True HDR output for AI video, built on LTX-2.3
LTX HDR converts 8-bit SDR — whether AI-generated or existing footage — into scene-linear 16-bit HDR EXR (float16, ~20-bit effective range), recovering the highlight and shadow detail SDR throws away. Your footage grades, relights, and composites alongside rendered or filmed plates, with production-grade dynamic range from one model.
.svg)
Key Capabilities
16-bit HDR EXR output
LTX HDR, an IC-LoRA on LTX-2.3, converts 8-bit SDR into scene-linear 16-bit HDR EXR (float16, ~20-bit effective range) — gradeable like rendered or raw footage.
SDR-to-HDR for existing footage
Convert any existing SDR clip to true HDR. The model reconstructs clipped highlights and crushed shadows instead of stretching values — no reshooting, no manual grading.
Consistent enhancement via IC-LoRA
Train a custom IC-LoRA on reference pairs to apply the same treatment across a full library — consistent, repeatable results, deployed via API for batch work.
Grade and composite AI footage like real plates
Post-production pipelines that need AI-generated video to hold up through a full color grade. LTX HDR outputs scene-linear 16-bit EXR, giving colorists and compositors the latitude to grade, relight, and composite AI footage alongside rendered or filmed elements. No more flat, locked 8-bit image with tonemapping baked in.
Upgrade existing SDR libraries to HDR
Teams managing large libraries of existing SDR video. SDR-to-HDR reconstructs the dynamic range lost in the original 8-bit format — recovering clipped highlights and crushed shadows rather than just stretching values — and outputs a scene-linear EXR ready for grading and finishing. No reshooting, no manual workflows.
.jpeg)
Proven in professional production
LTX HDR was developed with Gear Productions, Magnopus, and Asteria, and is already in live production — from broadcast deliverables to LED-wall virtual production to the Tribeca-selected short film ChikaBOOM!. AI assets now carry the dynamic range needed to sit alongside traditionally captured elements.
Consistent batch treatment via IC-LoRA
Studios running large clip volumes through the same treatment: train an IC-LoRA on reference pairs and apply consistent, repeatable enhancement across every clip in the batch. Train once, deploy via API — no manual per-clip work.
Designed for real-world deployment
Production-grade HDR output for post-production, VFX, and virtual-production teams integrating AI video into professional finishing pipelines.
Product teams, AI startups, and developers building AI-powered video features. Add production-grade video generation as a product capability, not a research project. One API, production-ready results, and no custom orchestration.
Brands, agencies, and creative teams producing high volumes of content. Turn existing assets into video at scale. Faster iteration, lower production cost, and more output from what you already have.
Teams that require full control over deployment and data. Run video generation in your own environment. On-premises, no cloud dependency, and full infrastructure ownership.
Platforms powering creative tools with multiple AI models. Upgrade your video output with a best-in-class engine. Improve generation quality, retain users, and differentiate with a model built for production, not prototypes.
How LTX HDR works
Input
Technical characteristics:
Input: AI-generated video, or an existing SDR video file (for the SDR-to-HDR path).
- LTX HDR path: the LTX HDR IC-LoRA checkpoint on LTX-2.3.
- Runs via LTX API (REST) or open-source on a ≥32GB-VRAM GPU; FP8 quantization lowers the memory footprint.
Output
Technical characteristics:
- Scene-linear 16-bit HDR EXR (float16, ~20-bit effective range), gradeable like rendered or raw footage and compositable alongside traditionally created assets.
- Reconstructed highlight and shadow detail — not a tonemap stretch.
- Side-by-side SDR/HDR frame review in LTX Studio.
- IC-LoRA path: consistent enhancement applied across every clip.
- Per-second API pricing — pay only for the output you generate.
Beta - HDR Video Generation
Convert SDR video to 16-bit HDR for greater dynamic range and post-production flexibility — built for professional grading and finishing workflows.
- Video-to-video only (SDR → HDR)
- Output delivered as per-frame 16-bit EXR (ZIP)
- Billed per second of input video
- Max duration depends on resolution tier (up to ~7s at 1080p, up to ~4s at 2560×1440p)
FAQs
What resolutions does LTX-2.3 upscaling support?
Up to 4K. The spatial upsampler supports x1.5 or x2 scale and is available in TI2VidTwoStagesPipeline and TI2VidTwoStagesHQPipeline. RetakePipeline and TI2VidOneStagePipeline do not include the spatial upsampler.
Can upscaling be applied to existing footage, not just LTX-generated video?
Both. The spatial upsampler can be applied to externally sourced footage or as Stage 2 of a full text/image-to-video generation.
Is the upscaling model open-source?
Yes. Spatial upscaler weights are on HuggingFace. Run locally on 80GB+ VRAM GPU (32GB with FP8 quantization) or call via the LTX API.
