LTX 2.3 Text to Video Fast: write the sound into the prompt (6s, 8s, 10s, 4K, 4cr)

LTX 2.3 Text to Video Fast has native audio. A silent prompt still gets a soundtrack you did not choose.

LTX 2.3 Text to Video Fast is fast text-to-video generation. It generates a video directly from a text prompt at 1080p/1440p/2160p, 6–20s duration, with optional native audio generation. Audio is true. Credits: 4. Max output: 4K. Durations: 6s, 8s, 10s, 12s, 14s, 16s, 18s, 20s. requires_image is false. is_fast is true. Optional is the trap. Optional native audio on a row whose audio field is true means a mute prompt still comes back with a score — just not the one you would have written.

Fast still writes a soundtrack

Fast is a latency adjective. It is not a silent adjective. The feature list includes Audio-generation next to Fast, Text-driven, High-quality, Multi-resolution. Write the sound: room, line, foley, or an explicit "no speech, soft room tone only." If you needed a truly silent catalog row, pick one whose audio field is false or empty.

6s, 8s, and 10s are the title lengths because they are the right place to hear whether the optional audio matched. The spec will let you go to 20s. Twenty seconds of an accidental radio is a long accidental radio.

Run it from the AI video generator. Lightricks' family is LTX. Best LTX model is the ranking. Prompting: LTX 2.3 Text to Video Fast prompting. The family notes also live in the LTX 2.3 prompting guide.

6 to 20 seconds at 1080p / 1440p / 2160p

Qualities listed: 1080p, 1440p, 2160p. Max output: 4K. Aspect ratios: 16:9 and 9:16. That is a narrower ratio set than many stills rows — pick one of the two before you generate. There is no 1:1 on this spec.

Even-second durations only, 6 through 20. No 4s, no 5s, no 7s, no 30s. If the brief is a 5-second Turbo-shaped hook, this is the wrong floor. If the brief is a 30-second single shot, this is the wrong ceiling.

Four credits is the catalog credits field. The price matrix also shows a higher 4K option in credits; the live quote is what charges. Neither number is USD. Do not assume "Fast + 4 credits" means a free-feeling 20s 4K with a custom score. Read the quote. Write the sound anyway.

Optional audio is a prompt job

"Optional native audio generation" in the description is not "defaults to silence." Treat it as a control you must address in the prompt (and in the UI, if a toggle exists). The catalog audio value is true. Native audio is the term. Models with audio is the ranked list.

Because this is text-to-video, the model invents the picture and the track. That is correct for B-roll with atmosphere. It is incorrect for a SKU. Lock a still elsewhere if the product has to match. LTX 2.3 Fast will not honor a die line you never uploaded (requires_image is false).

4K video instead of 1080p is the credits scenario if you are choosing a quality tier. Best 4K AI video model is the ranking. This page is the Fast row's audio rule.

Write the room, then pick a length

Name the first frame, the move, the last frame, and the soundtrack. Then pick 6s or 8s or 10s until that sentence works. Then climb the even numbers toward 20s. Fast will let you skip the climb. The soundtrack will still be whatever you failed to specify.

FAQ

If audio is optional, can I leave it out of the prompt?

You can. You may still get a soundtrack you did not choose, because the row's audio field is true and the description includes native audio generation. Specify the track or specify the absence of speech.

What lengths exist?

Even numbers from 6 to 20 seconds: 6, 8, 10, 12, 14, 16, 18, 20. No 4s, no 5s, no 30s on this slug.

Is this 4K or 1080p?

Both, plus 1440p. Qualities are 1080p / 1440p / 2160p, max output 4K. Pick the tier in the UI. Do not prompt a resolution the quote did not include.

Does Fast take a start image?

requires_image is false and the category is text-to-video. Prompt-only. If the still is locked, use an image-to-video row rather than asking Fast to re-imagine the product for up to 20 seconds with a score.