By Higgsfield Genjutsu editorial teamPublished Nov 6, 2026Updated Nov 6, 2026
How long can a Genjutsu video be?
On this site, a single render starts from a source video between 3 and 30 seconds. The output is designed to follow the duration of that source, so one run is a short clip rather than a full timeline.
Longer pieces are made from several approved clips, recast shot by shot and assembled in an editor. That is the same workflow used for multi-shot music videos, ads and series content.
Plan the length before you upload.| Target | Source length | How to produce it |
|---|
| Quick identity or motion test | 3–6 seconds | One 720p run, reviewed before scaling |
|---|
| Short social clip | 6–15 seconds | One run; finish audio and captions in post |
|---|
| Maximum single run | Up to 30 seconds | One source clip; check the credit estimate first |
|---|
| Longer edit | Multiple 3–30 second clips | Recast each shot, then assemble in an editor |
|---|
Why is the source limited to 3–30 seconds?
The model needs enough motion to understand the performance, but every extra second costs GPU time and provider processing. A 3-second minimum avoids near-still inputs; a 30-second maximum keeps one render reviewable and affordable.
The limit also protects consistency. The longer the clip, the more chances for identity drift, hand artifacts or camera changes. A short, clean take is easier to judge than a long sequence with cuts.
Does the output match the source length?
The render is designed to preserve the source timing, so the downloaded clip should follow the source duration. Check the actual file after download instead of assuming, especially if an engine fallback was used.
If the output is shorter or longer than expected, first confirm the duration of the uploaded file. The server measures MP4/MOV duration and recalculates the estimate from that file, so a source that was not trimmed as expected changes the result.
How do you make a video longer than 30 seconds?
Build it as a sequence of shots. Keep each source section within 3–30 seconds, recast one shot at a time, and assemble the approved clips on an editing timeline. The model does not take a 60-second source or assemble a multi-shot timeline for you.
Plan the cuts before generating. A 60-second piece might be four 15-second shots, a 90-second piece six 15-second shots, and so on. Keep the character references and prompt structure consistent across the series.
- Write a shot list with duration and purpose for each clip.
- Keep every source section within the 3–30 second limit.
- Recast one shot at a time and review before continuing.
- Assemble, trim and transition in an editor.
- Add music, captions and the final grade after assembly.
How does duration affect Genjutsu credits?
On this site, credits are calculated from source duration and selected quality. A longer source or a 1080p finish costs more than a short 720p test. The generator shows the estimate before submission and the server recalculates from the uploaded file.
That makes the first test a cost decision as well as a creative one. Start with the shortest section that proves identity, motion and camera, then extend only after the setup is approved.
- Short 720p test: lowest credit cost.
- Longer 720p clip: more credits, more consistency risk.
- 1080p finish: higher rate for the same duration.
- Multiple shots: multiply the per-shot cost.
- Failed jobs are designed to refund the charged credits.
What should you do at the 30-second limit?
Do not upload a longer source and hope it is trimmed automatically. Cut the exact section you need before upload, keep it continuous and remove internal edits or speed ramps.
If the action needs more than 30 seconds, split at a natural cut and recast each side. Keep a small overlap between sections so the editor has room to blend the transition.
- Trim the source to the exact section you need.
- Split long actions at a natural cut point.
- Keep a few frames of overlap for the edit.
- Recast each section with the same references.
- Assemble and review the full sequence at phone size.
What are the common length mistakes?
The most common mistake is submitting a long, multi-cut source because it contains the right performance somewhere inside it. The model sees the cuts and the extra footage becomes another variable.
The second mistake is judging a character setup on a 30-second 1080p run instead of a 3–6 second 720p test. That spends more credits before the reference, prompt and framing are proven.
- Long source with internal cuts.
- Untrimmed camera resets or dead time.
- Testing identity on the maximum duration first.
- Changing duration and quality in the same retry.
- Expecting one run to cover a full multi-shot edit.
Key takeaways
- One source clip can be 3–30 seconds on this site.
- The output is designed to follow the source duration; verify the file.
- Longer videos are multi-shot projects assembled in an editor.
- Duration and quality determine the credit estimate.
- Trim and test a short 720p section before scaling.
Higgsfield Genjutsu