LTX 2.5

Generate up to 20 seconds 4K videos with sync'd audio.

LTX 2.5: Video With Native Audio, Up to 4K and 20 Seconds

LTX 2.5 is Lightricks’ video model that generates video with synced audio baked in. Write a prompt or attach a starting image, pick a length up to 20 seconds, and get a finished clip with sound at up to 4K.

It comes in two tiers. Pro is the quality pick at 720p or 1080p, 6 to 10 seconds. Fast trades some polish for range: it’s cheaper per second and unlocks 1440p, 4K, 48 fps, and durations up to 20 seconds.

What can you do with it?

  • Generate a scene from text. Describe the shot, the light, the pacing. LTX 2.5 renders the clip and its soundtrack together: ambience, effects, and voices that match the picture.
  • Animate a still image. Attach a first frame and describe the motion. The video starts exactly on your image and follows its aspect ratio.
  • Control the start and end. Add a last frame too and LTX generates the transition between the two.
  • Direct the camera. Pick a camera move (dolly in, dolly out, left, right, jib up, jib down, static, or a focus shift) instead of hoping the prompt lands it.
  • Go long. On the fast tier a single generation runs up to 20 seconds, enough for a full beat instead of a snippet.

How to use it

  1. Write the prompt. Be concrete about subject, camera, light, and sound. Up to 5000 characters, so there’s room for detail.
  2. Attach frames if you have them. A first frame anchors the look; an optional last frame pins the ending.
  3. Pick the tier. Pro for the best image quality, Fast for 4K, high frame rates, and long durations at a lower price.
  4. Set length and size. Leave duration on Auto to let the model fit the length to the content, or pick 6 to 20 seconds. Resolution goes from 720p to 4K, in 16:9 or 9:16.

What you get

One mp4 with native audio, 6 to 20 seconds long, at your chosen resolution and frame rate. Audio is included at every resolution; you can switch it off if you only want the picture.

Tips

  • Describe the sound too. The audio is generated from the same prompt as the video. “Rain on a tin roof, distant thunder” gets you a soundtrack, silence about sound gets you a guess.
  • Use Auto duration for dialogue and action. The model picks a length that fits the motion instead of padding or cutting.
  • Pick a camera move for product and architecture shots. A slow dolly in reads as intentional; unprompted camera drift rarely does.
  • Start at 1080p, upscale the keeper. Iterate cheap, then re-run the winning prompt at 1440p or 4K on the fast tier.
  • Match the frame rate to the feel. 24 fps for a filmic look, 48 or 50 for smooth motion and sports-style clips.

Related tools

Auto-subtitles Brefnet Background Remover Bria Background Remover Crystal Image Upscaler Crystal Video Upscaler Demucs Track Splitter

All AI models & tools

See pricing and start creating on Mitte

Mitte AI Models & Tools Presets Pricing Enterprise Creators Careers About Blog