ltx-2.3-22b-distilled-lora-384-1.1.safetensors
LTX 2.3 Distilled LoRA 384 v1.1 (Official)
Official Lightricks rank-384 distillation LoRA, v1.1. Applied on the dev model, it converts dev-style inference (many steps, CFG > 1) into distilled-style (8 steps, CFG = 1) — the official high-rank counterpart to Kijai's dynamic-rank LoRA. Place in models/loras/.
Released 2026-04-27 · Source: Lightricks/LTX-2.3 (HuggingFace) — Official v1.1 rank-384 distillation LoRA. Higher-rank counterpart to Kijai's dynamic-rank LoRAs.
Download ltx-2.3-22b-distilled-lora-384-1.1.safetensors
Direct HuggingFace download. 7.61 GB · Free.
No 16GB GPU? Try ltx-2.3-22b-distilled-lora-384-1.1.safetensors online — free generation included
Skip the 7.61 GB download and ComfyUI setup. Generate a 6-second video using this exact model in your browser, ~30 seconds.
Technical details
ltx-2.3-22b-distilled-lora-384-1.1.safetensors is Lightricks' official v1.1 distillation LoRA at a fixed rank of 384. 'Rank 384' is the LoRA's inner dimension applied uniformly across layers — much higher than Kijai's dynamic-rank LoRAs (which average ~105–111), which is why this file is ~7.6 GB versus their ~2.7 GB. The higher rank captures more of the dev→distilled transformation, trading file size and a little VRAM for fidelity.
Like all distillation LoRAs, it is applied on top of the dev model and only does its job when the sampler is set to the distilled profile: 8 steps, CFG = 1. Loading it without changing the sampler produces nothing useful — the LoRA modifies how the model responds to those specific settings.
This is the official Lightricks LoRA. Kijai's dynamic-rank LoRA (ltx-2.3-22b-distilled-1.1_lora-dynamic_fro09_avg_rank_111_bf16.safetensors) is the lighter community alternative; both target the same v1.1 distilled behavior.
When to choose ltx-2.3-22b-distilled-lora-384-1.1.safetensors
Pick the rank-384 official LoRA when you want the closest match to the official distilled checkpoint's quality via a LoRA, and you have VRAM to spare for a 7.6 GB adapter on top of the dev model. It is the highest-fidelity distillation LoRA available.
If VRAM is tight on a 16 GB card, use the Kijai dynamic-rank LoRA instead (~2.7 GB, rank ~111) — nearly the same speedup at a fraction of the footprint.
If you do not need LoRA flexibility at all, the standalone distilled FP8 checkpoint (ltx-2.3-22b-distilled-1.1_transformer_only_fp8_scaled.safetensors) is simpler — one file load, no LoRA composition.
Prefer this v1.1 over the v1.0 rank-384 LoRA (without '-1.1') unless you are reproducing an older result.
Will this run on my GPU?
Minimum: 16GB VRAM.
Recommendation: Pair with the dev model (dev FP8 on 16 GB, BF16 dev on 32 GB+) and set the sampler to 8 steps, CFG = 1. At rank 384 / 7.6 GB it is the heaviest, highest-fidelity distillation LoRA — prefer it when you want maximum distilled quality and have the VRAM headroom.
How to use ltx-2.3-22b-distilled-lora-384-1.1.safetensors
- Download the file from HuggingFace.
- Place it in ComfyUI/models/loras/ inside your ComfyUI directory.
- Restart ComfyUI (or refresh the model list from the menu).
- Load a compatible workflow — see below.
Compatible official workflows:
- LTX-2.3_T2V_I2V_Single_Stage_Distilled_Full.json— T2V / I2V Single Stage Distilled
- LTX-2.3_T2V_I2V_Two_Stage_Distilled.json— T2V / I2V Two Stage Distilled
- LTX-2.3_ICLoRA_Union_Control_Distilled.json— ICLoRA Union Control Distilled
- LTX-2.3_ICLoRA_Motion_Track_Distilled.json— ICLoRA Motion Track Distilled
- LTX-2.3_ICLoRA_HDR_Distilled.json— ICLoRA HDR Distilled
Don't want to run this locally? Try ltx-2.3-22b-distilled-lora-384-1.1.safetensors online with a free generation — no GPU, no install, ~30 seconds per clip.
ComfyUI says it can't find ltx-2.3-22b-distilled-lora-384-1.1.safetensors?
Some published workflow JSONs reference this file under a custom subdirectory. If ComfyUI shows a "cannot find model" error and your workflow references one of these path-prefixed variants:
- loras/ltx-2.3-22b-distilled-lora-384-1.1.safetensors
- ltx\ltx-2.3-22b-distilled-lora-384-1.1.safetensors
- ltx23\ltx-2.3-22b-distilled-lora-384-1.1.safetensors
The prefix before the slash or backslash is a subdirectory the workflow author used. The actual file is the same ltx-2.3-22b-distilled-lora-384-1.1.safetensors — you have two fixes:
- Create the matching subdirectory inside ComfyUI/models/loras/ and place the file there. Example: if the workflow references loras/ltx-2.3-22b-distilled-lora-384-1.1.safetensors, create the corresponding subfolder under ComfyUI/models/loras/ and put ltx-2.3-22b-distilled-lora-384-1.1.safetensors inside it.
- Or open the workflow JSON in a text editor and replace the prefixed string with just ltx-2.3-22b-distilled-lora-384-1.1.safetensors. ComfyUI then resolves it directly from ComfyUI/models/loras/.
On Windows the separator is \, on macOS/Linux it is / — they refer to the same nested folder regardless of platform.
Common issues
LoRA loads but generation is still slow / many steps▼
Sampler still at dev defaults. The LoRA cannot change step count on its own. Fix: Set KSampler steps = 8 and CFG = 1. The LoRA is calibrated for exactly these settings.
OOM after adding the LoRA on a 16 GB card▼
At 7.6 GB this rank-384 LoRA plus the dev FP8 transformer, Gemma encoder, and VAE can exceed 16 GB. Fix: Use the lighter Kijai dynamic-rank LoRA (~2.7 GB) on 16 GB, or enable model offload. Reserve the rank-384 LoRA for 24 GB+.
Output overcooked when applied to a distilled file▼
Double-distillation — the file already has 'distilled' baked in, and the LoRA distills again. Fix: Apply this LoRA only on dev-base files (ltx-2.3-22b-dev.safetensors or a dev transformer variant), never on a file with 'distilled' in the name.
ComfyUI doesn't see the file after I downloaded it▼
Make sure the file is in ComfyUI/models/loras/ (not a subfolder). Restart ComfyUI fully — the menu refresh sometimes misses new files. Filename must match exactly: ltx-2.3-22b-distilled-lora-384-1.1.safetensors.
CUDA out of memory error when loading the model▼
ltx-2.3-22b-distilled-lora-384-1.1.safetensors needs ~16GB VRAM minimum. If you're hitting OOM: • Enable Sequential Offloading in ComfyUI settings • Lower the resolution (768×512 instead of 1280×704) — both dimensions must be divisible by 32 • Reduce frame count (65 frames instead of 161) — must be 8n+1 • Use a smaller variant — see Related models below.
How do I apply this LoRA in ComfyUI?▼
Load it in a 'LoraLoader' node and connect it after your model loader. Pair this LoRA with the dev base model (not the distilled one) for the right behavior. LoRA strength 1.0 is the trained value — start there.
Get notified when LTX 2.3 Distilled LoRA 384 v1.1 (Official) updates
Occasional updates on what's new in LTX 2.3 — new FP8 quants, LoRAs, IC-LoRA releases — with our hands-on verdict on whether they're worth re-downloading. No fixed cadence.
No spam. Sent occasionally when there's real news. Unsubscribe in one click.
Related models
ltx-2.3-22b-distilled-1.1_transformer_only_fp8_scaled.safetensors
ltx-2.3-22b-distilled-1.1_lora-dynamic_fro09_avg_rank_111_bf16.safetensors
ltx-2.3-22b-distilled-lora-1.1_fro90_ceil72_condsafe.safetensors
LTX-2.3-OmniNFT-RL-Lora_bf16.safetensors
ltx-2.3-22b-distilled-1.1_transformer_only_mxfp8_block32.safetensors
ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled_v3.safetensors