2-3#request-width)

- **Type**: `integer`
- **Required**: conditional — Required when inputs.video is not provided or settings.reframe is provided
- **Min**: `128`
- **Max**: `2048`

Width of the generated media in pixels.

**Learn more** (2 resources):

- [Dimensions](https://runware.ai/docs/learn/dimensions) (learn)
- [Image Outpainting: Dimensions Critical For Outpainting](https://runware.ai/docs/learn/image-outpainting#dimensions-critical-for-outpainting) (learn)

### [height](https://runware.ai/docs/models/lightricks-ltx-2-3#request-height)

- **Type**: `integer`
- **Required**: conditional — Required when inputs.video is not provided
- **Min**: `128`
- **Max**: `2048`

Height of the generated media in pixels.

**Learn more** (2 resources):

- [Dimensions](https://runware.ai/docs/learn/dimensions) (learn)
- [Image Outpainting: Dimensions Critical For Outpainting](https://runware.ai/docs/learn/image-outpainting#dimensions-critical-for-outpainting) (learn)

### [duration](https://runware.ai/docs/models/lightricks-ltx-2-3#request-duration)

- **Type**: `float`
- **Required**: conditional — Required when inputs.video is not provided
- **Min**: `1`
- **Max**: `20`

Length of the generated video in seconds. The total number of frames produced is determined by duration multiplied by the model's frame rate (fps). With `inputs.video`, the seconds added to the source video.

### [fps](https://runware.ai/docs/models/lightricks-ltx-2-3#request-fps)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `120`
- **Default**: `25`

Frames per second for video generation. Higher values create smoother motion but require more processing time.

### [seed](https://runware.ai/docs/models/lightricks-ltx-2-3#request-seed)

- **Type**: `integer`
- **Min**: `0`
- **Max**: `9223372036854776000`

Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range.

**Learn more** (1 resource):

- [Seed](https://runware.ai/docs/learn/seed) (learn)

### [steps](https://runware.ai/docs/models/lightricks-ltx-2-3#request-steps)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `100`
- **Default**: `20`

Total number of denoising steps. Higher values generally produce more detailed results but take longer.

**Learn more** (1 resource):

- [Steps](https://runware.ai/docs/learn/steps) (learn)

### [CFGScale](https://runware.ai/docs/models/lightricks-ltx-2-3#request-cfgscale)

- **Type**: `float`
- **Min**: `1`
- **Max**: `20`
- **Step**: `0.01`
- **Default**: `4`

Guidance scale representing how closely the output will resemble the prompt. Higher values produce results more aligned with the prompt.

**Learn more** (1 resource):

- [Cfg Scale](https://runware.ai/docs/learn/cfg-scale) (learn)

### [lora](https://runware.ai/docs/models/lightricks-ltx-2-3#request-lora)

- **Path**: `lora.model`
- **Type**: `array of objects (3 properties)`
- **Min items**: `1`

With LoRA (Low-Rank Adaptation), you can adapt a model to specific styles or features by emphasizing particular aspects of the data. This technique enhances the quality and relevance of generated content and can be especially useful when the output needs to adhere to a specific artistic style or follow particular guidelines.

Multiple LoRA models can be used simultaneously to achieve different adaptation goals.

**Examples**:

```json
"lora": [
  {
    "model": "<lora-model-air>",
    "weight": 0.8
  }
]
```

**Learn more** (1 resource):

- [Loras](https://runware.ai/docs/learn/loras) (learn)

#### [model](https://runware.ai/docs/models/lightricks-ltx-2-3#request-lora-model)

- **Path**: `lora.model`
- **Type**: `string`
- **Required**: true

LoRA model identifier.

#### [weight](https://runware.ai/docs/models/lightricks-ltx-2-3#request-lora-weight)

- **Path**: `lora.weight`
- **Type**: `float`
- **Min**: `-4`
- **Max**: `4`
- **Step**: `0.01`
- **Default**: `1`

Strength of the LoRA influence. A value of 0 means no influence. Higher values increase the influence, and negative values can be used to steer away from the LoRA's style.

#### [transformer](https://runware.ai/docs/models/lightricks-ltx-2-3#request-lora-transformer)

- **Path**: `lora.transformer`
- **Type**: `string`
- **Default**: `both`

Transformer stages to apply LoRA. Some video models use separate high-noise and low-noise processing stages, and LoRAs can be selectively applied to optimize their effectiveness.

**Allowed values**:

- `high` Apply LoRA only to the high-noise processing stage (coarse structure and early generation steps).
- `low` Apply LoRA only to the low-noise processing stage (fine details and later generation steps).
- `both` Apply LoRA to both stages for full coverage.

**Settings**

Technical parameters to fine-tune the inference process. These must be nested inside the \`settings\` object.

### [enhancePrompt](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-enhanceprompt)

- **Path**: `settings.enhancePrompt`
- **Type**: `boolean`
- **Default**: `true`

Enable automatic prompt enhancement for cinematic results.

### [extend](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-extend)

- **Path**: `settings.extend`
- **Type**: `object (2 properties)`

Extension controls for the source video.

#### [context](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-extend-context)

- **Path**: `settings.extend.context`
- **Type**: `float`
- **Min**: `1`
- **Max**: `20`

Seconds of the source video used as context for the new seconds. Must not exceed the source length, and `context` plus `duration` must not exceed 505 frames at the source frame rate (about 21 seconds at 24 fps).

#### [mode](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-extend-mode)

- **Path**: `settings.extend.mode`
- **Type**: `string`
- **Default**: `end`

Side of the source video that receives the new seconds.

**Allowed values**:

- `end` Adds the new seconds after the last frame of the source video.
- `start` Adds the new seconds before the first frame of the source video.

### [reframe](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe)

- **Path**: `settings.reframe`
- **Type**: `object (6 properties)`

Reframe controls for the source video.

#### [placement](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement)

- **Path**: `settings.reframe.placement`
- **Type**: `object (5 properties)`

Position of the source video in the output frame, given as an `anchor` or as a rectangle of `x`, `y` and at least one of `width` and `height`. Omitted, the video is centered at the largest size that fits. A rectangle side left out follows the video's aspect ratio, both sides together must match it within one pixel, and the placed video must be at least 1 pixel per side and at most 2x its original size.

##### [width](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement-width)

- **Path**: `settings.reframe.placement.width`
- **Type**: `float`
- **Max**: `1`

Width of the video, as a fraction of the output width. `x` plus `width` must not exceed 1.

##### [height](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement-height)

- **Path**: `settings.reframe.placement.height`
- **Type**: `float`
- **Max**: `1`

Height of the video, as a fraction of the output height. `y` plus `height` must not exceed 1.

##### [anchor](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement-anchor)

- **Path**: `settings.reframe.placement.anchor`
- **Type**: `string`

Edge or corner of the output frame the video is placed against, at the largest size that fits. Anchors other than 'center' are rejected when that fit enlarges the video more than 2x.

**Allowed values**: `center` `top` `bottom` `left` `right` `top_left` `top_right` `bottom_left` `bottom_right`

##### [x](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement-x)

- **Path**: `settings.reframe.placement.x`
- **Type**: `float`
- **Min**: `0`

Left edge of the video, as a fraction of the output width.

##### [y](https://runware.ai/docs/models/lightricks-ltx-2-3#request-settings-reframe-placement-y)

- **Path**: `settings.reframe.placement.y`
- **Type**: `float`
- **Min**: `0`

Top edge of the video, as a fraction of the output height.

## Response Parameters

### [taskType](https://runware.ai/docs/models/lightricks-ltx-2-3#response-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `videoInference`

Identifier for the type of task this response belongs to.

### [taskUUID](https://runware.ai/docs/models/lightricks-ltx-2-3#response-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier echoed from the original request, used to match async responses to their tasks.

### [videoUUID](https://runware.ai/docs/models/lightricks-ltx-2-3#response-videouuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID of the output video.

### [videoURL](https://runware.ai/docs/models/lightricks-ltx-2-3#response-videourl)

- **Type**: `string`
- **Format**: `uri`

URL of the output video.

### [videoBase64Data](https://runware.ai/docs/models/lightricks-ltx-2-3#response-videobase64data)

- **Type**: `string`

Base64-encoded video data.

### [videoDataURI](https://runware.ai/docs/models/lightricks-ltx-2-3#response-videodatauri)

- **Type**: `string`
- **Format**: `uri`

Data URI of the output video.

### [seed](https://runware.ai/docs/models/lightricks-ltx-2-3#response-seed)

- **Type**: `integer`

The seed used for generation. If none was provided, shows the randomly generated seed.

### [NSFWContent](https://runware.ai/docs/models/lightricks-ltx-2-3#response-nsfwcontent)

- **Type**: `boolean`

Flag indicating if NSFW content was detected.

### [cost](https://runware.ai/docs/models/lightricks-ltx-2-3#response-cost)

- **Type**: `float`

Task cost in USD. Present when `includeCost` is set to `true` in the request.

## Examples

### text-to-video (text-to-video)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/d0338f5a-8393-4f30-973d-9204c51a2660.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "9073a860-902e-4beb-8114-83e9cf9b445b",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "A polished six-second chamber-pop music-video segment set inside a bright, stylized watchmaker’s workshop. Three adult watchmakers in crisp seafoam coveralls sit shoulder to shoulder at one long ivory workbench covered with brass gears, loupe lenses and open mechanical watches. In precise choreography, they tap tiny tools against brass trays, spin three oversized balance wheels, then sweep their hands aside simultaneously as dozens of gears rotate in place. Begin with an extreme macro shot of a ticking escapement; the camera rapidly pulls backward along the bench into a symmetrical medium-wide composition while the performers’ rhythmic motions remain sharp and perfectly synchronized. Warm morning sunlight streams through frosted factory windows, creating clean geometric beams, soft amber highlights and delicate shadows across an airy palette of ivory, seafoam, copper and pale apricot. Elegant practical production design, tactile metal detail, subtle 35mm film grain, playful but sophisticated editorial energy. The soundtrack is a tightly synchronized instrumental phrase combining crisp clock ticks, delicate brass-tool percussion, pizzicato strings and one warm analog-synth bass pulse; every visible tap and spinning mechanism lands on the beat. No singing, dialogue, lettering or logos.",
  "negativePrompt": "dark noir lighting, neon alley, rooftop, storm, science-fiction setting, steampunk costumes, chaotic movement, handheld shake, distorted hands, extra fingers, duplicated people, floating tools, illegible text, captions, logos, watermarks",
  "width": 1920,
  "height": 1088,
  "duration": 6
}
```

---

### first-frame (first-frame)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/303dc5cc-7e0f-49a3-a76e-c372dd031cba.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "6d4fdfa6-ad94-4231-8e37-fbfef8270e3e",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "Continue directly from the supplied opening frame as a polished music-video performance shot. In one uninterrupted five-second take, the percussionist drives her raised mallet down for a powerful first strike; the drumhead visibly flexes at impact and a small burst of salt jumps beneath her boots. She immediately pivots counterclockwise in a forceful, grounded dance step while the camera makes a smooth low arc in the opposite direction, creating strong parallax across the turquoise evaporation ponds and distant salt mounds. Her burnt-orange sleeves and trouser legs snap naturally in the dry wind, the drum remains securely strapped to her body, and her movement has convincing weight and rhythm. She completes the turn facing near camera and lands a second precise strike on the final beat. Preserve the real working salt-flat setting, expansive 1280x704 landscape framing, bleached white, turquoise and burnt-orange palette, low golden sunlight, long shadows, tactile salt crystals and subtle 16mm grain. Synchronized audio: two deep resonant bass-drum impacts, a sparse distorted electronic bass pulse, boots crunching on salt and steady open wind; no vocals or speech. Premium art-directed folk-electronic music video, coherent anatomy, stable wardrobe and instrument, realistic motion, no cuts.",
  "negativePrompt": "Lyrics, captions, titles, logos, extra people, extra limbs, duplicated mallets, warped drum, floating objects, costume changes, jump cuts, camera shake, slow motion, fantasy effects, neon lighting, night scene, rain, dialogue.",
  "width": 1280,
  "height": 704,
  "duration": 5,
  "inputs": {
    "frameImages": [
      "https://assets.runware.ai/assets/inputs/abc90ebf-3af2-4899-b620-3b49da70a132.jpg"
    ]
  }
}
```

---

### first-last-frame (first-last-frame)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/aaa2e849-b9d7-481c-ab2f-8082fce4c430.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "3ba49d0d-67ba-42d5-9af1-d87c2384b076",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "Create a single continuous vertical documentary b-roll shot connecting the supplied first and last frames. In a clean biomaterials laboratory, the technician steadily lifts the wet bacterial-cellulose sheet from the shallow fermentation tray by its two near corners. The cellulose separates from the liquid with believable adhesion and weight, unfolding into one coherent translucent ivory sheet without tearing, stretching unnaturally, or changing shape. Clear liquid streams from the lower edge and falls back into the tray in fine droplets, producing small concentric ripples. Her hands rise together at a measured pace while her posture and focused expression remain calm and consistent. The camera performs a subtle stabilized push-in with a gentle upward tilt, naturally following the sheet from the tray to its upright final position and arriving precisely at the supplied last frame. Preserve the technician's identity, clothing, safety equipment, hand anatomy, laboratory layout, lighting direction, and restrained ivory, pale-aqua, silver, and green-gray palette throughout. Soft diffuse morning light reveals the sheet's fibrous texture and wet sheen. Shallow depth of field, premium editorial science footage, realistic physics, temporal stability, no cuts, no time lapse, no dialogue. Synchronized location audio only: quiet laboratory ventilation, a soft wet peel as the cellulose releases, delicate dripping water, and faint liquid ripples in the steel tray.",
  "negativePrompt": "Text, captions, logos, labels, branding, jump cuts, time lapse, duplicated hands, extra fingers, fused gloves, identity changes, wardrobe changes, disappearing objects, warped tray, torn sheet, rubbery material, excessive splashing, camera shake, dramatic color shifts, music, speech.",
  "width": 704,
  "height": 1280,
  "duration": 7,
  "inputs": {
    "frameImages": [
      {
        "image": "https://assets.runware.ai/assets/inputs/20314441-db62-45ad-afcd-37ac178cf77c.jpg",
        "frame": "first"
      },
      {
        "image": "https://assets.runware.ai/assets/inputs/023ff0ab-d645-4bf9-ae3f-1050f8b932da.jpg",
        "frame": "last"
      }
    ]
  }
}
```

---

### multiple-keyframes (multiple-keyframes)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/4ec5f71e-4123-495f-99bc-f152482e07fc.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "a622acc3-2503-41e5-a07c-56e115d2e0b6",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "Create a polished ten-second music-video segment that connects the three supplied keyframes as one continuous, chronologically coherent tracking shot. Preserve the exact identity, age, silver bob, orange waterproof bib overalls, teal sweater and white boots of the woman, along with the same oyster mushroom room, rack layout and lighting throughout. Begin with the symmetrical wide view at the aisle entrance. The camera smoothly dollies backward as she advances toward it, transforming practical night-shift movements into precise post-punk choreography: two measured steps, a shoulder snap, a sharp side-step and one sweeping arm between the racks. Arrive naturally at the supplied middle pose without a cut. Continue retreating down the aisle as she pivots, crosses her feet and glides toward the ventilation fan; mushrooms and racks pass with convincing parallax while fluorescent fixtures remain temporally stable. Reach the later low-angle pose beside the fan, then let her hold the backward lean for the final beat as the fan gently lifts her sweater and hair. Grounded human anatomy, realistic foot contact, controlled expressive performance, coherent industrial environment, subtle handheld texture over a smooth dolly move, crisp detail, restrained 1980s factory-video character without imitation or retro degradation. Chalk-white mushrooms and gray metal dominate, punctuated by teal fabric and safety orange. Generate synchronized audio as a dry minimalist post-punk instrumental: steady motorik kick and snare, rubbery bass guitar, clipped muted guitar accents and the room's ventilation hum tuned into the rhythm. Choreographic accents land exactly on the drum hits. No singing, speech, captions, logos or on-screen text.",
  "negativePrompt": "Cuts, scene changes, extra people, duplicated limbs, identity changes, wardrobe changes, disappearing mushrooms, warped racks, unstable architecture, floating feet, exaggerated fantasy spores, fog, neon lighting, camera whip, strobing, subtitles, lyrics, logos, watermarks",
  "width": 1344,
  "height": 768,
  "duration": 10,
  "inputs": {
    "frameImages": [
      {
        "image": "https://assets.runware.ai/assets/inputs/cdb66638-6612-4785-ac54-77e1f9fccdcf.jpg",
        "frame": "first"
      },
      {
        "image": "https://assets.runware.ai/assets/inputs/befc8583-8f74-4aa0-9377-c265f5ba4fb6.jpg",
        "frame": 100
      },
      {
        "image": "https://assets.runware.ai/assets/inputs/21ee7cb6-624c-4882-b641-043e8d3b0bca.jpg",
        "frame": 200
      }
    ]
  }
}
```

---

### audio-driven (audio-driven)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/d7c3d28b-f598-41d1-90e9-e51e8199d7e0.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "4074b772-f214-4f28-bafc-2bf713928bcc",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "A polished eight-second square product demonstration for an accessible countertop automatic jar opener, presented in one continuous stable shot. Bright, optimistic assistive-living catalog art direction with soft daylight, a peach laminate counter, pale aqua tiled background, and restrained cobalt accents. The compact opener has a matte ivory body, cobalt non-slip base, hinged upper arm, and a clear circular gripping collar. No logo, labels, captions, or on-screen text. Three-quarter overhead close view keeps the entire device, glass raspberry jam jar, lid, controls, and the demonstrator's hands clearly visible. Synchronize each action precisely to the supplied narration: on “Place the jar on the base,” an older adult's hand sets the closed jar upright into the cobalt cradle; on “Lower the arm,” the hand smoothly lowers the hinged mechanism over the metal lid; on “Press once,” one finger presses the single large tactile button, then both hands withdraw fully; on “The opener grips, turns,” the clear collar closes around the lid and rotates it with convincing controlled torque while the jar remains stable; on “and releases the lid hands-free,” the arm rises slightly and reveals the loosened lid resting neatly above the open jar. Finish with the open jar and raised mechanism held cleanly for recognition. Realistic mechanical movement, consistent product geometry, natural hand anatomy, crisp commercial lighting, subtle contact shadows, premium industrial design, calm reassuring mood, no cuts, no camera shake, no extra objects entering frame.",
  "width": 1024,
  "height": 1024,
  "duration": 8,
  "inputs": {
    "audio": "https://assets.runware.ai/assets/inputs/e0b477f8-0ae0-4251-8080-285546f18d49.mp3"
  }
}
```

---

### video-extend (video-extend)

[Watch video](https://assets.runware.ai/examples/lightricks-ltx-2-3/86679214-192a-4de4-9b60-b9ba1196d3d5.mp4)

```json
{
  "taskType": "videoInference",
  "taskUUID": "95c8e17e-fefa-419c-9716-1d4c6930ede2",
  "model": "lightricks:ltx@2.3",
  "positivePrompt": "Immediately after the source clip's final frame, the same marine engineer continues walking at exactly the same pace and passes through the open watertight bulkhead. The stabilized camera follows directly behind him with uninterrupted forward tracking, preserving the original lens, height, composition and natural walking motion. As the camera crosses the threshold, the plain corridor opens naturally into the full environment: a vast, historically accurate 1920s ocean-liner engine room extending several decks upward and deep into the ship, with monumental triple-expansion steam engines, steel catwalks, ladders, gauges, polished connecting rods and dense period pipework. The engineer remains identical in clothing, proportions and movement. Match the muted blue-gray steel, dim amber caged lighting, damp industrial surfaces and documentary realism of the source. The machinery operates with slow, heavy synchronized motion while faint steam drifts near the floor. No cut, transition, time jump or change to anything already visible in the source. Synchronized audio continues seamlessly with his boots ringing on steel grating, a deep mechanical thrum, rhythmic piston movement, subtle pipe vibration and occasional restrained steam hiss. Over five seconds, the camera follows him several steps into the engine room and finishes on a wide, immersive reveal of its immense scale.",
  "duration": 5,
  "settings": {
    "extend": {
      "mode": "end",
      "context": 3
    }
  },
  "inputs": {
    "video": "https://assets.runware.ai/assets/inputs/92b1c584-826f-476f-ae04-7ec4319ac8f7.mp4"
  }
}
```