LTX 2.5 is Lightricks' open-weights AI video generation model. It renders 6–20 second clips at up to native 4K with HDR, generates audio in sync with the picture, and holds several shots in one generation with characters, light and sound consistent across the cuts. Its Diffusion Fidelity Rendering approach allocates rendering compute by scene complexity, and a Gemma 4 12B text encoder reads your prompt.

Are LTX 2.5, LTX Video and LTXV the same thing?

Same family, latest generation. LTX Video — often shortened to LTXV — is Lightricks' open-source video model line, and LTX 2.5 (also written LTX-2.5) is its newest generation. If you've seen "LTX Video 2.5" or "LTXV 2.5", they all refer to the model this page tracks.

Is LTX 2.5 really open weights, and can I use it commercially?

Yes. The weights are published under the LTX-2.x Community License: commercial use is free for teams under $10M in annual revenue, LoRA and IC-LoRA fine-tuning is allowed, and the model runs on consumer GPUs from 16GB of VRAM. On self-hosted hardware Lightricks quotes faster-than-real-time rendering — a 10-second 720p clip in 6.8 seconds on 2×GB200. Videos you generate on Molyin come with standard commercial usage under our terms.

How does native multi-shot work?

One generation can contain several shots — no stitching. You write the cuts as prose ("cut to…", "the camera now opens on…") and the model keeps characters, scenery, lighting and the audio bed consistent across them. Lightricks recommends 2–4 cuts per generation; beyond that, coherence starts to compete with clip length.

What is Auto duration?

Instead of fixing a clip length, you can let LTX 2.5 choose it: the model reads the beats of your prompt — how many actions, lines and cuts it holds — and picks a duration that fits, within the 6–20 second range. Handy for dialogue scenes, where forcing a fixed length either rushes or pads the delivery.

Does LTX 2.5 generate sound?

Yes — audio is native and generated in sync with the picture: dialogue in quotes gets spoken, ambience and effects match what's on screen. It's toggleable, so you can render silent when you only need visuals. The open-weights release also documents an audio-to-video capability that drives the picture from an existing soundtrack.

Can LTX 2.5 really render 4K?

Yes — the Fast tier reaches native 4K with HDR, and an EXR/RAW pipeline is part of the release for serious color grading. At 1440p and 4K, clips come in 6, 8 or 10 seconds; the full 6–20 second range belongs to 720p and 1080p. A practical workflow is drafting at 720p and re-running the winning prompt at 4K.

Can I use LTX 2.5 on Molyin today?

Yes — LTX 2.5 is live on Molyin right now, in the generator at the top of this page. Fast and Pro tiers both run on the same credits as every other model, and every sample on this page is a one-click Remix on the real model.

How much does LTX 2.5 cost on Molyin?

Per-second credit pricing is live on our credits page: LTX 2.5 Fast runs 20–67 credits per second from 720p up to 4K, and the Pro quality tier runs 27–38 at 720p and 1080p — all from the same balance as every other model. No separate plan, no new wallet.

How do I write good prompts for LTX 2.5?

For a single shot, write one flowing present-tense paragraph of about 4–8 sentences — set the scene, then let actions unfold in order, with dialogue inside quotes so it gets spoken. For multi-shot, describe the transitions and the audio continuity in prose rather than as a list. Camera language maps to eight motion presets — dolly in/out/left/right, jib up/down, static, focus shift — and prompts can run up to 5,000 characters. All five samples on this page are official prompts; hit Remix to study how they're built.

LTX 2.5 or MiniMax H3 — which should I use?

They're complements. MiniMax H3 is live on Molyin right now, with always-on stereo sound, native 2K and a big reference intake — up to 9 images plus videos and audio. LTX 2.5 brings open weights, native multi-shot, deterministic camera presets and a 4K HDR finishing pipeline. The practical answer: multi-shot scenes, deterministic camera moves and 4K masters point to LTX 2.5; big reference stacks and native 2K point to MiniMax H3 — both live, same credits.

LTX 2.5 (LTX Video 2.5) — Lightricks Open-Weights 4K Video Model

LTX 2.5

Lightricks' open-weights video model — multi-shot scenes with synced sound, up to native 4K HDR, and rendering compute spent where each scene actually needs it. Live on Molyin right now — Fast and Pro tiers in the generator below.

Failed generations auto-refunded
Credit packs never expire
Every model's price is public
Cancel anytime in one click

6–20 s
Clip length: Up to 4K
Four resolution tiers: Synced
Native audio, toggleable: Open weights
Self-host & fine-tune

Official samples with the full prompt behind each one

A small-town oil strike on live television, a backlit brown bear in morning mist, a frog yoga class that loses its composure, a golden-hour office brought to life from a single frame, and a shower of welding sparks — all five clips come from Lightricks' official LTX 2.5 sample prompts, and each one ships the complete prompt that made it. Hit Remix to load a prompt straight into the generator above, and run it on LTX 2.5 right now.

LTX 2.5
Text to video
20s
1080P

EXT. SMALL TOWN STREET – MORNING – LIVE NEWS BROADCAST The shot opens on a news reporter standing in front of a row of cordoned-off cars, yellow caution tape fluttering behind him. The light is warm, early sun reflecting off the camera lens. The faint hum of chatter and distant drilling fills the air. The reporter, composed but visibly excited, looks directly into the camera, microphone in hand. Reporter (live): “Thank you, Sylvia. And yes — this is a sentence I never thought I’d say on live television — but this morning, here in the quiet town of New Castle, Vermont… black gold has been found!” He gestures slightly toward the field behind him. Reporter (grinning): “If my cameraman can pan over, you’ll see what all the excitement’s about.” The camera pans right, slowly revealing a construction site surrounded by workers in hard hats. A beat of silence — then, with a sudden roar, a geyser of oil erupts from the ground, blasting upward in a violent plume. Workers cheer and scramble, the black stream glistening in the morning light. The camera shakes slightly, trying to stay focused through the chaos. Reporter (off-screen, shouting over the noise): “There it is, folks — the moment New Castle will never forget!” The camera catches the sunlight gleaming off the oil mist before pulling back, revealing the entire scene — the small-town skyline silhouetted against the wild fountain of oil.

LTX 2.5
Text to video
10s
1080P

A slow lateral tracking shot glides through old-growth pines, following a large brown bear as it moves along a shallow river bank, the camera drifting with it in parallel. Low morning sun cuts sideways through the trunks in hard parallel shafts, backlighting the bear's fur into a bright rim of gold against deep shadow; its breath clouds in the cold air with every step. Suspended mist and pollen turn slowly in the light beams. The bear pauses, lowers its head toward the water, then moves on. Telephoto compression, shallow depth of field, pine trunks sliding through the foreground, high dynamic range, cool green shadows against warm highlights, natural wildlife cinematography, 35mm film grain. Sound: shallow water running over stones, wind high in the pines, wet gravel underfoot, distant birdsong.

LTX 2.5
Text to video
20s
1080P

The camera opens in a calm, sunlit frog yoga studio. Warm morning light washes over the wooden floor as incense smoke drifts lazily in the air. The senior frog instructor sits cross-legged at the center, eyes closed, voice deep and calm. “We are one with the pond.” All the frogs answer softly: “Ommm…” “We are one with the mud.” “Ommm…” He smiles faintly. “We are one with the flies.” A pause. The camera pans to the side towards one frog who twitches, eyes darting. Suddenly its tongue snaps out, catching a fly mid-air and pulling it into its mouth. The master exhales slowly, still serene. “But we do not chase the flies…” Beat. “not during class.” The guilty frog lowers its head in shame, folding its hands back into a meditative pose. The other frogs resume their chant: “Ommm…” Camera holds for a moment on the embarrassed frog, eyes closed too tightly, pretending nothing happened.

LTX 2.5
Image to video
10s
1080P

A man enters from off frame left and crosses the sunlit office in a few unhurried strides. He sets a ceramic coffee mug down on the marble tabletop beside the stacked books and the small potted succulent, steam curling up through the low golden light. He pauses briefly, then walks toward the floor-to-ceiling glass and stops, silhouetted against the hazy skyline, looking out. Static locked-off camera, deep focus with the brass lamp anchoring the foreground; hard low sun raking in from frame right, long warm shadows crawling across the marble, gentle lens haze. 35mm film grain, natural cinematic realism, single continuous shot, no cuts. Sound: quiet footsteps on hard floor, the soft ceramic knock of the mug on marble, muffled city hum through glass.

LTX 2.5
Text to video
10s
1080P

A lateral tracking shot glides past a row of industrial presses in a cavernous steel mill. A welder's arc erupts in the mid-ground, throwing hard blue-white light and a shower of sparks that arc and die on the concrete floor. Overhead sodium lamps cut hazy orange cones through smoke. Silhouetted workers cross the frame in the foreground. High-contrast mixed color temperature, heavy atmospheric particulate, gritty texture, slow-motion spark trails, hard industrial realism.

What Is LTX 2.5?

LTX 2.5 is Lightricks' open-weights AI video model — the latest generation of the LTX Video family. Its core idea is Diffusion Fidelity Rendering: instead of spending the same compute on every frame, the model allocates rendering effort where a scene actually needs it, backed by a new diffusion video decoder. A Gemma 4 12B text encoder reads your prompt — with an optional prompt enhancer — and audio generates in sync with the picture, so dialogue, ambience and effects arrive together with the frames. Native multi-shot goes further: one generation can hold several cuts, and characters, scenery, light and sound stay consistent across them.

The spec sheet is built for finishing work, not just previews. Clips run 6–20 seconds in even steps at 24 or 25 fps in 720p or 1080p — or 6, 8 and 10 seconds at 1440p and native 4K with HDR, backed by an EXR/RAW pipeline for color grading. 48 and 50 fps are available for clips up to 10 seconds, framing covers 16:9 and 9:16, eight camera-motion presets — dolly, jib, static, focus shift — give you deterministic moves, and Auto duration lets the model pick the length that fits your prompt. The weights themselves are open under the LTX-2.x Community License: free commercial use under $10M annual revenue, LoRA and IC-LoRA fine-tuning, and it runs on consumer GPUs from 16GB of VRAM.

LTX 2.5 is live on Molyin — the panel above runs it right now, next to every model in our AI video generator, all on one credit balance. Browse every model we host, or check transparent per-second rates on the pricing page.

LTX 2.5 vs MiniMax H3

Two very different flagships — both live on Molyin today. Here is how their sheets line up, so you know which prompts to point where.

Spec	LTX 2.5	MiniMax H3
Duration	6–20s in even steps, plus Auto duration	Any length from 4 to 15 seconds
Resolution	720p · 1080p · 1440p · 4K with HDR	Native 2K default, 768P economy tier
Audio	Native synced audio, toggleable	Native stereo, always on
Multi-shot	Native — several cuts in one generation, consistency held across them	Multi-shot direction via timeline prompts
Inputs	Image-to-video with first & last frame	Up to 9 images + 3 videos + 3 audio clips
Open weights	Open — self-host and fine-tune with LoRA	Open weights announced
Price on Molyin	20–67 credits/s (720p–4K), Pro tier 27–38	18 / 29 credits/s (768P / 2K)
Reaches its best in	Multi-shot scenes, 4K HDR masters, deterministic camera moves	Available right now — stereo sound, big reference stacks, native 2K

Rates on both sides are current execution prices on Molyin — per second of video, from the same credit balance.

LTX 2.5 vs models live on Molyin

Model	Positioning	Release Date	On Molyin
LTX 2.5	Lightricks open-weights model — native multi-shot + 4K HDR	Aug 11, 2026 (open weights)	Live — generate above
Seedance 2.5	ByteDance flagship — 30s takes, directed second by second	Aug 7, 2026 (API)	Live — generate above
FLUX 3	Black Forest Labs — storyboard keyframes + take continuation	Aug 7, 2026 (video API)	Live — generate above
Wan 3.0	Alibaba's All-in-One model, up to 30s takes	Aug 6, 2026	Live — generate above
MiniMax H3	Omni-modal flagship, native 2K + stereo	Jul 31, 2026 · open weights announced	Live — generate above
HappyHorse	Expressive generation + instruction video editing	Jun 2026 (1.1)	Live — generate above
Seedance 2.0 family	Standard / Fast / Mini workhorses, up to 4K	Apr 2026 (API)	Live — generate above
Kling 3.0	Native audio in 5 languages + element references, up to 4K	Feb 4, 2026	Live — generate above
Veo 3.1	Google's cinematic model with audio	Oct 2025	Live — generate above

One subscription, one credit balance — every model available above is included.

LTX 2.5 full specifications

LTX 2.5 full specifications
Parameter	Value	Notes
Generation modes	Text to video · Image to video (first & last frame)	Prompt from scratch, animate a still, or pin both endpoints of the clip with a start and end frame.
Model tiers	Fast (up to 4K) · Pro (up to 1080p)	Lightricks ships two official tiers; their resolution ceilings differ.
Resolution tiers	720P / 1080P / 1440P / 4K	720p and 1080p carry the full duration range; 1440p and 4K render 6, 8 or 10-second clips.
Clip length	6–20 s	Even-second steps at 24/25 fps in 720p and 1080p — or hand the choice to Auto duration.
Frame rates	24 / 25 / 48 / 50 fps	At 48 or 50 fps, clips cap at 10 seconds.
Aspect ratios	16:9 · 9:16	Widescreen and vertical.
Audio	Native synced audio — optional toggle	Sound generates in sync with the picture; switch it off when you only need the visuals.
Multi-shot	Native — 2–4 cuts recommended	One generation holds several shots; characters, scene, light and sound persist across the cuts.
Auto duration	Supported	Leave the length blank and the model picks a duration that fits the beats of your prompt.
License	LTX-2.x Community License — open weights	Free commercial use under $10M annual revenue; weights published August 11, 2026.
Fine-tuning	LoRA / IC-LoRA	Trainable on consumer GPUs — 16GB of VRAM is the published minimum.
Seed control	Not supported	No fixed-seed reproduction — iterate by refining the prompt or remixing a result you like.

Specifications reflect Lightricks' official LTX 2.5 release; values will be wired to our live configuration when the model goes live on Molyin.

LTX 2.5 pricing on Molyin

Fast and Pro are billed per second as two tiers of the same model, from the same credit balance as everything else on Molyin. The numbers below are computed straight from our billing engine — and failed generations are refunded automatically.

AI video model credit rates by resolution — credits per second and per-clip examples
Model	Resolution	Credits/sec	5s Video	5s Video + 3s Reference
LTX 2.5	720p	20	100 credits	—
	1080p	29	145 credits	—
	1440p	43	215 credits	—
	4k	67	335 credits	—
LTX 2.5 Pro	720p	27	135 credits	—
LTX 2.5 Pro	1080p	38	190 credits	—

See credit pricing for every model on the full price list →

How LTX 2.5 will work on Molyin

Once live, LTX 2.5 runs like every other model here — three steps from idea to file.

Pick a model and a mode

Pick any live video model in the generator above — LTX 2.5 joins the same dropdown once it goes live.

Write the prompt

Describe one scene as a flowing present-tense paragraph and put dialogue in quotes — or hit Remix on a sample from this page to start from a prompt that's proven to work.

Generate and export

Choose resolution, length and audio, generate, preview in place, then download the original file.

LTX 2.5 FAQ

What creators ask about Lightricks' open-weights model while it makes its way to Molyin.

Start Creating with Molyin Today

Start Creating Free View Pricing