LTXV Text to Video
Generate videos from text prompts.
概要
This ComfyUI workflow turns a written prompt into a short video using the LTX-0.9.5 text-to-video checkpoint. It starts by loading the model and text encoder with CheckpointLoaderSimple and CLIPLoader, then converts your prompt and optional negative prompt into embeddings via CLIPTextEncode. Those embeddings are adapted for video with LTXVConditioning so the model can synthesize temporally coherent motion from your text description.
入力するもの
- ワークフローを読み込み中
ComfyUI で実行できます。 このワークフローはこのページでは実行できません。すべてが読み込まれた状態でComfy Cloudで開くか、ご自身のComfyUI用にダウンロードしてください。
Cloudで試すよくある質問
LTX-0.9.5 relies on rich textual conditioning to drive both appearance and motion. Brief prompts often lack detail about subject, setting, and dynamics, leading to generic or unstable video. Describe who/what, where, lighting, motion, and camera behavior.
CheckpointLoaderSimple loads the LTX-0.9.5 text-to-video checkpoint. CLIPLoader loads the CLIP text encoder used by CLIPTextEncode to turn your text and negative text into embeddings for LTXVConditioning.
Length is frames divided by FPS (set in EmptyLTXVLatentVideo and CreateVideo). Increase frames for longer clips. Smoother motion often benefits from more frames and careful prompting; you can also test different samplers in KSamplerSelect and adjust steps in SamplerCustom.
Add targeted negatives (e.g., "motion smear, motion artifacts"), try a different sampler, and tune steps/CFG in SamplerCustom. Reducing resolution can help stability; once the motion looks good, scale up cautiously. Ensure width/height are multiples of 16.