← WidemileCONTENT HISTORY
WHAT CHANGED
Update to Widemile
Snapshot Oct 24, 2026 · 13:07 UTC · version 1.1.0
Collection source: downloaded plugin package.
First saved snapshot
No earlier snapshot is available to establish a change.
Save Widemile and check future changes in Following.
Up to 3 plugins free · no card required.
Compare saved observations
Download comparison JSONFull technical diff · 0 changed fields
Full snapshot data
{
"description": "Create one fixed high-flip transition where the camera rises and travels forward over the subject's head, pitches through a top-down zenith view, and directly reveals a different character and location on the far side, with gentle slow-motion accents before and after the switch; use for zenith scene switches, overhead character reveals, fluid music-video transitions, transizioni zenitali, cambi scena sopra il soggetto, or transiciones cenitales, not for ordinary crane shots, orbits, barrel rolls, improvised camera moves, hard cuts, or static slideshows.",
"included_files": [],
"name": "vfx-zenith-transition",
"skill_md_contents": "---\nname: vfx-zenith-transition\ndescription: Create one fixed high-flip transition where the camera rises and travels forward over the subject's head, pitches through a top-down zenith view, and directly reveals a different character and location on the far side, with gentle slow-motion accents before and after the switch; use for zenith scene switches, overhead character reveals, fluid music-video transitions, transizioni zenitali, cambi scena sopra il soggetto, or transiciones cenitales, not for ordinary crane shots, orbits, barrel rolls, improvised camera moves, hard cuts, or static slideshows.\n---\n\n# Zenith transition VFX\n\n## Tool requirements\n\n- **Core evidence and bindings:** `prostudio_brief_get`, `prostudio_workspace_get`, `prostudio_references_list`, `prostudio_references_resolve`, `prostudio_reference_create`, `get_prostudio_media_catalog`, and `inspect_prostudio_media`.\n- **Final transition clip:** `prostudio_models_list`, `prostudio_generation_prepare`, `prostudio_video_generate`, `prostudio_job_status`, and `prostudio_generation_get`.\n- **Conditional support asset:** `prostudio_image_generate` only when the user chooses to create a missing still reference before the final transition generation.\n- This declaration does not activate tools. Use only the tools needed for the selected asset mode. If a required tool is absent from the current scoped catalog and `load_prostudio_tools` is available, call it with the missing outcome before claiming the capability is unavailable. If retrieval still exposes no matching tool, report the limitation. A model-catalog lookup alone is never completion for a requested transition.\n\nCreate one continuous cinematic clip using the **fixed high flip** defined below. The camera rises and travels forward over the first subject's head while pitching toward a controlled top-down apex, switches directly from the origin to the destination at the zenith crossing, and completes the same movement to reveal the new character and location on the far side.\n\n**The high flip is the only permitted transition movement.** Do not substitute a crane reveal, orbit, drone rise, lateral sweep, whip pan, roll, zoom, dolly, spiral, static top shot, or any other camera move. The result must contain only origin A, the single continuous high-flip crossing, and destination B. Never invent a third scene, bridge environment, extra character, inserted object, transition tunnel, abstract interlude, VFX flourish, or generated B-roll between the two supplied sides.\n\nKeep skill selection semantic. Do not add keyword routing, deterministic activation rules, catalog architecture changes, or unrelated UI changes as part of this workflow.\n\n## Establish both sides of the transition\n\n- Resolve the **origin side** and **destination side** separately: selected subject, wardrobe and pose, location, time of day, action, emotional beat, camera-facing direction, and desired amount of environmental change.\n- Resolve the shared delivery constraints: duration, aspect ratio, frame rate, audio behavior, realism or stylization, intended musical beat, and whether the destination reveal should feel surprising, triumphant, ominous, romantic, or energetic.\n- The workflow accepts exactly these source combinations:\n 1. **Two images:** one origin image and one destination image.\n 2. **Two videos:** one origin clip and one destination clip.\n 3. **One video and one image:** either asset may be the origin or destination.\n- Load `media-inspection` and inspect every selected asset before planning. Filenames, thumbnails, prompts, and summaries are not evidence of identity, framing, motion, exposure, perspective, or transition compatibility.\n- Load `reference-management` and bind each asset to one explicit purpose: `origin identity/environment/action` or `destination identity/environment/action`. Never allow the two identities or locations to become interchangeable.\n- Ask only for missing information that materially changes the fixed high flip, identity, or result. If an origin or destination asset is missing, ask for it or offer to generate that missing source still when the user permits generation. A generated still may become origin A or destination B; it must never become an additional intermediary asset. Never invent or replace the selected person without the user's choice.\n\n## Handle each asset combination\n\n### Two images\n\n- Animate both sides as living shots. The camera move alone is not enough: each still must gain believable subject motion, environmental motion, and depth before or after the zenith seam.\n- Give the origin subject a restrained pre-transition action such as a breath, look, step, turn, gesture, fabric response, or hair movement. Give the destination subject a distinct reveal action that continues the energy of the move.\n- Build layered parallax from foreground, subject, midground, and background. Add only motivated secondary motion such as wind, practical lights, haze, passing shadows, reflections, particles, foliage, traffic, or crowd movement.\n- Preserve face, anatomy, wardrobe, and location structure while animating. Avoid frozen-photo holds, cardboard cutout parallax, rubber limbs, face drift, or a destination image that appears only as a static final frame.\n\n### Two videos\n\n- Inspect both clips for a compatible temporal window, subject orientation, motion vector, camera height, exposure, and frame cadence. Choose origin and destination moments that can accept the same camera momentum.\n- Preserve purposeful source action while regenerating the continuous transition. Do not simply concatenate the clips or hide a hard cut behind generic motion blur.\n- Bind both videos only when the selected live model and mode support the required video references. If the current model cannot accept both moving references, choose another compatible live model or ask the user for explicit anchor images; do not silently ignore one clip or claim a deterministic composite.\n- Treat source audio as reference material, not guaranteed final audio. Define whether the final clip keeps a music bed, uses a transition whoosh, lands on a downbeat, or remains silent according to the user's intent and live model support.\n\n### One video and one image\n\n- Preserve the video side's important action, screen direction, and motion cadence.\n- Animate the image side with subject motion, layered parallax, environmental response, and a readable reveal beat. It must not feel frozen beside the moving source clip.\n- Match motion energy across the seam. A calm video should not land in frantic generated motion, and a fast performance clip should not resolve into an inert portrait unless the contrast is explicitly intentional.\n\n## Execute the fixed high flip\n\n- Author the result as one clip and one uninterrupted camera gesture with exactly these phases:\n 1. **Origin setup:** establish origin A and its living action in a readable front-facing or three-quarter composition. Hold the first light, fluid slow-motion accent briefly before the move.\n 2. **Upward-forward rise:** accelerate upward and forward along the subject's sagittal centerline so the camera physically travels over the head. Keep the subject as the visual anchor; do not orbit, arc sideways, pull backward, circle, or replace the rise with a generic crane reveal.\n 3. **Forward overhead flip:** during that same travel, execute one controlled forward pitch around the camera's horizontal axis. Pass through a clear near-nadir, top-down zenith frame. The movement is a high flip, not a roll: suppress yaw drift and keep the optical axis, subject center, and travel direction aligned.\n 4. **Direct A-to-B crossing:** switch from origin A to destination B at the cleanest near-zenith/inverted moment. Use only framing correspondence and the motion blur naturally produced by the high flip. There is no third visual beat and no intermediary content.\n 5. **Destination resolution:** continue the same forward pitch, direction, speed curve, and angular momentum into destination B; restore an upright, readable composition and ease into the second light, fluid slow-motion accent.\n- The fixed path is therefore: **rise and travel forward over A -> one forward overhead pitch -> direct zenith switch -> continue the same movement into B**. It is never a menu of alternative moves.\n- Reproduce the reference's spatial signature: begin with the origin subject readable from the front or three-quarter view; reveal progressively more of the top of the head and ground as the camera climbs; let the ground or an existing scene plane fill the composition around the near-nadir apex; carry the same pitch beyond the subject; then let the far-side destination enter from the opposite frame edge. The subject remains the central anchor throughout instead of sliding sideways.\n- Use a single continuous forward pitch through and past the top-down crossing, with one acceleration, one apex, and one deceleration. Do not add a second flip, reverse the rotation, spin in place, roll on the lens axis, or combine the move with another transition grammar.\n- Lock the transition axis before generation. Match the origin and destination subject center, head/body axis, scale, screen direction, camera distance, and perspective closely enough that the same move can pass through both sides without a teleport.\n- Keep origin A intact until the zenith seam and destination B intact after it. A scene change is intentional; morphing is not. Do not blend faces, wardrobes, limbs, buildings, or color palettes into an intermediate hybrid.\n- Do not manufacture material to hide the switch. A ground, table, chair, wall, ceiling, or other broad surface that already exists in the selected sources may naturally fill the frame as a consequence of the high flip and provide the direct match point. Never add clouds, smoke, fabric, hair wipes, flares, signs, debris, foreground objects, black frames, white flashes, portals, tunnels, liquid warps, or abstract full-frame smears for transition coverage.\n- Finish with enough time for destination B to read. Do not spend the whole duration flipping or reveal B only in the final few frames.\n\n## Shape the slow-motion accents\n\n- Use gentle slow motion as a pair of short dramatic accents: one immediately before the upward-forward rise and one immediately after the destination reveal. The high-flip crossing between them should remain fluid and slightly more energetic so the speed contrast supports the transition.\n- Ease into and out of both accents with smooth temporal ramps. Preserve natural body mechanics, fabric and hair response, camera inertia, and continuous motion blur; the result must not feel like a freeze frame, a sudden frame-rate change, or a speed-ramp preset pasted over the clip.\n- Keep the slow motion subtle unless the user asks for a pronounced overcranked look. It should create a brief sense of suspension around each character beat without delaying the overhead move or consuming the destination reveal.\n- For video inputs, respect the source cadence and available model behavior. Treat slow motion as visual timing intent unless the tool reports real high-frame-rate or interpolation support; never claim technically overcranked footage without evidence.\n- For image-derived sides, create the slow-motion feeling through deliberately restrained subject action, environmental response, parallax, focus drift, and camera easing. Do not simulate it by holding the source image static.\n- When music is present, let the first slow-motion accent anticipate the transition beat, carry the fastest high-flip motion through the apex, and let the second accent land just after the reveal beat before returning to normal cadence.\n\n## Direct the cinematic capture\n\n- Specify camera height, angle, path, rotation axis, focal length, focus behavior, depth of field, foreground occlusion, frame rate, shutter or motion-blur feel, exposure, color temperature, contrast, and light direction.\n- Favor a 24–35 mm full-frame-equivalent lens for a strong but controlled sense of movement. Use a wider or longer lens only when the references and location geometry support it; avoid fisheye distortion and stretched faces unless requested.\n- Treat the move as a precisely controlled virtual-camera high flip: one upward-and-forward translation over the subject's head, locked to the sagittal centerline, combined with one forward overhead pitch. Keep parallax, perspective, and motion blur consistent with the claimed speed and distance. Do not reinterpret it as a Technocrane, jib, cable-cam, drone, orbit, handheld, roll, or freeform move. Avoid camera clipping through the subject.\n- When the user requests an ARRI look, express it as capture intent: ARRI ALEXA 35 or LF-style image, natural skin tones, controlled highlight roll-off, detailed shadows, restrained grain, realistic motion blur, and a coherent LogC-style grade. Do not claim a real camera body, lens, codec, or color pipeline unless the generation tool reports it.\n- Match exposure and dominant light direction across the apex enough to preserve flow, while allowing the destination palette to change deliberately. Use the seam to bridge the grade rather than allowing an accidental brightness flash or crushed transition frame.\n- For music-video work, time lift acceleration, apex, reveal, and destination pose to musical structure. The apex can carry a riser, reverse swell, breath, impact, or whoosh; the destination reveal should land on a defined beat when audio is part of the request.\n\n## Prompt strategy\n\n- Load `generation-planning`, read the live video-model catalog, and select a currently available mode compatible with the supplied combination of images and videos. Do not hardcode a provider or promise unsupported multi-reference behavior.\n- Write one spatially explicit English prompt for the complete clip. Preserve dialogue and visible text in the user's requested language.\n- Assign concrete time ranges that fit the duration. A useful default for an 8-second clip is approximately 0.0–2.0 seconds for the living origin beat and subtle slow-motion accent, 2.0–4.6 seconds for the upward-forward rise and single forward high flip with the direct zenith switch, and 4.6–8.0 seconds for the destination resolution, subtle slow-motion accent, and readable destination action. Rebalance only the timing; never change the transition movement.\n- Use this structure and replace every bracketed field:\n\n ```text\n ONE CONTINUOUS FIXED HIGH-FLIP ZENITH TRANSITION, no montage and no separately generated shots.\n ORIGIN LOCK: [origin identity, wardrobe, location, action, light, source asset and role].\n DESTINATION LOCK: [destination identity, wardrobe, location, action, light, source asset and role].\n CAMERA PACKAGE: [ARRI-style or other capture intent, lens, frame rate, shutter/motion blur, focus, grade, smooth temporal-ramp intent].\n FIXED MOVEMENT RULE: the only transition move is one upward-and-forward camera travel over the origin subject's head, locked to the subject centerline, combined with one controlled forward overhead pitch through a clear near-nadir zenith frame. No orbit, roll, lateral arc, whip pan, zoom, drone pullback, alternate camera move, or second flip.\n PHASE 1 [start–end]: [living origin action] enters a light, fluid slow-motion accent; camera holds [start height and framing], keeps the subject centered, then accelerates upward and forward over the subject's head with [subject/environment animation and parallax]. Progressively reveal the top of the head and the ground plane exactly as the camera pitch increases.\n PHASE 2 [start–end]: while continuing the rise, camera performs one forward high flip around its horizontal axis, reaches a clear top-down zenith frame, and switches directly from origin A to destination B at [exact seam time]. Preserve [matched subject center, body axis, scale, screen direction, perspective, exposure, and natural motion blur]. No intermediary imagery or invented transition element.\n PHASE 3 [start–end]: continue the exact same forward pitch, direction, and angular momentum into [destination location], restore [upright horizon and framing], reveal [destination character and action], then ease into a second light, fluid slow-motion accent and settle long enough for B to read. Animate [subject and environment] when the destination source is a still image.\n AUDIO CONTINUITY: [music bed/source ambience], [riser or whoosh at apex], [reveal beat or impact], [dialogue placement if any].\n ```\n- State positive motion for every image-derived side: subject action, hair or wardrobe response, environmental movement, layered parallax, and focus evolution. Explicitly forbid a frozen still with camera-only movement.\n- Exclude hard cuts, dissolves, clip concatenation, teleporting viewpoints, repeated subjects, face or wardrobe morphing, hybrid locations, broken anatomy, camera-body intersections, sudden focal-length jumps, fisheye warping, roll or yaw drift, arbitrary full-frame blur, exposure pops, a horizon that never resolves, and destination reveals too brief to read.\n- Explicitly forbid any content between A and B: no third scene, intermediate environment, extra person, duplicate subject, generated prop, bridge shot, B-roll, portal, tunnel, particle burst, cloud bank, smoke wipe, fabric wipe, flare wipe, abstract texture, or decorative transition effect. The direct high-flip apex is the transition.\n\n## Generate and inspect\n\n- Prepare exactly one video generation plan and execute exactly one `prostudio_video_generate` action for the finished result. Never generate the two sides as separate shots and never use the Editor to assemble them into the effect.\n- Report the selected model, supported reference mode, bound origin and destination assets, duration, aspect ratio, audio behavior, and credit charge before the host's required confirmation. After approval, execute the prepared action without changing its arguments.\n- Poll the durable job and inspect the completed video before calling the transition successful. A completed status is not proof of visual continuity.\n- If a generated still support asset is necessary, create only that approved support asset first, inspect it, register or bind the exact durable media when needed, and then prepare the single final video generation.\n\n## Preflight checklist\n\nBefore generation, verify that:\n\n- there are exactly two clearly assigned sides and the origin/destination direction is explicit;\n- the asset mode is image–image, video–video, or mixed, and the live model supports the required bindings;\n- every image-derived side includes subject animation, environmental animation, depth, and parallax rather than a static hold;\n- both identities, wardrobes, locations, and actions are separately locked;\n- the camera path is the fixed high flip: centerline upward-forward travel over the subject's head, one forward overhead pitch, clear near-nadir apex, direct A-to-B switch, continuation of the same movement, and final settling composition;\n- no alternate camera movement, roll, yaw drift, orbit, side arc, zoom, drone pullback, whip pan, second flip, or reverse rotation appears in the plan;\n- the switch is direct and contains no third scene, intermediary imagery, invented object, extra character, bridge shot, portal, tunnel, decorative wipe, or generated transition flourish;\n- camera momentum, center of attention, frame cadence, exposure, and focus evolve coherently through the apex;\n- short slow-motion accents appear before the upward-forward rise and after the reveal, with smooth ramps, natural motion blur, and no frozen frames or abrupt cadence changes;\n- the destination receives enough screen time to communicate the new character and location;\n- the plan contains one generation and one final clip, not two generated shots or an Editor assembly;\n- the prompt contains no placeholders and forbids morphing, clipping, frozen stills, hard cuts, and an unresolved horizon.\n\n## Review and revision criteria\n\n- **Pass:** the camera begins in a readable origin shot, rises and travels forward over the subject's head along the centerline, performs one controlled forward overhead pitch through a clear near-nadir apex, switches directly from A to B there, and continues the same movement into a distinct destination character and location.\n- **Pass:** the high flip is unmistakably the only transition movement; its axis stays locked without orbit, roll, side sweep, reverse rotation, or a second flip.\n- **Pass:** the apex contains only the direct transition from the supplied origin to the supplied destination, with no invented visual content between them.\n- **Pass:** image inputs are visibly animated with believable human, environmental, and depth motion on their respective side of the transition.\n- **Pass:** the destination identity and environment appear cleanly after the seam, without a blended intermediate person or hybrid location.\n- **Pass:** the final horizon, framing, focus, exposure, and camera deceleration feel intentional and cinematic.\n- **Pass:** the origin and destination character beats receive subtle, fluid slow-motion accents while the overhead crossing retains continuous momentum.\n- **Fail:** the result reads as two clips joined by a hard cut, dissolve, generic blur, or teleport.\n- **Fail:** the camera uses any transition movement other than the fixed high flip, including a crane-only reveal, orbit, lateral arc, whip pan, zoom, dolly, drone rise or pullback, lens-axis roll, spiral, or static overhead shot.\n- **Fail:** any third scene, bridge environment, extra subject, inserted object, transition tunnel, portal, smoke/flare/fabric wipe, particle effect, abstract interlude, or generated B-roll appears between A and B.\n- **Fail:** either image side remains a frozen photo, only the camera moves, or parallax reads as layered cardboard.\n- **Fail:** the camera spins in place, rolls on the lens axis, drifts in yaw, clips through the subject, loses the visual anchor, never reaches a clear zenith view, reverses direction, adds another flip, or does not settle after the reveal.\n- **Fail:** slow motion becomes a frozen hold, stutters, creates optical-flow anatomy artifacts, starts or stops abruptly, or makes the zenith crossing feel sluggish.\n- **Fail:** identities morph, destination wardrobe leaks into the origin, locations blend before the seam, or the new character appears as a duplicate rather than a deliberate replacement.\n- For a localized failure, revise only the responsible timing, binding, axis lock, match, or high-flip constraint, then regenerate the complete single clip. Do not solve a failure by changing the movement, adding transition content, generating separate halves, or assembling multiple shots.\n"
}SHA-256 of public snapshot: 2ffd30dbdf60bb1553bd8e172078d21ab30ce6015a2cf881aced7c2127b732c0