usamaehsan/ltx-av

LTX-2.5 22B distilled: fast expressive text-to-speech (default) plus optional text/image/audio-to-video with native audio, in one model. Designed voices from text, no cloning.

Public
14 runs

Run usamaehsan/ltx-av with an API

Use one of our client libraries to get started quickly. Clicking on a library will take you to the Playground tab where you can tweak different inputs, see the results, and copy the corresponding code to use in your own project.

Input schema

The fields you can use to run this model with an API. If you don't give a value for a field its default value will be used.

Field Type Default value Description
prompt
string
Voice: describe the speaker, the emotion and the room, and put the exact words in quotes, e.g. A tired old man says softly: "I waited all night for you." Quiet room, no music. Video: describe the shot.
video
boolean
False
false = voice only (mp3/wav, fast). true = video with native audio (mp4). Giving an image, keyframe, audio, control video or reference image turns video on.
duration
number
0

Max: 20

Seconds. 0 = auto: voice lets the model pick from the prompt (rounded up to 0.5 s); video follows the audio input if given, else the model picks.
seed
integer
-1
Random seed. -1 = random.
steps
None
8
Voice: denoising steps from the distilled schedule (8 = full quality).
refine
boolean
False
Voice: extra 3-step refine pass.
ancestral
boolean
True
Voice: ancestral sampler (what LTX-2.5 uses).
output_format
None
mp3
Voice: audio format (mp3 = mono 96 kbps, wav = stereo 16-bit).
trim_text
boolean
True
Voice: run the text encoder on the real prompt tokens only (fast).
image
string
Video: optional first-frame image (image-to-video)
keyframe_2
string
Video: optional second keyframe, reached at keyframe_2_at seconds (-1 = last frame).
keyframe_2_at
number
-1

Min: -1

Video: position of keyframe_2 in seconds; -1 = last frame
keyframe_3
string
Video: optional third keyframe
keyframe_3_at
number
-1

Min: -1

Video: position of keyframe_3 in seconds; -1 = last frame
keyframe_4
string
Video: optional fourth keyframe
keyframe_4_at
number
-1

Min: -1

Video: position of keyframe_4 in seconds; -1 = last frame
keyframe_strength
number
1

Max: 1

Video: how hard keyframes are enforced (1.0 = exact).
generated_keyframes
integer
0

Max: 8

Video: model-invented evenly spaced keyframes first. 0 = off.
audio
string
Video: optional audio track to drive the video (lip sync / timing).
audio_start_time
number
0
Video: seconds to skip into the audio file
camera_move
None
none
Video: camera motion. 'none' leaves it to the prompt; 'static' holds it still.
camera_strength
number
1

Max: 2

Video: camera move strength
control_video
string
Video: control-signal video (depth/pose/canny for 'union', motion splines for 'motion_track').
control_type
None
none
Video: which in-context adapter reads control_video.
control_strength
number
1

Max: 1

Video: how closely the output follows control_video
reference_image
string
Video: a person, product, logo or place to put INTO the video (best as a reference sheet on black). Cannot be combined with control_video.
width
integer
768

Min: 256

Max: 1536

Video width
height
integer
1280

Min: 256

Max: 1536

Video height

Output schema

The shape of the response you’ll get when you run this model with an API.

Schema
{
  "type": "string",
  "title": "Output",
  "format": "uri"
}