Your model stops at ten seconds. Four ways to keep going, and a 10× price spread

Most video models stop somewhere between six and sixteen seconds, and the usual answer—generate a second clip and cut—breaks continuity. The other answer is to hand the clip you have to a model whose whole job is to continue it. Four models on LUVI do that, and on September 22, 2026 the same job cost between 450 and 4,510 credits depending on which one you pick.

What does each one charge?

We quoted all four against the same ten-second source clip, extending it at 720p:

• PixVerse v6 Video-Extend: 450 credits for five more seconds
• Grok Imagine Extend-Video: 1,040 credits for six more seconds, its default length
• Wan 2.5 Video-Extend: 1,050 credits for five more seconds
• FLUX 3 Video (Extend): 4,510 credits for five more seconds

That is a tenfold spread for the same request, which makes this one of the decisions in the catalog where the model choice matters more than the prompt.

The lengths available differ too. FLUX 3 will add up to twenty seconds in one pass and PixVerse up to fifteen, while Grok's extend defaults to six. If you need a long continuation, one call on a model with a high ceiling is simpler than chaining two on a model with a low one—though not necessarily cheaper.

What do you write in the prompt?

Only what happens next. This rule is the same across all four families, and it is the one people break: recapping the source clip gives the model information it already has, and spends the prompt's weight on the part you are not generating.

The camera holds as the mist lifts off the water and a single bird crosses the frame.

No description of what came before, no re-statement of the subject, no style words the source already carries. The extension inherits look, framing and subject from the clip itself; what it does not inherit is intention, so the new action and the new sound are the only things worth writing.

What about the sound?

It depends on the family, and the difference is worth planning around.

• FLUX 3 has a generate_audio switch that is on by default, so the new section arrives with sound it generated for itself.
• Grok Imagine has no audio switch at all, so the extension will have sound whether you asked or not—write an AUDIO: clause describing what it should be.
• Wan 2.5 generates audio natively and reads sound as content, so the effects belong in the sentence.
• PixVerse v6 has a sound switch, also on by default, and the same rule applies: the toggle decides whether there is audio, the prose decides what it contains.

In every case, an extension whose sound does not match the source will read as a cut even when the picture is continuous. If your source clip is silent, say so; if it has room tone, name the same room tone.

Which one should I use?

• Cheapest continuation, picture-led: PixVerse v6, at 450 credits for five seconds, with resolution tiers down to 360p if the extension is a draft.
• A continuation that has to carry dialogue or a scored bed: FLUX 3, which is the most expensive by a wide margin but generates picture and sound jointly and reaches twenty seconds in one call.
• Already working in Wan or Grok: stay there. Both extend at around a thousand credits for five or six seconds, and staying inside one family keeps the look consistent, which is the whole point of extending rather than cutting.

A last practical note: a clip you extended can usually be extended again, but each pass inherits the drift of the one before it. Two passes are often fine; by the third, the look has usually moved far enough from the original that a fresh generation would have been the better call.

How we checked

• Models: PixVerse v6 Video-Extend, Grok Imagine Extend-Video, Wan 2.5 Video-Extend and FLUX 3 Video (Extend).
• Settings: the same 1280×720 source clip of ten seconds, extended at 720p, five seconds on three models and the six-second default on Grok.
• Date and what was read: September 22, 2026—live credit quotes and the models' parameter contracts in LUVI, plus LUVI's own prompting guides for the four families.
• Results: 450, 1,040, 1,050 and 4,510 credits respectively. FLUX 3 offered durations up to twenty seconds, PixVerse up to fifteen.
• Drawbacks: we compared prices, contracts and prompting rules, not results. No outputs were generated for this post, so it makes no claim about which model produces the cleanest join, or about how far the look drifts after two passes—that last point comes from the families' own prompting guidance rather than a measured test.

In LUVI

Pick the clip from your library, open one of the extend models, and attach the clip as its reference—then write the next beat and nothing else. The credit estimate above the Generate button will show you the spread immediately if you switch models with the same clip attached, which is the fastest way to feel how wide it is. The help center covers how resolution and duration change the price on the tiers.

New accounts start with 200 credits, and verifying a phone and a card adds up to 1,500 more. You can create a LUVI account and see how a continuation holds up against your own footage.