LTX 2.3 Fast
Lightricks' quick open video model, served by NeverNo's own provider with no outside lab reading your prompt, with a start image and an optional end image.
Updated November 9, 2026
A humpback whale gliding through golden clouds above a 1950s seaside town at dusk
At a glance
| Maker | Lightricks |
|---|---|
| Strengths in the app | Fast |
| Arena Elo (Artificial Analysis) | 839 |
| Typical wait | about 49 s |
| Who can use it | Paid plans |
| How it’s paid | Credits, from the first use |
| Privacy | Private: an open model run by NeverNo on providers chosen because they don’t keep requests. Your history stays encrypted in your browser. |
Price
What the app charges, read from the same table the servers bill with. 1 credit = $0.01.
| 5-second clip | 38 credits ($0.38) |
|---|---|
| 10-second clip | 75 credits ($0.75) |
What it is
LTX 2.3 Fast is a video model from Lightricks, offered in NeverNo's Video Studio on paid plans and paid in credits. It makes clips of 5 or 10 seconds in 16:9 or in vertical 9:16, at 1280 by 704 pixels, and it accepts a starting image and an optional ending image, so you can say where a shot begins and where it should land. It is tagged fast: a typical 5 second clip takes about 49 seconds to render.
What sets it apart in the catalog is who stands between you and the clip. Every other video model in the Studio is a third-party lab's model, reached in anonymized form: NeverNo hides your name, account and IP address, but the lab still reads your prompt and applies its own rules. LTX 2.3 Fast is private instead. NeverNo serves it through its own provider, so no outside lab's moderation layer looks at what you write, and NeverNo's own hard limits still apply to every request. The finished video is kept in your browser, not on our servers.
Be clear-eyed about the quality. On the Artificial Analysis video arena, read on November 7, 2026, LTX 2.3 Fast has an Elo of 839 for text-to-video and 830 for image-to-video, versus 1128 and 1117 for P-Video 2 Pro, and 1172 for Wan 3.0 from text. It is not tagged for sound and it is not tagged for legible lettering, so signs and titles in the picture may come out garbled. The Studio's Auto mode ranks models by Elo, which means it rarely picks LTX for a normal request.
That makes LTX 2.3 Fast a deliberate choice rather than a default: you pick it by hand when keeping the prompt away from a lab matters more than polish, when you want to storyboard a motion between two stills, or when you simply want a quick, silent draft of an idea before you commit credits to a larger model. A few habits make the most of it. Because a draft costs under a minute, treat the first take as a thumbnail sketch: judge the framing, the speed of the motion and whether the subject stays recognizable from the first frame to the last. Keep prompts concrete, with one subject, one place and one kind of light. If the result drifts, shorten the sentence rather than lengthening it. When a take finally works, you can lock it in by reusing the same wording with a different start image, which gives a family of shots that look as if they belong to one film. Students, hobby filmmakers and concept artists can use it to previsualize a storyboard panel, a title sequence or a loop for a music video before deciding what deserves a longer production. Whatever the subject, a steady rhythm of tiny trials beats one ambitious guess: change a single noun, rerun, compare, and keep notes beside each saved clip.
Strengths
Private: served by NeverNo's own provider, so no outside lab's moderation layer reads your prompt.
Start image and optional end image, which lets you control both ends of the motion.
About 49 seconds for a typical 5 second clip, quick enough to try several takes of one idea.
Both 16:9 and vertical 9:16, at 1280 by 704 pixels, in 5 or 10 second clips.
Limits
The lowest arena Elo among the Studio's video models: 839 from text and 830 from an image (Artificial Analysis, read November 7, 2026).
No sound track and no tag for legible text: a sign or a title in the picture can come out garbled.
Video needs a paid plan, and each clip is paid in credits.
When to choose it
Animating a still between two frames
Give the model a starting picture and an ending picture, say a shop front at dawn and the same street at dusk, and let it invent the movement in between. Useful for storyboards, product teasers and music-visual loops.
Cheap drafts before a big render
Test the idea, the framing and the pace with a silent draft, then send the refined prompt to a model with sound and a higher arena rating once you know exactly what you want from the shot.
Vertical clips for social feeds
Pick 9:16 for stories and short-video feeds. A 5 second loop of drifting clouds, a rotating object or a walking silhouette fits the format without any editing.
Prompt tips
Describe one clear movement, such as a slow push toward a lighthouse or a dancer turning, rather than a sequence of events; a short clip holds a single idea.
With a start image, write what changes (the light, the camera, the subject's motion) instead of describing the picture again.
Keep words out of the frame. This model is not tagged for lettering, so add titles later in your editor.
Frequently asked questions
Is LTX 2.3 Fast private?
It is served by NeverNo's own provider, with no outside lab's moderation layer, and your clips are stored in your browser. NeverNo's own hard limits still apply.
How long does a clip take?
A typical 5 second clip takes about 49 seconds to render.
Does LTX 2.3 Fast need a paid plan?
Yes. Video in the Studio is for paid plans, and each clip is paid in credits.
Can I use an ending image?
Yes. It takes a starting image and an optional ending image, in 16:9 or 9:16.
LTX 2.3 Fast or P-Video 2 Pro?
P-Video 2 Pro scores far higher on the arena (1128 versus 839 from text, read November 7, 2026), makes sound and writes legible text, but it is an anonymized lab model. LTX 2.3 Fast stays private with NeverNo's own provider.