← Back to models

Lightricks

LTX-2.3

Active

LTX-2.3 is Lightricks’ open-weight audio-video model family, available for local workflows and through Fast and Pro API variants.

Capabilities

Text-to-videoYesImage-to-videoYesVideo-to-videoVaries by variantVideo editingVaries by variant

Local deployment evidence

The official LTX open-source overview provides the local installation and inference workflow and links the downloadable model resources.

Hardware entry summary

Lowest documented entry VRAM
16 GB
Applies to
LTX Desktop

The 16 GB figure is the lowest documented entry point and applies only to the LTX Desktop Quick Start profile. It is not a family-wide requirement for every checkpoint or workflow.

View full hardware profiles ↓

Local checkpoints

LTX-2.3 22B Dev

Variant ID: ltx-2.3-22b-dev

Active
Text-to-videoYesImage-to-videoYesVideo-to-videoYesVideo editingUnknown
Access mode
Local checkpoint
Checkpoint type
Full
Precision
BF16
Maximum duration
Unknown
Maximum resolution
Unknown
Frame rates
Not specified
Aspect ratios
Not specified
Native audio
Yes
Reference image
Yes
Pricing
No provider inference fee

Notes

Official full BF16 development checkpoint for flexible local inference and training. There is no provider inference fee for local execution, but hardware, electricity, storage, and processing time still cost money. The model license can impose commercial licensing conditions, so this does not mean unrestricted free commercial use.

LTX-2.3 22B Distilled 1.1

Variant ID: ltx-2.3-22b-distilled-1.1

Active
Text-to-videoYesImage-to-videoYesVideo-to-videoYesVideo editingUnknown
Access mode
Local checkpoint
Checkpoint type
Distilled
Precision
Not specified
Maximum duration
Unknown
Maximum resolution
Unknown
Frame rates
Not specified
Aspect ratios
Not specified
Native audio
Yes
Reference image
Yes
Pricing
No provider inference fee

Notes

Official distilled 1.1 checkpoint documented for an eight-step workflow with CFG 1. There is no provider inference fee for local execution, but hardware, electricity, storage, and processing time remain user costs. The model license can impose commercial licensing conditions; this is not unrestricted free commercial use.

Hardware profiles

Quick Start — LTX Desktop local generation

Applies to
LTX Desktop
Official minimum VRAM
16 GB
Official recommended VRAM
Unknown
Official minimum system RAM
16 GB
Official recommended system RAM
32 GB
Storage
160 GB

Quick Start requirements for LTX Desktop local generation on Windows or Linux with an NVIDIA GPU. macOS uses the API-only path. LTX Desktop is beta, and 16 GB VRAM is a workflow-specific entry point rather than a guarantee for every checkpoint, pipeline, resolution, or duration.

Official documentation (External link)

Full open-source workflow

Applies to
LTX-2.3 22B Dev, LTX-2.3 22B Distilled 1.1
Official minimum VRAM
32 GB
Official recommended VRAM
80 GB
Official minimum system RAM
32 GB
Official recommended system RAM
64 GB
Storage
100 GB

Official full open-source requirements: at least 32 GB GPU VRAM, 32 GB RAM, and 100 GB storage. The recommended environment is an A100 80 GB or H100, at least 64 GB RAM, and 200 GB or more SSD storage. This is a different workflow context from the 16 GB LTX Desktop Quick Start.

Official documentation (External link)

LTX Desktop product-page guidance

Applies to
LTX Desktop
Official minimum VRAM
Unknown
Official recommended VRAM
Unknown
Official minimum system RAM
Unknown
Official recommended system RAM
Unknown
Storage
Unknown

The official product page is internally inconsistent and is preserved as guidance rather than a single minimum: its Windows installation section recommends 32 GB+ VRAM and lists 160 GB free disk, while its FAQ names an RTX 5090 with 32 GB VRAM, 32 GB RAM, and 60 GB+ free disk. These figures are not merged with the separate 16 GB Quick Start profile.

Official documentation (External link)

Overview

Provider
Lightricks
Status
Active
Publication status
Published
Record scope
Family
Release date
2026-02-23
Last verified
2026-08-06

Specifications

Family maximum duration
Varies by variant
Family maximum resolution
Varies by variant
Frame rates
Varies by variant
Aspect ratios
Varies by variant
Portrait support
Yes
Native audio
Yes
Reference image
Yes
Character consistency
Unknown

Hardware

Local deployment
Yes
Tested minimum VRAM
Unknown
ComfyUI support
Yes

Licensing

Open weights
Yes
Open source
Unknown
Commercial use
Conditional
License
LTX-2 Community License Agreement

Local applications

LTX Desktop

Variant ID: ltx-desktop

Preview
Text-to-videoYesImage-to-videoYesVideo-to-videoUnknownVideo editingUnknown
Access mode
Local application
Checkpoint type
Not applicable
Precision
Not specified
Maximum duration
Unknown
Maximum resolution
Unknown
Frame rates
Not specified
Aspect ratios
Not specified
Native audio
Yes
Reference image
Yes
Pricing
No provider inference fee

Notes

Official beta desktop application for local generation on supported Windows systems with NVIDIA GPUs. There is no provider inference fee in local mode, but hardware, electricity, storage, and processing time still cost money. The underlying model license can impose commercial conditions, so this is not unrestricted free commercial use.

Availability

API availability
Yes
Web app availability
Yes

Performance and pricing

Speed category
Unknown
Estimated clip cost
Unknown
Pricing
Varies by variant

Hosted API options

LTX-2.3 Fast

Variant ID: ltx-2-3-fast

Active
Text-to-videoYesImage-to-videoYesVideo-to-videoNoVideo editingNo
Access mode
Hosted API
Checkpoint type
Not applicable
Precision
Not specified
Maximum duration
20 seconds
Maximum resolution
4K
Frame rates
24 fps, 25 fps, 48 fps, 50 fps
Aspect ratios
16:9, 9:16
Native audio
Yes
Reference image
Yes
Pricing
Custom pricing

Notes

The 20-second maximum applies only to 1080p at 24 or 25 fps. At 48 or 50 fps, and at 1440p or 4K, the maximum is 10 seconds. Audio-to-video, retake, extend, and reframe require Pro. Text/image generation costs USD 0.06/s at 1080p, USD 0.12/s at 1440p, and USD 0.24/s at 4K.

LTX-2.3 Pro

Variant ID: ltx-2-3-pro

Active
Text-to-videoYesImage-to-videoYesVideo-to-videoYesVideo editingYes
Access mode
Hosted API
Checkpoint type
Not applicable
Precision
Not specified
Maximum duration
10 seconds
Maximum resolution
4K
Frame rates
24 fps, 25 fps, 48 fps, 50 fps
Aspect ratios
16:9, 9:16
Native audio
Yes
Reference image
Yes
Pricing
Custom pricing

Notes

Pro supports audio-to-video, retake, extend, HDR upscale, and video reframe endpoints. Its documented generation maximum is 10 seconds. Text/image generation costs USD 0.08/s at 1080p, USD 0.16/s at 1440p, and USD 0.32/s at 4K; Pro-only endpoints use separate rates.

Best use cases

  • Local or private text-, image-, and video-conditioned generation workflows
  • Hosted generation and Pro-only audio-to-video or video-editing workflows

Limitations

  • Fast and Pro differ in duration, endpoint support, and price
  • Official hardware guidance differs by local workflow and even between sections of the Desktop product page
  • The community license applies revenue-based conditions to commercial use

Hardware notes

Official requirements describe separate workflow contexts. The 16 GB entry point belongs only to the LTX Desktop Quick Start; the full open-source workflow begins at 32 GB VRAM; and the Desktop product page contains additional, internally differing 32 GB/60 GB/160 GB guidance. Review the complete profiles rather than applying one value to every variant. These are official statements, not Aenica test results.

Commercial use notes

Conditional: the official license permits commercial use for entities below USD 10 million in annual revenue; entities at or above that threshold need a paid commercial license. All users remain subject to the complete license terms.

Speed notes

Speed remains unclassified because this catalog has not run a standardized reproducible benchmark.

Pricing notes

Local checkpoints and local Desktop mode have no provider inference fee, but hardware, electricity, storage, and processing time still cost money. Model licensing can also impose commercial conditions, so local access is not unrestricted free commercial use. Hosted API prices differ by Fast or Pro variant, task, resolution, and billed duration.

Sources