LTX-2.5 Video Generator and Audio: What the Open-Weights Update Changes
LTX-2.5 is the newest version of Lightricks' open-weights video model, the line that made synchronized audio and video in one pass a standard expectation. It is a foundation-model update — better motion, native multi-shot, stronger prompt adherence, automatic duration — released under a community license that allows commercial work. This page is an external reference to what it does and what it costs to run — last checked 2026-11-21.
At a glance
| External reference | LTX-2.5 (Lightricks) — ltx.io model page, the Hugging Face model card and the fal.ai hosting page |
|---|---|
| Availability on Imgo | Not on Imgo. No Imgo integration, no credit price — this page is a reference. |
| Last checked | 2026-11-21 |
| What you can generate on Imgo today | The live Seedance 2 family — 2.0, Fast, Mini — with native synchronized audio, text-to-video and image-to-video, quoted in credits before each run. |
| Official source and billing unit | Weights and license from Lightricks (LTX-2.x Community License) — hosted APIs (fal.ai, LTX Console) bill per second of output at the host's published rate — self-hosting has no per-generation fee. |
Generate video with audio on Imgo today: Seedance 2.0, Seedance 2 Fast, Seedance 2 Mini.
Audio and video in one pass
The headline capability the LTX line carries into 2.5: video and its soundtrack are generated together, so dialogue, effects and ambience land in sync without a second model or an alignment pass. On hosted endpoints this holds across the input modes — text-to-video, image-to-video, and reference-driven generation — and on both quality tiers up to 4K output on the Fast tier. If your work is talking-head explainers, product demos with sound design, or social cuts that need a beat, the single-pass audio is the reason to care.
What the 2.5 update actually changes
- Diffusion Fidelity Rendering: rendering compute is allocated by scene complexity — the pitch is frame-accurate quality that holds on a cinema screen, not just on the first glance.
- Native multi-shot: connected shots that hold character, environment, lighting and voice — a scene, not a clip.
- Prompt adherence: a Gemma 4 12B text encoder plus a custom prompt enhancer, so complex instructions survive the trip to pixels.
- Auto duration: an optional duration head predicts a clip's natural length from the prompt when you do not set the frame count.
- A fine-tunable foundation: the release is explicitly a base you can tune on your own data and domain.
The license, stated plainly
The weights ship under the LTX-2.x Community License: commercial and production use at no cost, with one material boundary — transferring fine-tunes may require a paid license, and the full binding terms (including revenue thresholds) live in the license file on the model card. Read it before building a product on the weights — this page is a summary, not the contract.
What it costs to run
Three ways to run it, three cost shapes:
- Self-host the weights: no per-generation billing — you pay for the GPU time. This is the open-weights deal.
- Hosted API (fal.ai and others): billed per second of output, by tier. The rate is the host's — check the model page for the current number rather than trusting a figure quoted anywhere — including here.
- LTX Console: Lightricks' own hosted surface, with its own credit billing.
LTX-2.5 or a hosted model you can start with today?
The honest split: if you want to own the pipeline, fine-tune on your footage, and run on your own GPUs, the open weights are the point. If you want to brief a shot and get a finished clip with audio now, a hosted model does that with no infrastructure — and on Imgo the live equivalent is the Seedance 2 family: synchronized native audio, first- and last-frame control, image-to-video from a packshot or character sheet, and a credit quote before every run. The two are not enemies — studios that self-host LTX still draft on hosted models, and vice versa.
FAQ
Is LTX-2.5 free for commercial use?
The weights carry the LTX-2.x Community License, which allows commercial and production use at no cost, with conditions — notably that transferring fine-tunes may require a paid license. The license text on the model card is the authority — read it before shipping.
Can I run LTX-2.5 on Imgo?
No. Imgo does not host LTX models. Imgo's live video set today is the Seedance 2 family with native audio — see the live model directory.
Does LTX-2.5 generate audio, or video only?
Audio and video together, synchronized, in one pass — that has been the LTX-2 line's signature and carries through 2.5 on both quality tiers.
What is the difference between the two model sizes on the card?
The release publishes a distilled fast tier and a full-quality tier with the same capabilities at different speeds — hosted endpoints price them differently per second. Check the endpoint's page for the exact split.
Sources: ltx.io's LTX-2.5 model page (capabilities, license summary), the Lightricks/LTX-2.5 Hugging Face model card (weights, license, duration head), fal.ai's LTX-2.5 page (hosting, per-second billing shape) — all checked 2026-11-21.
Share
Related Posts
No image
AURA is Kyutai's open-weights answer to real-time talking avatars: audio in, streaming talking-head video out, on one GPU — with non-commercial weight licenses that matter before you build on it. Checked 2026-11-21.
No image
Suno Visuals is Suno's song-to-video feature: clips up to 15 seconds, lyric-synced, credits sold separately. What it covers, what it can't, and the direct-your-own alternative on Imgo's live Seedance 2 family — checked 2026-11-21.
No image
LTX-2.5 is Lightricks' open-weights audio-video model: multi-shot continuity, synchronized sound, auto duration and a community license that allows commercial use. What that means in practice — checked 2026-11-21, with the live Seedance 2 family as the generate-today option on Imgo.