> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.ltx.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.ltx.io/_mcp/server.

# VFX & Post-Production

> Choose an LTX VFX and post-production workflow for HDR, refinement, and high-resolution delivery.

LTX for Post-Production brings together four capabilities for color, detail, and resolution. They are designed to produce elements and images that can keep moving through a production pipeline instead of ending as flattened previews.

Use this section to choose the right workflow, then follow the capability guide for its exact inputs, outputs, version boundaries, and limitations. The guides assume that you are comfortable with frame sequences, color management, and node-based or Python production workflows.

This section covers open-source ComfyUI workflows and the native Python paths identified in individual guides.

## Choose a workflow

Choose the guide that matches your source material and production goal. Unless a card says otherwise, these guides focus on LTX-2.5 ComfyUI workflows.

#### [Native HDR](/open-source-model/vfx-post-production/working-with-exr-hdr)

Animate an EXR still or replace part of an EXR sequence while keeping HDR data available for compositing and grading.

**Available in:** LTX-2.5 ComfyUI; image-to-video also has a native Python path.

**Watch for:** The inpainting graph uses one specific LTX-2.3 IC-LoRA as a documented exception. Keep the rest of the model stack on LTX-2.5.

#### [SDR to HDR](/open-source-model/vfx-post-production/sdr-to-hdr)

Expand supported SDR video into gradeable HDR output with an EXR sequence and a BT.2020/HLG master.

**Available in:** LTX-2.5 ComfyUI and native Python, plus a separate legacy LTX-2.3 workflow.

**Watch for:** ACEScct is the LTX-2.5 working representation. The current ComfyUI graph and Python pipeline default to scene-linear ACEScg EXR. Keep LTX-2.3 LogC3 assets separate.

#### [Native Resolution](/open-source-model/vfx-post-production/native-resolution)

Use Tiled Fusion to process an IC-LoRA video task on a canvas larger than the adapter's trained spatial window.

**Available in:** LTX-2.5 ComfyUI with a workflow-matched IC-LoRA.

**Watch for:** Match the tile size to the adapter's trained window. Full HD, 4K, and 8K are selector values, not performance guarantees; use only workflow-matched IC-LoRAs.

#### [Refine & Restore](/open-source-model/vfx-post-production/refine-and-restore)

Add synthesized clarity and detail to archival or low-quality footage, with Tiled Fusion available for larger output canvases.

**Available in:** LTX-2.5 ComfyUI.

**Watch for:** The documented Tiled Fusion Upscale workflow uses Refine Details; Restore is a separate adapter and is not selected by that graph. This is generative refinement, not pixel-accurate restoration or a conventional upscaler. Test the selected output size on your target hardware.

## Prepare for production

### Frame sequences and timing

EXR sequences do not carry container timing. Supply the correct frame rate and confirm sequence ordering and frame count before generation. LTX video frame counts use the workflow's supported `8k+1` structure; a loader may trim an EXR folder to its longest valid prefix.

### Color management

Declare the color space that matches the pixels in an EXR source. Do not put display-encoded HLG or PQ values in an EXR file and label them as scene-linear. Match the output transform and viewer input transform to the generated EXR color space.

### Preview and master outputs

Treat a tonemapped preview as a viewing aid. Use the documented EXR sequence or HLG video as the production output, and validate it in the color-managed application used by your pipeline.

### Storage and quality checks

Frame sequences and high-resolution intermediates require substantial storage. Budget for source plates, generated EXRs, previews, and delivery masters. Validate representative frames, boundaries, motion, audio, and color transforms before committing a full batch.