taeltx2_3.safetensors
LTX 2.3 VAE
VAE by Kijai. Required for all ComfyUI workflows. Place in models/vae/.
Released 2026-03 · Source: Kijai/LTX2.3_comfy (HuggingFace) — Shipped alongside Kijai's initial LTX 2.3 ComfyUI port in March 2026.
Download taeltx2_3.safetensors
Direct HuggingFace download. ~0.5 GB · Free.
No 2GB GPU? Try taeltx2_3.safetensors online — free generation included
Skip the ~0.5 GB download and ComfyUI setup. Generate a 6-second video using this exact model in your browser, ~30 seconds.
Technical details
taeltx2_3.safetensors is the Tiny AutoEncoder (TAE) for LTX 2.3 — a distilled, much smaller replacement for the original full VAE. The 'tae' prefix marks it as a tiny encoder/decoder pair shipped specifically for ComfyUI's pipeline; the file is around half a gigabyte versus several gigabytes for a full VAE.
Every LTX 2.3 ComfyUI workflow — T2V, I2V, two-stage, IC-LoRA — references this exact filename in its LoadVAE / VAELoader node. The string is hardcoded in published workflow JSONs, so the filename matters: do not rename, do not substitute, and do not move it out of models/vae/.
The TAE handles both the encoder (image → latents for I2V) and the decoder (latents → frames at the end of inference). Quality-wise it is calibrated against the full VAE so artifacts are minimal, but if you see banding or color drift on long clips, the standalone BF16 video VAE is the heavier alternative.
When to choose taeltx2_3.safetensors
This is not an alternative — it is required. Every workflow on this site, every ComfyUI workflow shipped by Lightricks or the community, and every IC-LoRA pipeline loads taeltx2_3.safetensors. Download it first regardless of which transformer/quant you pick.
The one case where you would use something else is if a workflow explicitly references LTX23_video_vae_bf16.safetensors — that is the BF16 standalone video VAE, used by a small number of two-stage or HDR workflows that need the heavier decoder.
Will this run on my GPU?
Minimum: 2GB VRAM.
Recommendation: Required for all setups. Download this first regardless of your VRAM.
How to use taeltx2_3.safetensors
- Download the file from HuggingFace.
- Place it in ComfyUI/models/vae/ inside your ComfyUI directory.
- Restart ComfyUI (or refresh the model list from the menu).
- Load a compatible workflow — see below.
Don't want to run this locally? Try taeltx2_3.safetensors online with a free generation — no GPU, no install, ~30 seconds per clip.
ComfyUI says it can't find taeltx2_3.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:
- ltx/taeltx2_3.safetensors
- ltx23\taeltx2_3.safetensors
- ltx 2.3\taeltx2_3.safetensors
- vae/taeltx2_3.safetensors
The prefix before the slash or backslash is a subdirectory the workflow author used. The actual file is the same taeltx2_3.safetensors — you have two fixes:
- Create the matching subdirectory inside ComfyUI/models/vae/ and place the file there. Example: if the workflow references ltx/taeltx2_3.safetensors, create the corresponding subfolder under ComfyUI/models/vae/ and put taeltx2_3.safetensors inside it.
- Or open the workflow JSON in a text editor and replace the prefixed string with just taeltx2_3.safetensors. ComfyUI then resolves it directly from ComfyUI/models/vae/.
On Windows the separator is \, on macOS/Linux it is / — they refer to the same nested folder regardless of platform.
Common issues
VAELoader: file not found / 'taeltx2_3.safetensors not in list'▼
File placed in ComfyUI/models/ root, or in checkpoints/ instead of vae/. Fix: Move the file to ComfyUI/models/vae/taeltx2_3.safetensors. Click the refresh button on the VAELoader node so ComfyUI re-scans the directory.
VAE decoded image is solid green / black / corrupt▼
Partial download — HF's xet protocol can leave a truncated file on flaky connections that still looks like a valid .safetensors header. Fix: Delete the file and redownload. If you used a browser, switch to `huggingface-cli download` or `aria2c` with retry. Verify by checking file size matches HuggingFace's reported size (~500 MB).
Workflow loads but VAE node is red after rename▼
The filename is hardcoded in workflow JSONs as taeltx2_3.safetensors. Fix: Rename it back. If you have other VAE files in the directory, keep taeltx2_3.safetensors and let the others coexist.
ComfyUI doesn't see the file after I downloaded it▼
Make sure the file is in ComfyUI/models/vae/ (not a subfolder). Restart ComfyUI fully — the menu refresh sometimes misses new files. Filename must match exactly: taeltx2_3.safetensors.
CUDA out of memory error when loading the model▼
taeltx2_3.safetensors needs ~2GB 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 VAE 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_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-lora-384-1.1.safetensors
ltx-2.3-22b-distilled-lora-384.safetensors
ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors