Generate Video and Its Soundtrack Together — from Text, or Text Plus a First Frame
Lightricks' LTX-2.5 (distilled, 22B) generates picture and sound together from a prompt. Add an optional image and it becomes the first frame, animated from there. Output is a 5-second, 24 fps clip at up to 1920×1088, with a synchronized soundtrack muxed in. Runs on CNAPS infrastructure; no external API key required.
What it does
LTX-2.5 is Lightricks' open-weight video and audio model. A prompt alone gives a clip with its own synchronized soundtrack; add an image and it becomes the first frame, animated from there — text-to-video and image-to-video in one node.
The node runs the distilled checkpoint the way Lightricks' official DistilledPipeline does: eight steps at half size, a ×2 latent upscale, then three steps at full size, with no guidance. Detailed prompts work best; the optional Gemma 4 E2B-it prompt enhancer expands a short one.
To make a person in a photo speak along to a voice you provide, use the separate LTX-2.5 Audio-to-Video node.
Problem it solves
- Video + sound from one prompt — No separate Foley or music step for a first draft
- Animate a still — The image becomes the video's first frame
- Landscape or portrait — 1920×1088 (default), 1088×1920, 1536×1024, or 1024×1536
- Short-form clips — Five-second clips (121 frames) at 24 fps
- Better prompts from short ideas — The Gemma 4 E2B-it enhancer expands a one-liner into detailed cinematic instructions
Input/Output
Input
- Text (required) — prompt describing the video and its sound
- Image (optional) — first-frame image. When given, the model animates from it (image-to-video); it is resized and centre-cropped to the chosen resolution. Without it, the whole clip comes from the prompt (text-to-video)
Parameters (main view)
Option | Values · what it does |
Resolution (W×H, ratio) | 1920×1088 (16:9) (default) / 1088×1920 (9:16) / 1536×1024 (3:2) / 1024×1536 (2:3). Lightricks says the model is geared towards 1080p and up; 1536×1024 is the default size of the official pipelines. Portrait sizes are the same canvases turned vertical |
Frame count (8k+1) | 121 frames (~5 s @ 24 fps). LTX needs 8k+1 frames; 121 is the model card's example and the official pipelines' default. The soundtrack covers the same duration |
Advanced Options
Option | Values · what it does |
Prompt enhancer | Off (default) / On. Rewrites the prompt with Gemma 4 E2B-it before generation, expanding a short prompt into richer cinematic instructions (it also sees the image when one is attached). Adds about 10 GB of weights to load |
Random seed | Default -1 (random each run). Set a fixed value to repeat a result |
Output
- Video — generated 24 fps video with its synchronized soundtrack muxed in
How to Use
- Connect a Text Input with a detailed prompt — describe the subject, action, camera, setting, and the sound.
- (Optional) Connect an Image Loader to use an image as the first frame.
- Pick a resolution — 1920×1088 for landscape, 1088×1920 for vertical short-form.
- Connect the output to Video Viewer and run.
- If the result drifts from your idea, write a longer prompt or turn on the prompt enhancer.
Example Workflow
Prompt: "A ceramic cup on a wooden table beside a rain-streaked window. The camera slowly moves closer; soft morning light. Sound: steady rain on glass and a quiet kettle in the background."
Flow: Text Input (+ optional Image Loader) → LTX-2.5 → Video Viewer.
Limitations
- Prompt following depends heavily on prompting style — write a detailed prompt or turn on the prompt enhancer.
- Only the first frame can be set from an image, and the image is centre-cropped to the chosen resolution.
- About 71 GB of weights (81 GB with the prompt enhancer) are read for every clip, so each run takes a while.
How it compares
What you need | Use |
A new shot with sound, from text or a first frame | LTX-2.5 |
A person in one photo speaking a voice you provide | LTX-2.5 Audio-to-Video |
An existing talking video re-synced to a new voice | LatentSync 1.6 |
A deterministic pan or zoom over a still | Ken Burns (Still to Video) |
Model Source
- CNAPS model page: LTX-2.5
- Hugging Face: Lightricks/LTX-2.5
- GitHub: Lightricks/LTX-2
- Paper: arXiv 2601.03233
Compliance & Provenance
Item | Detail |
Provider | Lightricks (open weights) · Specialized |
License | LTX-2.x Community License |
EU AI Act risk class | Limited Risk |
Art. 50 transparency | Required — AI-generated outputs are marked |
For platform classification, acceptable use, and output marking, see the CNAPS AI Policy.
Run it now
Open CNAPS Studio, find LTX-2.5 in the Component panel, and start with the example workflow above.