Skip to content

Deploy LTX 2.x on RunPod

Use this page to choose an LTX template, deploy it on a suitable NVIDIA GPU and open ComfyUI. Both templates support video and audio generation from text, images or video; the main choice is between the newer LTX-2.5 model chain and the established LTX-2.3 workflows.

Quick recommendation

  • Start with LTX-2.5 INT8 ConvRot for new text-to-video or image-to-video projects.

  • Choose LTX-2.3 BF16/FP8 for the included DWPose and SDPose video-to-video workflows or when you need the documented LTX-2.3 precision profiles.

  • Choose a template.

  • Check its hardware and storage requirements.
  • Complete the LTX-2.5 model-access step when using gated official files.
  • Open the RunPod link and configure the pod.
  • Wait until the container log reports that the pod is ready.
  • Open Connect → HTTP Services → ComfyUI (port 8188).

Choose a template

Template Included generation modes Choose it when RunPod
LTX-2.5 Dev INT8 ConvRot Text-to-video, image-to-video, video-to-video and video transition You want the newer LTX-2.5 chain with INT8 ConvRot models and GPU-based profile selection Deploy LTX-2.5 →
LTX-2.3 Dev BF16/FP8 Text-to-video, image-to-video, video-to-video and video transition You want the existing BF16/FP8 profiles or pose-transfer workflows Deploy LTX-2.3 →

After deployment, use the ComfyUI workflow browser to select an included example. LTX-2.5 provides standard text-to-video and image-to-video workflows; LTX-2.3 includes DWPose and SDPose video-to-video transfer examples.

Before you deploy

Select a GPU listing with a valid region ID. Hosts without a region ID have been unreliable in the documented RunPod tests.

LTX-2.5 provisioning targets

The following values are conservative targets based on model size and ComfyUI offloading. They are not universal benchmarked minimums.

Profile Typical GPU VRAM System RAM
INT8 ConvRot transformer and encoder RTX 3090, RTX 4090 or RTX 5090 24 GB minimum; 32 GB preferred 64 GB minimum; 80 GB recommended
INT8 ConvRot transformer with BF16 encoder L40S, RTX A6000 or RTX 6000 Ada 48 GB 80 GB minimum; 96 GB recommended
High-memory profile A100/H100 80 GB or RTX PRO 6000 96 GB 80 GB or more 96 GB minimum; 128 GB recommended

The supplied LTX-2.5 profile uses a 40 GB VRAM threshold. GPUs with more than 40 GB receive the high-VRAM model set; GPUs at or below it receive the low-VRAM model set.

Allow at least 120 GB of persistent storage at /workspace for ComfyUI, one complete model profile, custom nodes and a modest amount of output. Longer or higher-resolution videos require more space.

LTX-2.3 tested minimums

Precision Tested GPU Minimum VRAM Minimum system RAM Tested output
FP8 mixed RTX A5000 or RTX 4090 24 GB 50 GB 1280×736, 20 seconds, 24 fps
BF16 L40S or RTX 6000 Ada 45 GB 60 GB 1920×1088, 20 seconds, 24 fps

Higher resolutions, longer videos, larger batches and concurrent model loading can require more VRAM, RAM and storage. See the full LTX hardware requirements.

Optional credentials

  • Set PASSWORD if you want a fixed Code-Server password.
  • Add HF_TOKEN when downloading gated official LTX-2.5 files or private Hugging Face resources.
  • Add CIVITAI_TOKEN only when downloading resources from CivitAI.

See RunPod configuration for all supported environment variables and secret names.

LTX-2.5 model access

The template documents a public LTX-2.5 model chain as well as optional official files from the gated Lightricks/LTX-2.5 repository. To use the gated files:

  1. Sign in to Hugging Face with the account that will own the token.
  2. Read the current LTX-2.x Community License Agreement.
  3. Select Agree and Access on the model repository.
  4. Create a Hugging Face read token with access to Lightricks/LTX-2.5.
  5. Store it in a RunPod secret and reference it through HF_TOKEN.
  6. Deploy or restart the pod and confirm in the logs that the model download succeeds.

Access is linked to the Hugging Face account that accepted the terms. A token from another account does not inherit that access. Follow the LTX-2.5 access guide for secret examples, verification and manual downloads.

Deploy and connect

  1. Open the deployment link for your chosen LTX version.
  2. Choose an available GPU that matches a documented profile and has a region ID.
  3. Confirm the system RAM and persistent /workspace storage.
  4. Add any required RunPod secrets or environment variables.
  5. Click Deploy Pod.
  6. Open the pod logs and wait for the final ready message. Initial startup can take several minutes while the container and models download.
  7. Expand the pod, select Connect, and open ComfyUI under HTTP Services.

For screenshots, normal startup checkpoints and troubleshooting, follow the full RunPod pod deployment guide.

Container date tags

RunPod templates use dated container tags. Keep the tag supplied by the template unless you are deliberately testing another build. LTX-2.5 and LTX-2.3 can remain on different validated tags when a newer container has not yet been tested with both model chains.

Continue after deployment

Reference

Topic Guide
GPU, RAM and storage requirements LTX hardware requirements
Gated LTX-2.5 model access LTX-2.5 access guide
Environment variables and network services RunPod configuration
Pod startup and troubleshooting RunPod pod deployment