How to Extend a Video With AI (Make Short Clips Longer)

June 23, 2026 · Updated September 30, 2026 · 8 min read

To extend a video with AI, you upload a short clip, choose how many seconds to add, and a model generates new footage that carries on from the clip's final frames. On our Video Extend tool that means a 2–15 second source clip, 4 to 15 added seconds per run, and 10 credits per added second. You get back the new footage, not your original with the extension attached, so the last step is joining the two in any editor.

What you need before you start

Video Extend runs ByteDance's Seedance 2.0, the only model the tool offers. ByteDance's documentation (as of September 2026) sets these limits on the video you feed it:

• Length: 2 to 15 seconds. Trim longer footage first, keeping the part that leads into the moment you want continued. The model only sees what you upload.
• Format: MP4 or MOV at 24–60 fps. The upload field also accepts WebM, but ByteDance lists only MP4 and MOV as inputs, so export to one of those.
• Size: roughly 480p or larger. Each frame needs at least 407,696 pixels. An 854×480 export clears that; 640×360 does not.
• No real people's faces. ByteDance says Seedance 2.0 doesn't support uploaded videos that contain real human faces, so clips of people usually fail. Credits for a failed job come back automatically.
• A clean ending. Not a documented limit, but the last frames drive the continuation. Cut fades to black, end titles and any hard cut in the final second before you upload.

How to extend a video, step by step

1. Upload the clip

Sign in, open Video Extend and choose your trimmed clip.

2. Set the extension length

The slider runs from 4 to 15 seconds and shows the credit cost as you drag. Start at 5: enough to judge whether the continuation holds, at a third of the price of 15. If it works, you can extend the result again.

3. Match the frame

Set the aspect ratio to your source's shape. The menu offers 16:9, 9:16, 1:1, 4:3 and 3:4 and starts on 16:9, so a vertical phone clip needs 9:16 — Seedance 2.0 renders whatever ratio you ask for, and a mismatch comes back reframed. Choose 720p or 1080p to match your source; both cost the same.

4. Write a continuation prompt (optional)

Describe what happens next in up to 1,000 characters, or leave the box empty and the model continues on its own. A prompt matters most when the shot could go several ways — a car nearing a junction, a camera approaching a doorway. Examples are below.

5. Generate and download

Credits are taken when you submit, and the result appears in My tasks with a preview and a Download link. A failed job is refunded automatically; a result you simply don't like isn't, so check the input list above before you spend.

6. Join it to your original

Put your original on a timeline, place the new clip straight after it, and check the cut before you export. The section on fixing the join covers what to do if it shows.

How much longer can you make a video?

Each run adds up to 15 seconds. To go further, extend the extension: the new clip is already short enough to upload again. Two 15-second runs turn a 10-second shot into 40 seconds for 300 credits.

Don't chain indefinitely. ByteDance's prompt guide warns that using a generated video as the input for another continuation degrades image quality, and that the loss compounds with every pass. Two or three links is a sensible ceiling; beyond that, generate fresh shots and cut between them.

For comparison, Google's Veo 3.1 API only extends videos Veo itself generated: 7 seconds per step, up to 20 steps and 148 seconds in total, at 720p, returned as one combined file (Gemini API docs, as of September 2026). Video Extend works on footage from any source, within the 2–15 second input limit.

Continuation prompts to try

ByteDance's prompt guide says to name the source clip directly as “Video 1”. Write “Extend Video 1”, not “reference Video 1”, which can make the model treat your clip as a style reference and invent a new scene. Describe the motion already under way, keep to one camera move, say what must stay the same, and end with constraints. These are starting points to adapt, not results we are showing:

Drone or landscape — the camera is already moving, so keep it moving.

Product turntable — check the label closely before you use the shot.

Ambient footage — try an empty prompt first; it costs the same.

Reveal for an end card — a held final shot leaves room for a caption.

Moving vehicle — constraints like “no other vehicles” help keep new elements out.

Animal in motion — ByteDance's guide favors slow, continuous movement over sudden bursts, so ease the action down.

Why the join looks unnatural — and how to fix it

• A jump or a short replay at the cut. ByteDance's docs say an extension usually includes only the tail footage of the original, so the new clip can open by repeating your last moment. Line up the matching frames and cut there. ByteDance's own fix for jump cuts: trim 6 frames from the end of the earlier clip and 1 frame from the start of the next, then check playback.
• The frame changes shape. The aspect ratio didn't match your source. Re-run with the right setting; cropping in the editor costs resolution.
• The sizes don't line up. The new clip's pixel dimensions can differ slightly from your source's. Use a timeline set to your source resolution and let the editor scale; lossless join tools that need identical encoding will fail.
• A color or brightness shift. Apply one grade to both clips, or cover the cut with a cross-dissolve a few frames long.
• The subject changes. Pin it down with two or three stable features, as ByteDance's guide recommends — “the red car with a black roof” — and slow the action down.
• A new scene instead of a continuation. Usually the prompt: it said “reference Video 1” or described something unrelated to the last frames. Start with “Extend Video 1” and describe the action already under way.
• Captions, logos or watermarks appear. Add “no text, no subtitles, no logo”. ByteDance recommends these phrases and notes stray subtitles are much less likely in landscape than in portrait.
• A click at the end of the audio. If the new clip has sound, ByteDance's guide suggests fading the last moment out with a volume envelope; a music bed across the join also hides it.
• Blur and blotches after several runs. That is the compounding loss from chaining. Go back to your sharpest source and chain fewer times.

What extending a video costs

Video Extend charges 10 credits for each second you add; the length of your original doesn't matter. At the prices on our pricing page (September 30, 2026) — Starter Pack $9 for 200 credits, Power Pack $59 for 1,800, Pro plan $49 a month for 2,000 — that works out to:

The Pro plan column assumes you use the whole monthly allowance. Budget for retries: if a 10-second extension takes three tries to land a clean join, that is 300 credits, or $13.50 in Starter Pack credits. The 20 free credits you get at sign-up don't cover even the 4-second minimum (40 credits).

When not to extend

• You only need a second or two. Slowing the clip by 10–20% or holding the last frame in your editor is free and exact.
• The next moment is a different shot. ByteDance's guide recommends extension for continuous takes and separately generated clips for scene changes and fast action. Make the next shot with the Video Generator and cut to it.
• It needs exact text or a specific event. The model invents what happens next; it can't know your logo reveal or your product's next frame.
• The change belongs inside the clip. A new sky, color or object is a job for Video Edit; our guide to editing a video with an AI prompt has prompts to try.

Try it

Open Video Extend, upload a trimmed clip and start with 5 seconds. For continuation prompts that go beyond the examples above, our guide to writing AI video prompts covers camera and lighting language that carries straight over.