CoolFace
Apppublic

R1000/ltx-2.3-cpu

sourceHugging Faceotherupdated 4mo agoView on Hugging Face
0likes
App README

LTX 2.3 CPU

[LTX 2.3](https://huggingface.co/Lightricks/LTX-2.3) CPU, cond_safe distill 1.1 + Sulphur-2 merge = 10Eros.

How it works

ComfyUI runs headless with --cpu --force-fp32 --cache-none. Models are GGUF (mmap'd from disk, per-layer dequant). Workflow matches RuneXX's proven LTX 2.3 GGUF pattern: DualCLIPLoaderGGUF + SamplerCustomAdvanced + CFGGuider.

ComponentSourceSizeFormat
<!--DiTunsloth/LTX-2.3-GGUF distilled-1.113.4 GBUD-Q3KM GGUF (Unsloth Dynamic 2.0)-->
DiTunsloth/LTX-2.3-GGUF distilled-1.1-Q2_K7.9 GBQ2_K GGUF
Text encoderunsloth/gemma-3-12b-it-GGUF~6 GBQ3KM GGUF
Text projectionKijai/LTX2.3_comfy2.2 GBbf16 safetensors
VAEKijai/LTX2.3_comfy (TaeL tiny VAE)22 MBsafetensors
LoRATenStrip/LTX2.3_Distilled_Lora_1.1_Experiments cond_safe662 MBsafetensors

Performance

First call downloads ~20 GB of models (cached on 50 GB ephemeral disk). Verified benchmark:

ResolutionDurationStepsTimeHardware
512x3202 sec8 (distilled-1.1)74 minfree CPU (2 vCPU)

Curl test

bash
TOKEN="hf_xxx"
SPACE="https://werecooking-ltx-2-3-cpu.hf.space"

EVT=$(curl -s -X POST "$SPACE/gradio_api/call/generate" \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{"data":["A woman walking through a neon-lit Tokyo alley", 2.0, 8, -1]}' \
  | python -c "import sys,json;print(json.load(sys.stdin)['event_id'])")
curl -sN "$SPACE/gradio_api/call/generate/$EVT" -H "Authorization: Bearer $TOKEN"

Logs

bash
curl -N -H "Authorization: Bearer $TOKEN" "https://huggingface.co/api/spaces/WeReCooking/ltx-2.3-cpu/logs/build"
curl -N -H "Authorization: Bearer $TOKEN" "https://huggingface.co/api/spaces/WeReCooking/ltx-2.3-cpu/logs/run"

Credits

License

Same as upstream LTX-2 Community License.