# Higgsfield Genjutsu Prompts: 8 Creative Briefs Copy eight original prompt examples for reference-led transformations and image-to-video experiments, with input notes, review criteria and focused revisions. By GenjutsuLite Team | Published 2026-10-29 | 10 min read HTML source: https://genjutsulite.com/hub/higgsfield-genjutsu-prompts Canonical HTML: https://genjutsulite.com/hub/higgsfield-genjutsu-prompts Category: Blog A useful Genjutsu prompt says what changes and makes the target unambiguous. It does not have to retell every frame of a performance that is already visible in your source video. Start with the inputs and intended edit, then add only the direction needed to explain it. These **eight original prompt examples are untested creative briefs**, written by the GenjutsuLite team. They are not official Higgsfield presets, recovered private prompts or demonstrations of generated results. A matching output is not guaranteed. **GenjutsuLite is independent of Higgsfield, including Higgsfield's own Genjutsu Lite.** Our [template workbench](/templates) uses labeled photo roles; there is no free-text prompt box in our purchase flow. Use this library to plan a scene or adapt a brief in a tool that accepts the specified inputs. ## Match the prompt to the input workflow | Prompt group | Required starting point | What the words should clarify | | --- | --- | --- | | Examples 1, 2, 5 and 6: motion transfer | Source performance plus visual references | The new cast or scene built around that performance | | Examples 3 and 4: object replacement | An existing shot plus replacement references | Which visible item changes and what should remain | | Examples 7 and 8: image-to-video | A starting image in a compatible tool | The action and camera movement after that image | Higgsfield's [Genjutsu introduction](https://higgsfield.ai/blog/higgsfield-genjutsu) describes reference-driven transformation and optional text direction. Our [tutorial](/hub/higgsfield-genjutsu) explains how to choose between its modes. Examples 7 and 8 use a different starting point: a still image, not an existing performance video. ## A small prompt formula you can adapt For a video transformation, write **target + replacement reference + scope of change + continuity goal**. For example, identify the central person's jacket, point to the intended jacket reference and state which surrounding parts of the shot you want kept. For image-to-video, try **subject action + scene motion + one camera move**. Runway's [Gen-4 prompting guide](https://help.runwayml.com/hc/en-us/articles/39789879462419-Gen-4-Video-Prompting-Guide) recommends starting simply, describing motion and using positive phrasing. We apply that advice to examples 7 and 8; it is not proof that identical wording works across every model. Choose duration, resolution and aspect ratio in the tool's actual settings when they are available. Writing a format into a prompt does not establish an export option or override a model's supported limits. Read the displayed cost before each paid run. ## Eight example prompts Use **Copy prompt** to copy an instruction. If clipboard access is blocked, the page offers manual selection. Replace details so they agree with your actual references, then review the full result using the notes below each example. ### 1. Character recast for a walking shot **Workflow:** Higgsfield Genjutsu — Motion transfer. **Prepare:** A source clip of one person walking, plus consistent reference images of the replacement character. ```text Rebuild the walking scene around the character in the reference images. The character wears the reference outfit in a quiet stone courtyard. Carry over the source performance, timing and camera path. Keep the character recognizable as the head turns. ``` **Review:** Inspect the face during the turn, the outfit at the end and the contact of each foot with the ground. **Next experiment:** Keep the original setting for your next attempt if changing both the cast and location makes it difficult to identify the problem. ### 2. One-person dance in a new setting **Workflow:** Higgsfield Genjutsu — Motion transfer. **Prepare:** A clearly visible dance performance and reference images of your intended performer. ```text Recast the dancer as the person in the reference images, wearing the reference outfit, on a small studio stage. Follow the source dance and camera framing. Keep the full performer readable through the turn and finish. ``` **Review:** Compare the timing of the main gestures and check hands, feet and the face during fast movement. **Next experiment:** Use a less occluded portion of the same performance before adding more scene detail. A motion description alone does not supply dance choreography. ### 3. Replace a clearly identified bag **Workflow:** Higgsfield Genjutsu — Objects swap. **Prepare:** An existing shot showing one visible bag and clear images of the replacement bag from relevant angles. ```text Replace the bag carried by the person on the left with the bag in the reference images. Match its visible shape, material and strap. Preserve the person, surrounding scene and original shot timing. ``` **Review:** Look at the strap as it passes the shoulder and the bag outline when an arm moves across it. **Next experiment:** Try a source where the complete bag stays visible. If exact lettering matters, check it separately rather than assuming a reference guarantees readable text. ### 4. Change one jacket in a finished shot **Workflow:** Higgsfield Genjutsu — Objects swap. **Prepare:** A clip with a clearly identifiable subject, plus images of the intended jacket. ```text Change the jacket worn by the central person to the jacket in the reference images. Match the collar, material and overall fit. Keep the person, trousers, background and existing performance consistent with the source shot. ``` **Review:** Check cuffs, the collar, moving fabric and the boundary between the jacket and hands. **Next experiment:** Isolate the jacket change before attempting an entire outfit replacement. Use references that agree about the garment design. ### 5. Restage a performance at dusk **Workflow:** Higgsfield Genjutsu — Motion transfer. **Prepare:** A source performance and references for the cast and intended dusk environment. ```text Rebuild the source performance in a quiet rooftop setting at dusk, using the reference character and outfit. Retain the source movement, pacing and camera path. Use soft evening light with a readable face and clear separation from the skyline. ``` **Review:** Inspect whether the face remains visible and whether the scene lighting stays coherent through camera movement. **Next experiment:** Start with a brighter setting if the subject becomes difficult to assess. Darkness can hide a problem rather than solve it. ### 6. Two distinct characters in one performance **Workflow:** Higgsfield Genjutsu — Motion transfer. **Prepare:** A source with two readable performers and separate, consistent visual references for each intended character. ```text Rebuild this two-person scene with the first reference character in the left position and the second reference character in the right position. Each follows the performance of the source person in that position. Keep their outfits distinct throughout the shot. ``` **Review:** Check whether identities switch when the performers cross or overlap, and whether clothing remains assigned correctly. **Next experiment:** Use a source where the two subjects remain separated before attempting a crossing movement. This instruction is not a guarantee of persistent identity. ### 7. A quiet portrait from a starting image **Workflow:** General image-to-video — not a Genjutsu source-video instruction. **Prepare:** A clear portrait image, in a tool that accepts an image and a motion prompt. ```text The subject turns slightly toward the window and gives a small smile. A gentle breeze moves a few strands of hair. The camera slowly moves closer in one continuous shot. ``` **Review:** Compare the face with the input and check whether the camera movement overwhelms the subtle action. **Next experiment:** Try a locked camera while keeping the same expression change. For image-to-video, motion is more useful than describing every visible detail again. ### 8. A simple product reveal from an image **Workflow:** General image-to-video — not a Genjutsu source-video instruction. **Prepare:** A clean product image with a useful starting composition. ```text The camera makes a slow lateral move past the product. A soft highlight travels across its surface. The product rests steadily on the table throughout one continuous shot. ``` **Review:** Watch the silhouette and surface details, especially any logo or small text that must remain accurate. **Next experiment:** Use a smaller camera move before combining motion with a new environment. A generated reveal is not evidence of exact product fidelity. ## Why a shorter revision is often more informative An overloaded brief might ask for a new face, outfit, location, lighting setup, camera orbit and several transitions together. If the result drifts, you cannot easily tell which request conflicted with the source or references. Our suggested revision process is to keep a baseline and change one thing. First evaluate whether the target replacement is recognizable. Next decide whether the setting needs to change. Only then consider a different look. This is a method for learning from attempts, not a claim that extra prompting removes model limitations. Keep references consistent with the text. If the prompt asks for a blue jacket while the reference shows a red coat, choose which instruction you actually mean. If the source hides a product behind a hand, a longer description does not provide the missing visual information. ## A review sheet for your next attempt | Record | Why it helps | | --- | --- | | Source clip or starting image | Keeps the input comparable | | Chosen mode and model | Prevents comparing unrelated workflows as if they were one | | Reference set and exact prompt | Makes a useful revision possible | | Output settings and displayed charge | Separates the creative brief from export and spending choices | | One observed defect and one proposed change | Keeps the next attempt focused | Review the start, a difficult middle frame and the end. Watch for identity shifts, changing product shapes and hands crossing the target. Also listen to the output: a visual prompt is not a promise of an identical soundtrack, licensed music or synchronized speech. ## Using the idea in a GenjutsuLite template Choose the scene that already contains the action you want. [Browse templates](/templates), watch the full preview and select **Recreate**. Supply one clear image per role; do not put a prose brief into a photo slot or look for a custom-video upload control that our workbench does not offer. The order is for a selected template, with its own full duration, available formats and audio disclosure. The current options are one video at 480p for USD 7.99 or 720p for USD 11.99; see [pricing](/#pricing) and the final checkout total. Videos are manually prepared after confirmed payment, with no fixed delivery time promised. Save the original order link for recovery and download. ## FAQ ### Is there a single best Higgsfield Genjutsu prompt? There is no universal brief that supplies suitable references, chooses the right source and guarantees the desired edit. Start with the specific target and use the shortest instruction that explains your intended change. ### Should I paste a long negative prompt? Check the selected model's guidance. Runway's linked guide recommends positive descriptions for Gen-4. Rather than treating a negative-prompt list as portable, state the wanted behavior clearly, such as “the camera remains locked.” ### Have these eight prompts been generation-tested? No. They are original planning examples. We have not generated comparison samples or validated third-party billing for this library. The review notes identify what to inspect when you test your own inputs. ### Will “same music” or “perfect face” guarantee those properties? No. Text does not establish audio rights or identity fidelity. Inspect the result and the selected tool's audio behavior. On GenjutsuLite, read the individual template's audio disclosure before purchasing. ### Can these prompts customize a GenjutsuLite order? Our current checkout does not accept custom prompts. [Select a template](/templates) whose action already suits your idea, or read [the alternatives comparison](/hub/higgsfield-genjutsu-alternatives) if you need a different workflow. ## Sources and further reading Checked October 29, 2026: [Higgsfield's Genjutsu introduction](https://higgsfield.ai/blog/higgsfield-genjutsu) and [Runway's Gen-4 video prompting guide](https://help.runwayml.com/hc/en-us/articles/39789879462419-Gen-4-Video-Prompting-Guide). The prompts, preparation suggestions and review notes are our own editorial examples. Continue with [the Genjutsu tutorial](/hub/higgsfield-genjutsu), [workflow alternatives](/hub/higgsfield-genjutsu-alternatives) or [photo preparation for our templates](/hub/photos-for-ai-video-templates).