It works fine on 8 GB VRAM might work on 6 GB too

[nitinmukesh]

Check this out
https://youtu.be/nur4_b4yzM0

[darkstorm2150]

Just to verify this isn't related to VRAM right ?

(LTXVideo) → C:\Users\thesi\Desktop\LTXVideo [main ≡ +10 ~1 -0 !]› python .\inference.py WARNING:__main__:Running generation with arguments: {'ckpt_dir': 'Lightricks/LTX-Video', 'num_inference_steps': 20, 'guidance_scale': 3.5, 'height': 512, 'width': 768, 'num_frames': 86, 'frame_rate': 25, 'prompt': "A french woman with blonde hair styled up, wearing a black dress with sequins and pearl earrings, looks down with a sad expression on her face. The camera remains stationary, focused on the woman's face. The lighting is dim, casting soft shadows on her face. The scene appears to be from a movie.", 'negative_prompt': 'low quality, worst quality, deformed, distorted, disfigured, motion smear, motion artifacts, fused fingers, bad anatomy, weird hand, ugly', 'seed': 55234234, 'output_path': 'outputs', 'num_images_per_prompt': 1, 'input_image_path': '', 'input_video_path': '', 'bfloat16': True, 'int8': False, 'disable_load_needed_only': False} WARNING:__main__:Padded dimensions: 512x768x89 Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 9.94it/s] Traceback (most recent call last): File "C:\Users\thesi\Desktop\LTXVideo\inference.py", line 369, in <module> main() File "C:\Users\thesi\Desktop\LTXVideo\inference.py", line 273, in main images = pipeline( File "C:\Users\thesi\.conda\envs\LTXVideo\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "C:\Users\thesi\Desktop\LTXVideo\ltx_video\pipelines\pipeline_ltx_video.py", line 928, in __call__ latents = self.prepare_latents( File "C:\Users\thesi\Desktop\LTXVideo\ltx_video\pipelines\pipeline_ltx_video.py", line 685, in prepare_latents latents = randn_tensor( File "C:\Users\thesi\.conda\envs\LTXVideo\lib\site-packages\diffusers\utils\torch_utils.py", line 67, in randn_tensor raise ValueError(f"Cannot generate a {device} tensor from a generator of type {gen_device_type}.") ValueError: Cannot generate a cpu tensor from a generator of type cuda.

[nitinmukesh]

Try again.
It happens sometimes when the code tries to unload text_encoder to CPU to save VRAM.