Skip to main content
Models & Technology

LTX-2.5 Model Debuts: AI Generates 10-Second 720P Videos in Just 6.8 Seconds with Native ComfyUI Integration

Open-world model company LTX announced yesterday (August 11) the launch of the LTX-2.5 model, natively integrated with ComfyUI. Generating a 10-second 720P video takes just 6.8 seconds on a setup with 2 NVIDIA GB200s.

LTX-2.5 Model Debuts: AI Generates 10-Second 720P Videos in Just 6.8 Seconds with Native ComfyUI Integration

Open-world model company LTX announced yesterday (August 11) the launch of the LTX-2.5 model, natively integrated with ComfyUI. Generating a 10-second 720P video takes just 6.8 seconds on a setup with 2 NVIDIA GB200s.

For access, users can obtain the model or call hosted generation services through Hugging Face, ComfyUI, and the LTX API. For licensing, organizations with annual recurring revenue (ARR) below $10 million (note: approximately RMB 67.586 million at the current exchange rate) can use it for free; larger companies must negotiate a commercial license.

LTX-2.5 Model Debuts: AI Generates 10-Second 720P Videos in Just 6.8 Seconds with Native ComfyUI Integration
LTX-2.5 Model Debuts: AI Generates 10-Second 720P Videos in Just 6.8 Seconds with Native ComfyUI Integration

In terms of pricing, LTX-2.5 Fast generates 720p video with audio at $0.09 per second (approximately RMB 0.61 at the current exchange rate), with a 10-second clip costing $0.90 (approximately RMB 6.1 at the current exchange rate).

Rank

Model

Cost per second

Cost per 10 seconds

Distinctive differentiator

Notes

1

Veo 3.1 Lite (Google)

$0.05

$0.50

Lowest price in the category, the lowest published price overall

Does not support 4K or clip extension

2

LTX-2.5 Fast (Lightricks)

$0.09

$0.90

The industry's only open-weights model; users with annual recurring revenue below $10 million (approximately RMB 67.586 million at the current exchange rate) do not need to self-host and are allowed to fine-tune it.

Scales to 4K at a cost of $0.30 per second (approximately RMB 2 at the current exchange rate); each generation can be up to 20 seconds long at 24/25 fps.

3

Veo 3.1 Fast (Google)

$0.10

$1.00

The cheapest closed-source 4K API path ($0.30 per second, approximately RMB 2 at the current exchange rate)

The budget offering in the Veo series

3

Gemini Omni Flash (Google)

$0.10

$1.00

Independently evaluated as a quality leader, ranking first in AI analysis of text-to-video as of August 2026.

Supports only 720p; maximum length of 10 seconds; best tool for iteration

5

LTX-2.5 Pro (Lightricks)

$0.12

$1.20

The premium optimized version in the only open-weights series—fast alignment, fonts, and typography (according to the vendor).

Maximum resolution is 1080p, with a duration of 10 seconds

6

FLUX 3 Video (Black Forest Labs)

$0.17

$1.70

The first model capable of delivering 20-second video clips with audio

High-definition tier; includes audio; draft tier costs $0.06 per second (approximately RMB 0.41 at the current exchange rate) ($0.60 per clip, approximately RMB 4.1, HD only)

7

HappyHorse 1.0 (Alibaba)

~$0.182

~$1.82

Arena quality leader at launch (April 2026), but surpassed since; the “open-source” claim has never been supported by verifiable downloadable weight data.

Third-party reseller pricing; audio included for free

8

Veo 3.1 (Google)

$0.40

$4.00

Premium version; 8× Veo 3.1 Lite; 1080p resolution is available instead of 720p at no additional cost

The most prominent metric in LTX's launch materials is speed: on 2 NVIDIA GB200 chips, it can generate a 10-second 720p image-to-video clip in 6.8 seconds. The same task takes 23.7 seconds through LTX's hosted API, but the API outputs at 1080p resolution (with no 720p option).