LTX 2.3 is the latest from Lightricks — two tiers (Fast for cheap iteration, Pro for max fidelity), text-to-video / image-to-video / video extension modes, durations from 6 to 20 seconds, resolutions through 4K (2160p), and optional native audio generation.
LTX 2.3 is Lightricks' latest video model — broadest feature surface of any video model on PixelDojo. Two quality tiers (Fast and Pro), three modes (text-to-video, image-to-video, video extension), durations from 6 to 20 seconds, three resolutions including 4K (2160p), and optional synced audio generation.
Pricing scales by tier × resolution × duration: Fast at 1080p starts at 2 credits/sec; Pro at 4K runs 8 credits/sec. Most production work lives at Pro 1080p (2 cr/sec) for hero shots, Fast 1080p (2 cr/sec) for iteration, and Fast at extended durations (up to 20s) for atmospheric / longer-form work.
6–20s
Duration range
3
Modes (T2V · I2V · Extend)
4K
Max resolution
Fast tier is built for rapid iteration — cheaper, faster turnaround, supports the longest durations (up to 20s). Pro tier is the premium fidelity option for hero shots and final renders. Same prompt language across both — start in Fast, finalize in Pro.
Fast tier supports clips up to 20 seconds — the longest single-shot duration on PixelDojo. Useful for atmospheric beats, multi-step action sequences, or longer-form social content that won't fit into the 5-10 second window most models cap at.
Toggle generate_audio on and LTX 2.3 layers a synced soundscape onto the clip — ambient sound, diegetic effects, basic music cues, all aligned to the visuals. Toggle off for clean tracks you'll score yourself.
Text-to-video composes from scratch. Image-to-video animates a starting frame (and optionally a second frame for transitions). Extend Video continues an existing clip — useful when you need 30+ seconds total and want consistent motion across chunks.
Each example shows the exact prompt that produced the result. Copy any prompt with one click.
Pro · 1080p · 6s · 16:9 · audio on
A vintage 1970s convertible drives down a Pacific Coast Highway turnout at golden hour, ocean glittering on the right, cliffside on the left, smooth tracking shot from above and behind the car, warm cinematic color grade
Pro tier earns its credits on cinematic motion shots — smooth camera arcs, dynamic composition, warm color grading all land cleanly. "Smooth tracking shot from above and behind" specifies both angle and motion type unambiguously.
Fast · 1080p · 8s · 16:9 · audio off
Anime-style watercolor of a girl with twin braids running through a sunlit meadow, dandelions bursting into seed around her, soft pastel palette, hand-drawn aesthetic, gentle camera follow
Fast tier at 8 seconds is the sweet spot for stylized atmospheric beats — cheaper than Pro, longer than most models allow at this price. Name the medium ("watercolor", "hand-drawn") to commit the model to non-photo aesthetics.
Pro · 1080p · 6s · 16:9 · audio on
Hyper-real macro of a glass perfume bottle rotating slowly on a wet onyx pedestal, soft rim light catching faceted edges, a single drop of perfume falls in slow motion from the dropper above, luxury commercial mood
Pro at 1080p for product spots — material fidelity (glass refraction, wet onyx, slow-motion droplets) reads cleaner at Pro. Native audio adds subtle ambient (room tone, the drop hitting the surface) that elevates it past muted spec footage.
Fast · 1080p · 8s · 9:16 · audio on
A young chef in a steamy industrial kitchen plates a delicate dish under hanging tungsten bulbs, hands moving with practiced precision, shallow depth of field, vertical mobile-first composition, documentary lighting
9:16 vertical at 8 seconds is mobile-first social sweet spot. Fast tier handles documentary-style action well at this aspect. "Hands moving with practiced precision" gives LTX a specific motion target instead of generic "cooking".
Fast tier at 1080p (2 cr/sec) is plenty for prompt iteration — same prompt language as Pro, faster turnaround, lower cost. Once you've nailed the prompt, regenerate the hero clip in Pro for fidelity. Don't burn Pro credits iterating.
6-second clips are the default sweet spot — short enough to iterate cheap, long enough for a single beat. Push to 10-20 seconds (Fast tier only) when you need multi-beat action, atmospheric hold, or longer narrative chunks. Don't pad short prompts into long clips.
generate_audio=true for finished content that ships as-is — synced ambient gives production polish. generate_audio=false when you'll score the clip yourself, lay vocal over it, or use it as B-roll in a larger edit. Same credit cost either way.
"Smooth tracking shot", "slow forward dolly", "gentle camera follow", "static locked-off". LTX reads these literally and executes one move cleanly. Combining two motion types in one prompt usually produces compromised motion.
LTX 2.3 text-to-video only supports 16:9 (landscape) and 9:16 (vertical) — no square 1:1, no widescreen 21:9. For other aspects, generate at 16:9 and crop, or use Kling/Runway/WAN where wider aspect range is supported.
Need a continuous 30 or 45 second piece? Generate the first 20s, then use Extend Video mode with a follow-up prompt — LTX continues from the final frame with consistent motion characteristics. Cheaper and more coherent than generating two separate clips.
| Setting | Values | Notes |
|---|---|---|
| Mode | Text-to-video · Image-to-video · Extend video | I2V supports optional end_image_url for first→last frame transitions. Extend continues an existing clip. |
| Tier | Fast · Pro | Fast for iteration and long durations. Pro for hero shots and final renders. |
| Duration | 6–20s (Fast) · 6–10s (Pro) | Pro is duration-capped at 10s. Fast extends to 20s for atmospheric / longer-form work. |
| Resolution | 1080p · 1440p · 2160p (4K) | Credits scale with resolution. 1080p is the production sweet spot. |
| Aspect ratio | 16:9 · 9:16 (T2V) · auto (I2V) | I2V auto-infers from source image. No 1:1 or 21:9 — generate then crop if needed. |
| FPS | 24 · 25 · 48 · 50 | 25 is the safe default. 48/50 for slow-motion effect downstream. |
| Audio | Toggle on/off | On layers synced ambient + diegetic sound. Off ships a clean track for post-scoring. |
| Pricing | 2–8 credits/sec | 1080p = 2 cr/sec (both tiers). 1440p Fast/Pro = 3/4. 2160p Fast/Pro = 6/8. |
Fast for iteration, longer durations (up to 20s), and any time fidelity isn't the limiting factor. Pro for hero shots, final renders, material-driven scenes (macro product, fabric texture), and anywhere the difference shows. Same prompt language across both — iterate cheaply in Fast, finalize in Pro.