LTX 2.5: what it is, the licence, hardware needs and cost

LTX 2.5 is LTX’s open-weights video-and-audio model, released on August 11, 2026, free under $10M annual revenue, needing a 32GB+ GPU to run locally or from $0.09 a second on its API; this site currently delivers images only, not video, and its photo editor, a different model, edits one photo in about 37 seconds for 10 credits. Free with sign-up: 10 credits = 1 image.

Your edited photo appears here (this site’s editor)
Studio background edit made on this site: the same woman on a plain light-grey studio backdrop
Input photo (AI-generated, fictional): a smiling woman with curly brown hair in a green sweater, in a parkYour photo
Type your own change, or leave this empty and pick a one-tap edit below.0/400

Style changes (cartoon, 3D, anime) are not reliable yet. If the result looks unchanged, try again or describe the change more specifically.

One-tap edits

  • Images only: this site does not make video
  • Not affiliated with LTX or Lightricks
  • Free with sign-up: 10 credits = 1 image
release (LTX press release)
Aug 11, 2026
parameters, read from the checkpoint file names
22B
annual revenue below which commercial use is free
$10M
GPU memory in LTX’s official minimum
32 GB+

What LTX 2.5 is

From LTX’s own pages

  • The Hugging Face page calls LTX-2.5 an open-weights video-and-audio generation model for local execution and fine-tuning. It makes video with synchronized sound from text, image or video input, and lists audio-to-video and text-to-audio as modes too. Native multishot generation is listed as new in 2.5.Source: Hugging Face model page ·
  • Released on August 11, 2026: LTX’s press release of that date says it is “available today as open weights on Hugging Face, natively in ComfyUI, and through the LTX API”. The model page itself states no date.Source: LTX press release ·
  • The page does not state a parameter count; the checkpoint file names say 22b, so we read it as a 22B-parameter model. The text encoder is a custom Gemma 4 12B. Variants include a distilled model (fixed 8 steps) and a full dev model you can fine-tune.Source: Hugging Face model page ·
  • LTX says what is new: a new diffusion video decoder, native multishot (character, scene and voice held across cuts), a Gemma 4 language backbone with a prompt enhancer, and a better distilled model. These are the maker’s claims; we did not test them.Source: LTX press release ·
  • Who makes it: LTX announced on July 8, 2026 that it is spinning off from Lightricks into an independent company. The weights and code still sit under the Lightricks name on Hugging Face and GitHub.Source: LTX newsroom ·
  • Our own read of Hugging Face’s public API on 2026-11-19: the repo was created on 2026-07-23, 19 days before the announcement, last changed on 2026-10-11, and shows 2,871,190 downloads and 12,586 likes by Hugging Face’s own counters. Access is gated: you agree to share your Hugging Face details first.Source: Hugging Face API, our read ·

The licence

  • The LTX-2.x Community License Agreement: commercial and production use at no cost for organizations under $10M in annual revenue; above that, a paid commercial agreement. Revenue counts the whole entity, including subsidiaries and affiliates. Passing on fine-tunes may need a paid license. The binding terms are in the licence file on GitHub, which we did not read.Source: Hugging Face model page ·
  • LTX’s licensing page says the threshold is total annual company revenue across the organization, and that under $10M you can self-host the weights, fine-tune and use LTX commercially at no cost, with no usage caps.Source: LTX licensing page ·
  • A wording difference we noticed: the press release says free for organizations under $10M in annual recurring revenue, while the model page and the licensing page say annual revenue. They are not the same thing, so read the licence text, not a summary. Because of the revenue limit, this is open weights under a community licence, not open source in the strict sense.Source: LTX press release vs licensing page ·

Hardware, and where the numbers disagree

  • LTX’s official minimum to run it locally: an NVIDIA GPU with 32GB+ of VRAM, 32GB of system memory, 100GB of free storage, CUDA 12.7 or higher and Python 3.12 or higher. Recommended: an A100 (80GB) or H100, 64GB+ of memory and a 200GB+ SSD.Source: LTX docs, system requirements ·
  • For less memory the model page lists an fp8 option (`--quantization fp8-cast` with CPU offload) and, for Diffusers, CPU offload and VAE tiling. It also lists int8 and NVFP4 builds for ComfyUI or Blackwell GPUs. No VRAM figure is given for these on the page.Source: Hugging Face model page ·
  • LTX’s press release says it runs on hardware you already own, “whether a data center GPU to a Mac”. It gives no specification for that, so we treat it as a claim.Source: LTX press release ·

Resolution and length

  • The hosted API models: LTX-2.5 Fast and LTX-2.5 Pro, up to 4K (3840×2160) in 16:9, 9:16, 1:1, 3:2 and 2:3. Fast makes 6 to 20 seconds at 720p or 1080p and 24 or 25 fps (6, 8 or 10 seconds at 48 or 50 fps, 1440p and 4K). Pro makes 6, 8 or 10 seconds at every size. The model can also pick the length itself.Source: LTX docs, LTX-2.5 model page ·
  • For the open weights the model page gives rules, not a limit: width and height divisible by 32, frame count of the form 8n+1, an example at 1920×1088, and a 4K example (3840×2176) using 2× upscaling. Its Diffusers example is 121 frames at 24 fps, which is about 5 seconds by our division. It states no maximum length.Source: Hugging Face model page ·
  • Vendor pages differ slightly: ComfyUI’s launch post says Fast runs 720p through 4K at 2 to 20 seconds, while LTX’s docs table limits 1440p and 4K to 10 seconds. We quote the docs.Source: ComfyUI blog ·

Not confirmed

  • Quality compared with other video models: LTX publishes its own comparison charts, which we did not read, and we found no independent test. We do not rank it.Source: LTX press release ·
  • Running it on less than 32GB: third-party articles cite lower figures for community quantised builds. We did not test any, so we state none.Source: LTX docs, system requirements ·
  • Other hosts and their prices (for example fal.ai, Replicate or kie.ai): we did not check, so the only price on this page is LTX’s own API. Also unchecked: the licence file text, and why the word is searched more now.Source: LTX docs, pricing ·

On this site: related pages

Prices

LTX 2.5 API prices, and this site

LTX’s API price per second of output video, read on 2026-11-19 from LTX’s pricing page; LTX sets these. The 10-second row is our own arithmetic. The open weights cost nothing to download but need your own hardware. This site makes no video; its numbers come from its own settings.
-LTX-2.5 Fast (API)LTX-2.5 Pro (API)This site (photo editor)
What it makesVideo, with soundVideo, with soundAn edit of one photo you upload (1024 × 1024)
720p$0.09 per second$0.12 per secondVideo not offered
1080p$0.13 per second$0.17 per secondVideo not offered
1440p$0.19 per second$0.25 per secondVideo not offered
4K$0.30 per second$0.39 per secondVideo not offered
Our arithmetic: one 10-second 1080p clip$1.30$1.70Not video
One edited photoNot what it doesNot what it does10 credits, from $0.13

How to run it

How to run LTX 2.5

  1. In ComfyUI

    ComfyUI added day-one support. Update to a version that has the LTX-2.5 templates (the launch post named 0.32.0) or use Comfy Cloud, put the weights from Hugging Face in your models folder, load the template and add your prompt.

  2. In Python: ltx-pipelines

    Clone the Lightricks LTX-2 repo (Python 3.12 or higher, CUDA 12.7 or higher), accept the licence on the Hugging Face page, download the split files with the Hugging Face CLI and run the distilled pipeline. Use the bf16 files there; the int8 files are for ComfyUI only.

  3. With Diffusers

    The model page lists a Diffusers version, Lightricks/LTX-2.5-Diffusers, and says it needs Diffusers installed from the main branch because LTX-2.5 was not yet in a release.

  4. Hosted: the LTX API

    LTX’s API bills per second of output video, in Fast and Pro tiers, from $0.09 per second at 720p. The model page says no inference provider on Hugging Face serves it.

Prompts

Photo edit prompts for this site’s editor

These are this site’s photo-editor prompts, for preparing a still photo before you give it to a video model as its first frame. They are not LTX prompts. We did not run them for this page, and we have not tested them in other tools. Results vary with the photo.

Prepare a clean first-frame photo

LongCat Video: another open-weights video model

Not run for this page. LTX-2.5 can start from an image, but this site does not make the video.

  • Remove background clutter

    Edit my uploaded photo: remove distracting clutter from the background (stray objects, litter, bins, cables, signs and people in the background) and fill those areas naturally with what is around them. Keep the framing as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Plain studio background

    Edit my uploaded photo: replace the background with a plain, softly lit light-grey studio backdrop with natural soft shadows. Keep the framing as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Professional headshot

    Edit my uploaded photo into a clean professional headshot: a smooth neutral grey-blue background with a gentle gradient, soft even studio light on the face, the same crop or a slightly tighter head-and-shoulders crop. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

Not run for this page. These edit one photo and return a still image.

  • Fix the lighting

    Edit my uploaded photo: fix the lighting only. Balance the exposure, lift dark shadows on faces, tame blown-out highlights and correct the white balance so skin tones look natural. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Warm golden-hour light

    Edit my uploaded photo: change only the light to warm, low golden-hour sunlight from one side, with soft long shadows and a warm glow, as if taken shortly before sunset. Keep the scene and framing. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Classic black and white

    Edit my uploaded photo into a classic black and white photograph with rich contrast, deep blacks, clean whites and fine film grain. Keep the composition exactly as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Watercolor painting

    Repaint my uploaded photo as a soft watercolor painting on textured white paper, with loose washes and visible brush edges, keeping the same composition. Keep every person and pet recognizable, and do not add or remove anyone. No text, logos or watermarks.

  • Sharpen and clean up

    Edit my uploaded photo: gently sharpen it, reduce noise and compression artifacts, and recover fine detail in hair, fabric and edges, without a plastic or over-smoothed look. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

Pricing on this site

Credit packs or a monthly plan.

Credits come in one-time packs ($4.99 for 50 credits, $9.99 for 120 credits, $19.99 for 300 credits); each image costs 10 credits. Or pick a plan from $9.90 a month (500 credits), cancel anytime.

  • Basic · $9.90/month500 credits a month · about 50 images · $0.20 per image
  • Pro · $19.90/month1500 credits a month · about 150 images · $0.13 per image

Billed every month and renews automatically. Cancel anytime in Settings; you keep the plan until the end of the month you paid for. Each month's credits are for that month; unused ones expire.

Refund policy

FAQ

Questions, answered

What is LTX 2.5?

LTX-2.5 is an open-weights model from LTX that generates video and synchronized audio from text, images or video. It was released on August 11, 2026, has about 22 billion parameters by its file names, and runs locally, in ComfyUI or through LTX’s API.

What is the LTX 2.5 licence, and is LTX 2.5 free to use?

It uses the LTX-2.x Community License. Commercial use is free for organizations under $10M in annual revenue, and a paid agreement is needed above that. LTX’s press release says “annual recurring revenue” while its licence page says “annual revenue”, so read the licence text.

When was LTX 2.5 released (LTX-2.5 release date)?

On August 11, 2026, according to LTX’s press release, with open weights on Hugging Face, day-one ComfyUI support and the LTX API. The Hugging Face repo itself was created earlier, on 2026-07-23, by our read of its API.

How much VRAM does LTX 2.5 need?

LTX’s official minimum is an NVIDIA GPU with 32GB+ of VRAM, 32GB of system memory and 100GB of storage; it recommends an A100 (80GB) or H100. Third parties report lower figures for community quantised builds; we did not test any.

How long and how sharp are LTX 2.5 videos?

Through the API: up to 4K, and up to 20 seconds for Fast at 720p or 1080p (24 or 25 fps), 10 seconds at most for Pro and for Fast at 1440p and 4K. The open weights state no maximum length. All of this is the maker’s statement; we did not test it.

How do I run LTX 2.5 in ComfyUI or from Hugging Face?

In ComfyUI, update to a version with the LTX-2.5 templates, add the weights and load the template. From Hugging Face, accept the licence on the Lightricks/LTX-2.5 page, then use ltx-pipelines from the Lightricks LTX-2 repo or the Diffusers version.

How much does the LTX 2.5 API cost?

LTX bills per second of output video: Fast is $0.09 at 720p, $0.13 at 1080p, $0.19 at 1440p and $0.30 at 4K; Pro is $0.12, $0.17, $0.25 and $0.39. By our arithmetic a 10-second 1080p clip is $1.30 on Fast. We did not check other hosts.

Can I make LTX 2.5 video on this site?

No. This site is not LTX or Lightricks and currently delivers images only, not video. Its editor uses a different model and edits one photo in about 37 seconds. Partly: new accounts get 10 free credits after signing in, enough for 1 free image (once per account, valid for 7 days, while the day's new-account gifts last). Each image costs 10 credits, from $0.13 on a monthly plan. You can type your own request (up to 400 characters) or tap a one-tap edit.

Just need one photo edited?

Upload it, write the change, and download the result in about 37 seconds. Images only, not video.

Edit a photo here

Independent site. Not affiliated with LTX, Lightricks, Hugging Face or ComfyUI, and this site does not run LTX 2.5 or make video. LTX 2.5 facts come from the makers’ own pages, read on 2026-11-19; licence terms, limits and prices may have changed.