uv run python scripts/generate.py "A cat walking through a garden" \
--pipeline distilled \
--height 512 --width 768 \
--frames 65 --steps 8
==================================================
LTX-2 MLX Video Generation
Prompt: A cat walking through a garden
Resolution: 768x512, 65 frames
Steps: 8, CFG: 5.0, Seed: 42
Model variant: distilled
Compute dtype: FP16 (memory optimized)
Text encoder: Gemma 3 at weights/gemma-3-12b
Latent shape: 9x16x24
[1/5] Encoding prompt...
Loading tokenizer from weights/gemma-3-12b...
Loading Gemma 3 model...
Loading Gemma shards: 100%|███████████████████████| 5/5 [00:20<00:00, 4.13s/it]
Loaded 626 weight tensors
Loading text encoder projection...
Loading text encoder weights from weights/ltx-2/ltx-2-19b-distilled.safetensors...
Loaded 30 weight tensors
Tokenizing prompt...
Token count: 7/1024
Running Gemma 3 forward pass (48 layers)...
Got 49 hidden states
Processing through text encoder pipeline...
Processing through connector...
libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity)
/opt/anaconda3/lib/python3.12/multiprocessing/resource_tracker.py:254: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d
Always break at connector part. My device is M2 Max 96GB, is that too weak?
uv run python scripts/generate.py "A cat walking through a garden" \
==================================================
LTX-2 MLX Video Generation
Prompt: A cat walking through a garden
Resolution: 768x512, 65 frames
Steps: 8, CFG: 5.0, Seed: 42
Model variant: distilled
Compute dtype: FP16 (memory optimized)
Text encoder: Gemma 3 at weights/gemma-3-12b
Latent shape: 9x16x24
[1/5] Encoding prompt...
Loading tokenizer from weights/gemma-3-12b...
Loading Gemma 3 model...
Loading Gemma shards: 100%|███████████████████████| 5/5 [00:20<00:00, 4.13s/it]
Loaded 626 weight tensors
Loading text encoder projection...
Loading text encoder weights from weights/ltx-2/ltx-2-19b-distilled.safetensors...
Loaded 30 weight tensors
Tokenizing prompt...
Token count: 7/1024
Running Gemma 3 forward pass (48 layers)...
Got 49 hidden states
Processing through text encoder pipeline...
Processing through connector...
libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity)
/opt/anaconda3/lib/python3.12/multiprocessing/resource_tracker.py:254: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d
Always break at connector part. My device is M2 Max 96GB, is that too weak?