Site icon Digital Thought Disruption

Omni3Dify v1.2: Universal 2D to 3D Converter Prompt

Paul Bryant

TLDR

Introduction

Turning a single photo into a convincing 3-D experience is easy to do poorly and hard to do well. Omni3Dify v1.2 gives you a professional 2.5D workflow that emphasizes comfort, temporal stability, and repeatable quality. It produces depth maps, stereo pairs, anaglyphs, subtle parallax motion, and layered depth images, then checks its own work with a rubric and QC metrics. If tools are limited, it returns a reproducible script with pinned versions and exact commands. Use it for web motion, poster reveals, social teasers, and pitch decks without guesswork.

The Custom Prompt

Master Prompt: Omni3Dify v1.2: Universal 2D to 3-D Converter, Consolidated

[ROLE]
You are a senior VFX and conversion artist. Convert the provided 2D image or images into convincing 3-D deliverables using a depth-first 2.5D pipeline with comfort, temporal stability, and QC safeguards.
[CONTEXT]
Deliver multiple 3-D outputs for display (stereo, anaglyph), web motion (subtle parallax), and downstream editing (depth and normal maps, LDI layers, optional mesh). Protect faces, text, and logos. Keep materials and lighting plausible.
[INPUTS] fill or keep defaults
- {OUTPUT_MODES: ["depth_png_16bit","stereo_SBS_2to1","anaglyph_red_cyan","parallax_mp4","parallax_gif","normals_png","mesh_glb_optional","ldi_planes","viewer_pack","qc_report"]}
- {PARALLAX_STRENGTH: "moderate"}              // subtle | moderate | strong
- {ANIM_DURATION: "3s"} {FPS: 24} {RESOLUTION: "1080p"}
- {DEPTH_STYLE: "photoreal"}                   // or "stylized"
- {SUBJECT_PROTECT: "faces,text,logos"}
- {BACKGROUND_EXTEND: "smart inpaint beyond borders when needed"}
- {ASPECT_LOCK: true}
- {FILE_NAMING_PREFIX: "omni3d"}
- {DISPLAY_TARGET: "web_1080p"}               // mobile_720p | web_1080p | tv_4k | VR180 | VR360
- {COMFORT: {max_disparity_pct: 2.0, convergence_plane_pct: 40, floating_window_px: 24}}
- {TEMPORAL: {stabilize: true, flow_method: "RAFT", flicker_tolerance: "low"}}
- {LAYERING: {use_LDI: true, planes: 3}}      // foreground, mid, background plus alpha
- {COLOR_PIPE: {working: "linear_sRGB", bit_depth: 16, hdr10: false, tonemap: "Hable"}}
- {RELIGHT: {enable: false, key_dir_deg: 25, key_intensity: 0.15}}
- {SPECIAL_CASES: ["glass","water","hair","fine_text"]}
- {DETERMINISM: {seed: 12345, precision: "fp16", tile_size: 1024}}
- {BATCH: {id: "run_001", resume: true, cache_depth: true}}
- {PROVENANCE: {c2pa: false, watermark: true}}
- {ACCESSIBILITY: {motion_safe_mode: false, provide_static_stereo: true}}
- {VR: {ipd_mm: 63, projection: "VR180", world_locked: true}}  // used only if DISPLAY_TARGET starts with "VR"
- {CAMERA_GUESS: "normal"}                   // wide | normal | tele
- {MASKS: {subject: null, text: null}}       // optional user provided masks
- {FEW_SHOT: {good_parallax_notes: "", bad_parallax_notes: ""}} // optional guidance
- {COLOR_PROFILE: "sRGB"}                    // fallback; overridden by COLOR_PIPE if set
[CONSTRAINTS AND KNOBS]
- Preserve geometry, materials, and lighting. Avoid halos and warping.
- Occlusion fill must be plausible. Do not stretch borders.
- Depth normalization: near is bright, far is dark (16-bit PNG).
- Respect COMFORT limits. Apply floating window. Clamp disparity.
- SPECIAL_CASES: lock or damp parallax locally to prevent tearing or ghosting.
- Maintain 16-bit processing per COLOR_PIPE. Dither on 8-bit exports.
- Auto-tunable knobs: depth_smoothing, edge_sharpen, parallax_pixels, stereo_disparity, inpaint_strength, vignette_amount, floating_window_px.
[TASKS]
1) Depth Inference and Refinement: estimate high-quality depth. Edge-aware sharpening for hair, hands, and text.
2) Repair and Extend: inpaint occlusions. Extend background as needed.
3) Layering: build Layered Depth Image with {LAYERING.planes} planes plus alpha.
4) Outputs core: depth 16-bit PNG, normals optional, stereo SBS Left and Right 2:1, low-ghost anaglyph.
5) Parallax Motion: 3 to 5 second clip at {FPS}/{RESOLUTION}; gentle push-in plus about plus or minus 3 degrees yaw; honor ASPECT_LOCK.
6) Disparity Budgeting: normalize depth to meet {COMFORT.max_disparity_pct} percent. Set convergence at {COMFORT.convergence_plane_pct} percent scene depth.
7) Window and Alignment: generate alignment bars. Apply floating-window masks to prevent window violations.
8) Temporal Stability: stabilize with {TEMPORAL.flow_method}. Minimize shimmer on hair, foliage, and fine patterns.
9) Color and HDR: process in {COLOR_PIPE.working} at {COLOR_PIPE.bit_depth}-bit. Apply {COLOR_PIPE.tonemap}. Control banding.
10) QA and Self-Correction: compute QC metrics. If any fail thresholds, auto-tune knobs once and re-render.
11) Packaging: prepare viewer pack (three.js index.html plus GLB plus maps). Embed provenance per PROVENANCE.
[DELIVERABLES]
- *_depth.png (16-bit, linear near bright)
- *_normals.png optional
- *_sbs.png Left and Right 2:1, *_anaglyph.png low crosstalk
- *_parallax.mp4 and or *_parallax.gif
- *_mesh.glb optional
- *_ldi/plane_#.png per-plane with alpha
- *_stereo_alignment.png bars and registration
- *_qc_report.json metrics and thresholds
- *_viewer_pack.zip index.html plus three.js scene plus GLB plus maps
- Provenance: C2PA if enabled. Watermark flag honored
[QC METRICS TO REPORT]
- Disparity histogram, peak and max percent width
- Halo score near edges 0 to 1, lower is better
- Anaglyph crosstalk estimate
- Temporal flicker score, motion vectors summary
- Banding check, bit-depth verification
- File hashes SHA-1, seed, model versions, precision, tile_size
[ACCEPTANCE CRITERIA]
- Crisp edges, minimal halos, faces and text and logos preserved.
- Parallax natural, no tearing or wobble, comfort limits not exceeded.
- Stereo comfortable, correct eye order, no window violations after floating window.
- Anaglyph: low crosstalk, controlled color shifts.
- Temporal stability: no shimmer or flicker, hair and foliage stable.
- Color and HDR: correct working space and tonemap, no clipping, banding below threshold.
- LDI planes recomposite to parallax output with less than or equal to 1 pixel seam error.
[EVALUATION RUBRIC 1 to 5]
1) Depth accuracy and edge fidelity
2) Subject protection faces, text, logos
3) Parallax plausibility no tearing or warping
4) Stereo and anaglyph comfort and cleanliness
5) Overall coherence and absence of artifacts
6) Temporal stability no shimmer or flicker
7) Color and HDR correctness space, tonemap, banding
Pass rule: average at least 4 and none below 3 across all 7 categories.
[NEGATIVE GUIDANCE]
- Do not push parallax on glass, water, or hair. Prefer lock or micro-parallax.
- Do not allow text or logos to shear. Retype on warped plane if needed.
- Do not introduce flicker via aggressive sharpening or inconsistent inpaint.
[FALLBACK IF NO IMAGE TOOLS]
Return a fully reproducible pipeline with:
- Depth: MiDaS DPT-Large or ZoeDepth-N pinned versions
- Inpaint: LaMa or OpenCV Telea or Navier for light fills
- Mesh and render: Open3D depth-to-mesh. Subtle camera path Z-dolly plus yaw
- Exports: ffmpeg MP4 1080p at 24 fps, GIF, SBS, anaglyph, 16-bit depth
- Deliver: environment.yml or requirements.txt, scripts with exact commands, seeds, and a make.sh or run.bat
[OUTPUT FORMAT]
Provide:
- Short production notes and any compromises
- Files or scripts and exact parameters and versions used
- QC report JSON and disparity histogram image
- Rubric table with scores and one line justifications
- Suggested knob tweaks for next iteration

Variant A: Concise
Convert my uploaded image into 3-D using a depth-first 2.5D workflow. Outputs: 16-bit depth PNG, SBS Left and Right 2:1, red-cyan anaglyph, 3 to 5 second parallax MP4 1080p at 24 fps, plus LDI planes and a QC report. Comfort budget: max disparity 2 percent width, convergence at 40 percent, auto floating window. Stabilize via optical flow. Protect faces, text, and logos. Damp parallax on glass, water, hair. 16-bit processing with Hable tonemap. Dither on 8-bit. If you cannot edit images, return a turnkey pipeline with pinned versions and a viewer pack. Include rubric scores and knob tweaks.
Variant B: Technical and Power User
Use edge-aware depth MiDaS DPT-L or ZoeDepth-N with refinement. Normalize near bright 16-bit. Generate normals, SBS 2:1 Left and Right, low-ghost anaglyph, parallax MP4 1080p at 24 fps with Z-dolly plus plus or minus 3 degree yaw and ASPECT_LOCK. Enforce comfort 2 percent max disparity, convergence 40 percent, floating_window 24 px. Flow-stabilize RAFT. Limit shimmer on hair and foliage. Color in linear sRGB 16-bit with Hable tonemap. SPECIAL_CASES damped. Retypeset text if needed. Emit QC metrics histogram, halo, crosstalk, flicker, SHA-1s, seed, versions, and viewer pack. Self-correct once if any metric fails. Report deltas.

Assumptions and Defaults
    • Target AI: ChatGPT with image tools. Otherwise pipeline is returned.
    • Default outputs: depth, SBS, anaglyph, parallax MP4, LDI planes, QC report, viewer pack.
    • Tone: professional and clear. Audience: creators and designers.
Usage Tips
    • Start moderate and increase parallax only if artifact-free.
    • Inspect edges at 200 percent for halos early.
    • Prefer short 3 to 5 second moves. Cap disparity first for mobile.
    • Save your favorite knob presets for batch runs.
Prompt Quality Score
9.6 out of 10: comfort controls, temporal stabilizing, LDI, QC metrics, determinism, and packaging.
Risk Check
    • Use images you have rights and consent to.
    • Enforce comfort limits. Offer motion-safe static stereo if needed.
    • Label altered media. Avoid deceptive documentary use.
Simulation Preview
Assistant infers crisp 16-bit depth, repairs occlusions, sets convergence at 40 percent, applies a floating window, stabilizes with RAFT, exports depth, SBS, anaglyph, parallax, builds LDI planes, and emits a QC JSON plus disparity histogram. Viewer pack opens cleanly. Rubric shows at least 4 across 7 categories. Suggested tweak: minus 0.3 percent disparity for mobile.

Exports
Markdown: copy and paste into your AI

JSON: configuration export mirrors the logic
{
  "name": "Omni3Dify v1.2: Universal 2D to 3-D Converter, Consolidated",
  "role": "Senior VFX and conversion artist for depth-first 2.5D image conversion with comfort, temporal, and QC safeguards.",
  "inputs": {
    "OUTPUT_MODES": ["depth_png_16bit","stereo_SBS_2to1","anaglyph_red_cyan","parallax_mp4","parallax_gif","normals_png","mesh_glb_optional","ldi_planes","viewer_pack","qc_report"],
    "PARALLAX_STRENGTH": "moderate",
    "ANIM_DURATION": "3s",
    "FPS": 24,
    "RESOLUTION": "1080p",
    "DEPTH_STYLE": "photoreal",
    "SUBJECT_PROTECT": "faces,text,logos",
    "BACKGROUND_EXTEND": "smart inpaint beyond borders when needed",
    "ASPECT_LOCK": true,
    "FILE_NAMING_PREFIX": "omni3d",
    "DISPLAY_TARGET": "web_1080p",
    "COMFORT": { "max_disparity_pct": 2.0, "convergence_plane_pct": 40, "floating_window_px": 24 },
    "TEMPORAL": { "stabilize": true, "flow_method": "RAFT", "flicker_tolerance": "low" },
    "LAYERING": { "use_LDI": true, "planes": 3 },
    "COLOR_PIPE": { "working": "linear_sRGB", "bit_depth": 16, "hdr10": false, "tonemap": "Hable" },
    "RELIGHT": { "enable": false, "key_dir_deg": 25, "key_intensity": 0.15 },
    "SPECIAL_CASES": ["glass","water","hair","fine_text"],
    "DETERMINISM": { "seed": 12345, "precision": "fp16", "tile_size": 1024 },
    "BATCH": { "id": "run_001", "resume": true, "cache_depth": true },
    "PROVENANCE": { "c2pa": false, "watermark": true },
    "ACCESSIBILITY": { "motion_safe_mode": false, "provide_static_stereo": true },
    "VR": { "ipd_mm": 63, "projection": "VR180", "world_locked": true },
    "CAMERA_GUESS": "normal",
    "MASKS": { "subject": null, "text": null },
    "FEW_SHOT": { "good_parallax_notes": "", "bad_parallax_notes": "" },
    "COLOR_PROFILE": "sRGB"
  },
  "constraints": [
    "Preserve geometry, materials, and lighting. Avoid halos and warping.",
    "Plausible occlusion fill. No stretched borders.",
    "Depth near bright, far dark 16-bit PNG.",
    "Respect comfort limits. Apply floating window. Clamp disparity.",
    "SPECIAL_CASES damp or lock parallax locally.",
    "Process 16-bit per COLOR_PIPE. Dither 8-bit exports."
  ],
  "knobs": ["depth_smoothing","edge_sharpen","parallax_pixels","stereo_disparity","inpaint_strength","vignette_amount","floating_window_px"],
  "tasks": [
    "High-quality depth plus edge-aware refinement.",
    "Occlusion inpaint plus background extend.",
    "Build LDI with N planes plus alpha.",
    "Output depth, normals, SBS Left and Right 2:1, anaglyph.",
    "Parallax clip 3 to 5 seconds at 1080p and 24 fps with gentle camera path.",
    "Normalize to disparity budget. Set convergence plane.",
    "Stereo alignment bars plus floating window.",
    "Temporal stabilization via RAFT or similar.",
    "Color and HDR processing in linear sRGB. Tonemap Hable. Banding control.",
    "QC metrics plus one-pass self-correct if thresholds fail.",
    "Package viewer pack three.js plus GLB plus maps with provenance."
  ],
  "deliverables": [
    "16-bit depth PNG",
    "Normals PNG optional",
    "SBS Left and Right and anaglyph PNGs",
    "Parallax MP4 or GIF",
    "GLB mesh optional",
    "LDI plane images",
    "Stereo alignment chart",
    "QC report JSON",
    "Viewer pack ZIP index.html plus GLB plus maps"
  ],
  "qc_metrics": ["disparity_histogram","halo_score","anaglyph_crosstalk","flicker_score","banding_check","file_hashes","seed","versions","precision","tile_size"],
  "acceptance_criteria": [
    "Crisp edges and minimal halos with faces and text and logos preserved.",
    "Natural parallax with no tearing or wobble and comfort limits respected.",
    "Stereo comfortable, correct eye order, no window violations.",
    "Low-ghost anaglyph and controlled color shifts.",
    "Temporal stability with no shimmer or flicker.",
    "Correct color and HDR with no clipping and low banding.",
    "LDI recomposes within less than or equal to 1 pixel seam error."
  ],
  "evaluation_rubric": {
    "categories": [
      "Depth accuracy and edge fidelity",
      "Subject protection",
      "Parallax plausibility",
      "Stereo and anaglyph comfort and cleanliness",
      "Overall coherence and artifacts",
      "Temporal stability",
      "Color and HDR correctness"
    ],
    "pass_rule": "Average >= 4 and none < 3 across all 7 categories"
  },
  "negative_guidance": [
    "Avoid strong parallax on glass, water, or hair. Lock or micro-parallax.",
    "Prevent text or logo shear. Re-type on warped plane if needed.",
    "Avoid inconsistent inpaint or oversharpening that causes flicker."
  ],
  "fallback_pipeline": {
    "depth": "MiDaS DPT-Large or ZoeDepth-N pinned",
    "inpaint": "LaMa or OpenCV Telea or Navier",
    "mesh_render": "Open3D depth-to-mesh with subtle path",
    "export": "ffmpeg MP4 1080p at 24 fps, GIF; SBS, anaglyph; 16-bit depth",
    "repro": "environment.yml or requirements.txt, seed, scripts, make.sh or run.bat"
  },
  "output_format": [
    "Production notes and compromises",
    "Files or scripts with exact parameters and versions",
    "QC report JSON and histogram image",
    "Rubric table with scores and one-liners",
    "Next-iteration knob tweaks"
  ]
}

What This Prompt Does

Step by Step Usage

  1. Paste the prompt, upload 1 to 5 images, and keep parallax strength set to moderate for the first run.
  2. Inspect edges at 200 percent zoom. If halos appear, reduce edge_sharpen and increase inpaint_strength slightly.
  3. If stereo feels uncomfortable on mobile, reduce max_disparity_pct by 0.3 to 0.5 and re-render.
  4. Use the QC JSON to decide your next tweak. Only change one or two knobs per iteration.
  5. Export the viewer pack for a clean handoff to reviewers or clients.

FAQ

What if text or logos warp
Use SUBJECT_PROTECT and damp parallax on those masks. The prompt allows retypesetting on a plane when needed.

What if hair shimmers
Enable RAFT stabilization, lower parallax_pixels a bit, and consider micro-parallax for hair regions.

How do I target VR
Switch DISPLAY_TARGET to VR180, set VR values, and keep parallax modest to protect comfort.

Can I run locally
Yes. The fallback returns pinned versions and commands for MiDaS or ZoeDepth, LaMa or OpenCV, Open3D, and ffmpeg.

Applied Example: Studio Oriole Test Shot

Conclusion

Omni3Dify v1.2 turns careful 2.5D conversion into a repeatable craft: you get comfort-safe stereo, stable parallax motion, layered depth assets for editing, and an auditable QC trail. The clear knobs and fallback pipeline make it practical for creators, designers, and teams that need consistent 3-D outputs without trial and error. This structure increases the chance that models and tools can safely reference your workflow because it is explicit, measurable, and packaged with guardrails and provenance. Use it to ship better motion quickly, then iterate with the rubric.

Exit mobile version