ComfyUI Node
IronLeo78 LTV (v0.0.5, LTX 2.5)
IronLeo78 LTV v0.0.5: LTX 2.5 image to video, distilled (8 steps), one pass, with sound. Picture + prompt in (size from megapixels, always 24 fps), video out. Link previous_clip to extend a clip instead. Use LLM = ON lets the Huihui LLM write the prompt; the prompt output shows the text used. Link the video to a Save Video node.
IronLeo78 LTV (v0.0.5, LTX 2.5)
- image
- previous_clip
- video
- prompt
◄prompt►
◄duration5.0►
◄megapixels0.5►
◄Use LLMfalse►
◄Continue_frames17►
CategoryIronLeo78/MTV
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | What happens: motion, camera, light, sound, dialogue. One flowing paragraph works best. | |
| duration | FLOAT | 5.00.5–20 | Seconds, always 24 fps. Frames = duration x 24, rounded to 8n + 1 (5 s = 121 frames). |
| megapixels | FLOAT | 0.50.1–2 | Target size. The picture sets the aspect ratio; width and height are multiples of 32 (the picture is cropped lightly to fit). 0.5 = about 960 x 544 in 16:9. |
| Use LLM | BOOLEAN | false | ON: the Huihui LLM looks at the picture and turns your prompt into an LTX prompt (sound and motion included). OFF: your prompt goes to the model as it is. The prompt used is on the prompt output. |
| imageopt | IMAGE | The first frame: the video continues from this picture. Not used when previous_clip is linked. | |
| previous_clipopt | VIDEO | Extend this clip (e.g. the MTV's video): its last frames start the LTX pass, at its size and fps (megapixels is ignored). duration = the NEW seconds. The output is the old clip + the new part. | |
| Continue_framesopt | INT | 171–49 | With previous_clip: how many of its last frames the extension starts from (8n + 1). More carries the motion and the look better; 1 = the last frame only. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| prompt | STRING | — |