AI video longer than 8 seconds: models that reach 30

The short answer

The longest clip you can create here in one generation is 30 seconds, in Seedance 2.5, Wan 3.0 and Wan 3.0 Prime. Next comes LTX-2.5 Fast with up to 20 seconds, and a row of models with up to 15 seconds, such as Kling 3.0 and MiniMax H3. The Veo 3.1 family stops at 8 seconds.

For example, a 30 second clip at 720p with sound in Wan 3.0 costs ₪11.25112.48 crWan 3.0, 720p · 30s · with sound right now.

When you need more than 30 seconds, you join several clips into one video. How to do that without the character or the style changing between the parts is explained below.

How long a clip can be in each model

The table includes only lengths that appear on the makers’ official pages and are also available here on the site. Models whose maker does not publish the length were left out, even if you can choose a long length in them.

More models reach 10 seconds too, for example: Kling 2.6 makes 5 or 10 seconds, Hailuo 2.3 makes 6 or 10 seconds, with 10 seconds only at 768p, and Ray 3.2 makes 10 seconds from a text prompt and 5 seconds when you start from an image.

What a long clip costs

In most models the price follows the seconds, so a long clip costs more. In Wan 3.0, for example, 5 seconds at 720p with sound cost ₪1.8818.75 crWan 3.0, 720p · 5s · with sound right now, and 30 seconds cost ₪11.25112.48 crWan 3.0, 720p · 30s · with sound. In some models the price is set per clip: in Veo 3.1 Lite a 4 second clip costs ₪0.707.03 crVeo 3.1 Lite, 720p · 4s · with sound right now and an 8 second clip costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound.

The table is computed as the page loads, with the same calculation the studio shows before every generation:

One long clip or several short ones

There are two ways to reach a long video, and each has a different advantage:

• One long generation keeps continuity: the same scene, the same light and continuous sound. ByteDance writes that Seedance 2.5 keeps the looks and the voices of several characters consistent through the clip
• Several short clips give you control over every shot. If one shot does not come out well, you make only that one again, and you can also mix different models for different shots

There is also a middle way: several shots inside one generation. Wan 2.6 has a "Multi-shot" switch, and Alibaba describes it as a model that tells a short story in several shots. Google’s official prompting guide for Veo 3.1 includes an example of an 8 second clip split in the prompt into four time segments, each with its own shot.

How to build a long video from several parts

When you need a minute or more, you build the video from several parts and join them in an editing app. A few steps that help the joins look natural:

1. Plan the shots in advance. Write what happens in each part, and give each one a length the model supports
2. Keep the same character. Use the same reference images or the same start image in every part. The full explanation is in the guide how to keep the same character
3. Continue from the end of the previous part. Save the last frame of the part as an image, with a screenshot for example, and upload it as the start image of the next part
4. Keep the same settings. The same model, the same resolution, the same aspect ratio and the same style description in every part
5. Add one soundtrack for the whole video. A voiceover and music created separately tie the parts together, and a model such as Eleven Music v2.5 can also create music without vocals

To get a full minute, you can for example join two 30 second clips. Right now each such clip at 720p with sound in Wan 3.0 costs ₪11.25112.48 crWan 3.0, 720p · 30s · with sound.

Continuing a clip you already made

Wan 2.7 can continue an existing clip: you upload a clip of 2 to 10 seconds, choose a total length of up to 15 seconds, and the model continues it from where it ended. According to Alibaba, it extends the clip you uploaded up to the length you chose. Right now such a continuation to 15 seconds at 720p costs ₪5.6256.24 crWan 2.7, 720p · 15s.

There is another way too: Kling O3 takes a reference video and makes a new shot from it, for example the next shot or the same camera move. According to Kling, the result is a new clip and not an edit of the clip you uploaded.

Tips for a long clip that looks good

• Describe a beginning, a middle and an end. A clip of 20 or 30 seconds has room for several actions, and it helps to write them in order
• Do not overload it. One clear action every few seconds works better than a long list of events
• Check the price before the full length. In the studio the exact price updates at once when you change the length or the resolution
• Add Hebrew separately. None of these video models lists Hebrew on an official speech language list, so a Hebrew voiceover is created in a voice model such as Eleven v3 and added in editing

There is no subscription here and no monthly quota: you pay per generation, from the same balance, in any model and at any length it supports.

Questions and answers