LTX 2: write the sound into the prompt (6s, 8s, 10s, 4K, 4cr)
LTX 2 has native audio. A silent prompt still gets a soundtrack you did not choose.
LTX 2 has native audio. A silent prompt still gets a soundtrack you did not choose.
LTX 2 is Lightricks' LTXV2 row: latest video generation with enhanced control. Category is text-to-video. No image is required. Audio is on. Max output is 4K. Durations: 6s, 8s, 10s, 12s, 14s, 16s, 18s, 20s. Qualities: 1080p, 1440p, 2160p. Styles: Cinematic, Documentary, Abstract. Frame rates: 24fps, 30fps, 60fps. Motion levels: Low, Medium, High. The catalog lists 4 credits. The matrix also prices a 4K option at 16 credits — 4 is the listed row / SD-HD figure, not a promise that 20s of 2160p costs 4.
The floor is 6s, the ceiling is 20s, and the mix is already on. Write it.
Six seconds is the short take
There is no 4s and no 5s on this ladder. Even steps from 6 to 20. If the idea is a four-second punch, you are padding to 6s or you are on another model. If the idea is a twenty-second hold, this is one of the rows that will actually go there. Do not brief 8s because another article said 8s is the default. Brief 6s to test the mix, 8s or 10s when the sentence is that long, 20s when you have a reason that fills twenty seconds of native audio.
Native audio means a silent prompt still ships a soundtrack. Name the mix: dry room, no VO, a single whoosh, traffic, silence if you will replace the bed. "We will add music later" without writing "no score" is how LTX 2 scores it for you and you fight the take in the edit.
4K / 2160p is the ceiling, with 1440p and 1080p below it. A wrong mix at 4K is a sharper wrong mix. Iterate the prompt and the motion level at 1080p. Spend 2160p on the take whose sound you would keep.
The AI video generator is the launcher. The LTX roster and best LTX model sit next to this slug. Best 4K video is the pixel map. This page is LTX 2's mix rule at a 6s floor.
Control is the rest of the row
Enhanced control in the description is the other half: motion Low / Medium / High, 24 / 30 / 60 fps, Cinematic / Documentary / Abstract. Those knobs do not replace a mix line. High motion with no sound note is a loud clip with a mystery bed. 60fps will not add dialogue. Abstract will not mute the model.
No image required: this is text-to-video. If a pack must appear, lock a still and change rows. LTX 2 from a sentence will invent the pack and then soundtrack it. That is two surprises.
Listed 4 credits is the cheap-looking number on a 4K-capable, 20-second, native-audio row. Treat it as the listed figure, not as a dare to max every knob. 20s at 2160p with High motion and a mystery score is how a 4-credit name becomes the week's spend. The matrix's 4K option at 16 is the reminder that pixels are not free just because the title says 4cr.
What a silent prompt costs
A soundtrack you strip or live with, for 6–20 seconds, at whatever quality you picked. Write sound in the same sentence as the action: "low camera, dry foley, no music, no voice." A locked spoken line is not this row — native text-to-video audio is not lipsync. Caption after; 4K does not typeset.
Pixverse 5.6 text-to-video is a different story (5 / 8 / 10s, 1080p, 75 credits). LTX 2 is 6–20 even, 4K, 4 credits listed.
Can you hear a mix?
If no, LTX 2 invents one for at least six seconds. If the mix is "replace later," write no-score or pick a silent model. If the mix is a line a mouth must match, change rows — keeping LTX 2 only to throw the baked speech away wastes native audio.
FAQ
Does LTX 2 always add sound?
Audio is marked true. A prompt that never mentions sound still leaves the model free to score. Write the mix, including no music if that is the job. The shortest clip is already 6s of that decision.
Can I generate 4 seconds at 4K?
Not 4 seconds. The ladder starts at 6s and steps by twos to 20s. 4K / 2160p is on the quality list. Short-plus-4K means 6s at 2160p, not a 4s 4K take this row does not list.
Is it really 4 credits for 4K?
4 credits is the listed catalog figure; the matrix prices SD/HD at 4 and 4K at 16. Quote both. Do not plan a 20-second 2160p master as a 4-credit sketch. Test at 6s 1080p. Promote the take whose sound you would keep.
Do I need to upload an image?
No. Text-to-video, image not required. If the still is the contract, this is the wrong label. LTX 2 will invent the object and then give it a soundtrack you also did not specify.