LTX-2.5 VRAM requirements in ComfyUI, file by file

LTX-2.5 is Lightricks' open video model with sound. ComfyUI's own workflow for Distilled downloads 44.9 GB and needs 24.3 GB on the card at its largest stage, the diffusion stage. The smallest card listed here that loads that set with room to spare is the RTX 5090.

What LTX-2.5 is

Lightricks' open video model from August 2026, a 22-billion-parameter transformer that generates sound with the picture. It comes distilled, for few-step generation, and as dev, the undistilled model. LTX-2.3, the release before it, is still in wide use and shares the same line.

Distilled: 22B · Dev: 22B Text encoder: Gemma 4 12B LTX-2 Community License August 2026

Which LTX-2.5 file for your card

For each memory size, the highest-precision diffusion file that loads with room to spare (a file that only just fits is used when nothing else does), and what is left for working memory while it samples. "Streams" means even the smallest file is larger than the card: ComfyUI still runs it by streaming weights from system RAM, more slowly. FP4 files are offered only to the Blackwell cards that run them natively.

Distilled

ComfyUI's template for Distilled: INT8 ConvRot diffusion model, INT8 ConvRot text encoder. Stages: text encoder 15.4 GB, prompt enhancer 5.2 GB, diffusion 24.3 GB. Download 44.9 GB.

Dev

ComfyUI's template for Dev: BF16 diffusion model, INT8 ConvRot text encoder. Stages: text encoder 15.4 GB, prompt enhancer 5.2 GB, diffusion 44.9 GB. Download 65.4 GB.

Check another card, or pick a file by hand, in the ComfyUI VRAM checker.

Every LTX-2.5 file ComfyUI loads

Sizes are the exact byte counts Hugging Face reports. Files marked "template" are the ones ComfyUI's official workflow downloads.

What Lightricks says it needs

• "GPU: NVIDIA GPU with a minimum 32GB+ VRAM - more is better" (Lightricks' own pipeline, minimum, source)
• "GPU: NVIDIA A100 (80GB) or H100" (Lightricks' own pipeline, recommended, source)

Those figures and this page measure different things. A vendor's figure covers its own script at its default resolution and length, working memory included. The tables here count the weights each stage holds, which is exact, and leave the working memory as the space that remains, because it depends on your resolution, frame count and attention kernel and nobody publishes it per model. If generation runs out of memory with a file that loads, lower the resolution or the frame count before stepping down a precision. ComfyUI's README says it "can run even the biggest open source models on as low as 4GB vram + 8GB ram" by streaming weights (source), so "streams" means slower, not impossible.

LoRA training for LTX-2.5

Only configurations a trainer's own documentation gives a VRAM figure for. Settings change the figure a great deal, so read the setting column before trusting the number.

What LTX-2.5 is good at, and what to watch

Good at

• Video with audio, 24 frames a second, about five seconds at the default length.
• Lightricks ships ComfyUI-ready files itself, including spatial and temporal upscalers.
• Lightricks documents a low-VRAM LoRA training config for 32 GB cards.

Watch out for

• Lightricks gives a 32 GB minimum for its own pipeline and recommends an A100 or H100.
• The text encoder is Gemma 4 12B, and ComfyUI's template also loads a small Gemma 4 model to enhance prompts.
• Free for companies under US$10 million in annual revenue; above that, a paid agreement. The repo is gated: accept the terms on Hugging Face first.

Questions

How much VRAM does LTX-2.5 need?

In ComfyUI's own workflow for Distilled, the largest stage is the diffusion stage at 24.3 GB, and the whole workflow downloads 44.9 GB. Working memory for generation comes on top and grows with resolution and length. Lightricks itself says: "GPU: NVIDIA GPU with a minimum 32GB+ VRAM - more is better" (Lightricks' own pipeline, minimum).

Can I run LTX-2.5 on 8 GB of VRAM?

Not fully. The smallest file here, INT8 ConvRot at 21.5 GB, needs 24.3 GB at its largest stage, more than 8 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.

Can I run LTX-2.5 on 12 GB of VRAM?

Not fully. The smallest file here, INT8 ConvRot at 21.5 GB, needs 24.3 GB at its largest stage, more than 12 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.

Can I run LTX-2.5 on 16 GB of VRAM?

Not fully. The smallest file here, INT8 ConvRot at 21.5 GB, needs 24.3 GB at its largest stage, more than 16 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.

Which LTX-2.5 file should I download?

The highest precision that loads fully on your card. On a 24 GB card that is none of them: even INT8 ConvRot streams. ComfyUI's template downloads INT8 ConvRot with the INT8 ConvRot text encoder. FP4 files (NVFP4) are made for Blackwell cards (RTX 50, RTX PRO 6000, DGX Spark); on older cards, use FP8 or INT8.

• Files and sizes: Lightricks/LTX-2.5 on Hugging Face, read from the Hugging Face API. Template files: ComfyUI's LTX-2.5 guide.
• Model and licence: Lightricks/LTX-2.5 and the LTX-2 Community License.
• Card memory: manufacturers' specifications, as on each GPU page. The checker covers NVIDIA cards; ComfyUI also runs on AMD and Apple silicon, where file formats and speed differ and we have not sized them.
• Fit rule: a stage loads fully under 85% of usable memory and is tight up to 95%, the same margins the LLM pages use.