24 credits each

Dataset Captioner

Caption images as before, or analyze training videos and export an editable LTX 2.5 dataset.json.

Images or training videosNo file-count limit. Images up to 20 MB / videos up to 500 MB each.24 credits per file, charged separately when its queue turn starts.

Include in video captions

DialogueMusicSounds / SFX

Caption format / extra instructions

Upload instructions .txt151/1200

LTX format: one concise caption string. Separate visual look, motion, and selected audio. Use [SPEECH], [MUSIC], or [SFX] only when present. For SFX-only captions, uncheck Dialogue and Music.

Dataset media folderExample output: videos/clip-01.mp4. Add your trigger later with process_dataset.py --lora-trigger.

Upload images or training clips to begin.

Editable LTX caption

The generated description will appear here for review.