DepthForge — AI Image & Video to 3D

Voyde Procedural in Modeling


OFFICIAL USER DOCUMENTATION

DEPTHFORGE

AI Image & Video to 3D Relief

VERSION 0.3.0  ·  BLENDER 4.2+

Everything you need to install DepthForge, generate image and video depth, customize procedural 3D reliefs, and export your results.

SECTION 01

Installation & Setup

DepthForge comes with its AI models, ONNX Runtime, and video decoding components already bundled. No additional dependency installation is required.

Choose the Correct Package

Download the ZIP matching both your operating system and Blender version.

Platform Blender 4.2–5.0 Blender 5.1+
Windows x64 Python 3.11 Python 3.13
macOS Apple Silicon Python 3.11 Python 3.13
macOS Intel Python 3.11 Python 3.13
Linux x64 Python 3.11 Python 3.13

Installation Steps

STEP 01 — Open Preferences

Launch Blender and navigate to Edit → Preferences → Get Extensions.

STEP 02 — Install from Disk

Select Install from Disk and choose your DepthForge ZIP package. Do not extract the ZIP beforehand.

STEP 03 — Open the Add-on

In the 3D Viewport, press N to open the sidebar and select the DepthForge tab.

✓ No API key  ·  No internet  ·  No administrator access  ·  No extra downloads

SECTION 02

Image to 3D Workflow

Convert photographs, artwork, and heightmaps into editable 3D relief geometry.

  1. Open the DepthForge sidebar.
  2. Select Image as the source type.
  3. Choose your source image.
  4. Select the desired depth estimation mode.
  5. Adjust the Detail setting.
  6. Choose Horizontal (XY) or Vertical (XZ) orientation.
  7. Generate your 3D relief.

Image Processing Modes

AI DEPTH — QUALITY

Uses Depth Anything V2 Small to estimate scene depth from visual features. Recommended for photographs and detailed scenes.

AI DEPTH — FAST

Uses a compressed inference model for faster processing while retaining similar depth structure.

LUMINANCE

Maps image brightness directly to height without AI inference. Suitable for logos, grayscale heightmaps, patterns, and artistic displacement.

TIP: For grayscale terrain maps, use Luminance Mode. For photographic depth estimation, start with AI Depth — Quality.

Image Detail

Supported processing detail: 252–1512 px. Default: 518 px. Increasing Detail can improve depth-map resolution but requires more processing resources.

Updating an Existing Relief

Use the depth update command to process a new source image and update an existing relief without recreating its entire procedural setup.

SECTION 03

Video to Animated 3D

DepthForge generates animated depth geometry from video footage or numbered image sequences.

Supported Inputs

Video: MP4, MOV, MKV, AVI, WebM, M4V, MPG, GIF, MXF

Sequences: PNG, JPG, TIFF, EXR, BMP, WebP, TGA

Video decoding also depends on the underlying codec and encoding.

Processing Workflow

  1. Select Video as the source type.
  2. Choose your video file or image sequence.
  3. Select Video Depth Anything or Per Frame.
  4. Set First Frame and Frames to select the desired clip range.
  5. Adjust Detail according to your system resources.
  6. Start processing and monitor progress.
  7. Play the Blender timeline to preview your animated 3D relief.
VIDEO DEPTH ANYTHING

Uses a temporal depth estimation model to process consecutive video frames. It operates with 32-frame windows, improving consistency across time compared with independent frame estimation.

PER FRAME MODE

Processes frames individually using an image model, with consistent depth scaling and motion-aware temporal smoothing.

Temporal Smooth: Default 0.6.

Background Processing

Video processing runs as a background task. Blender remains responsive, with a progress indicator, estimated remaining time, and Cancel or ESC control.

Depth Sequence Output

DepthForge saves the processed result as a 16-bit PNG sequence.

depth_0001.png
depth_0002.png
depth_0003.png
...

Default output location:

//DepthForge/<clip_name>_depth

For unsaved projects, DepthForge uses its user data directory. Existing output directories are not overwritten.

Timeline Synchronization

Use Start Frame and Frames to adjust relief timing. Optionally synchronize Blender's scene range and frame rate with the processed clip.

SECTION 04

Geometry Nodes Controls

DepthForge generates a procedural relief with live, non-destructive controls.

Control Description
Height Adjusts relief depth and elevation.
Size Controls the overall relief dimensions.
Resolution Defines grid vertex density.
Black Point Adjusts the lower depth range.
White Point Adjusts the upper depth range.
Gamma Controls the depth response curve.
Smooth Softens abrupt depth variations.
Invert Reverses near and far depth direction.
Cutoff Removes selected depth regions.
Stretch Cut Removes elongated faces along depth discontinuities.
Thickness Adds side walls and a flat back/base.

Recommended workflow: Start with moderate Resolution, adjust Height, refine the depth range using Black/White Point and Gamma, then apply Smooth and Cutoff if needed.

SECTION 05

Creative Presets

PHOTO 2.5D

For photographic depth surfaces and 2.5D scenes.

BAS-RELIEF

For shallow sculptural and embossed effects.

CUTOUT

For depth-masked geometry and background removal.

TERRAIN

For landscape-style geometry and heightmap workflows.

Presets provide starting configurations. All Geometry Nodes settings remain adjustable after selection.

SECTION 06

Export & Conversion

EXPORT DEPTH

Save image-based depth as a 16-bit PNG heightmap. Supported depth transformations can be incorporated into the export.

CONVERT TO MESH

Convert the Geometry Nodes relief into regular Blender mesh geometry. For video reliefs, the currently displayed frame is converted.

OPEN DEPTH FOLDER

Open the directory containing your processed video depth frames.

TIP: Duplicate your relief before using Convert to Mesh if you want to preserve the procedural version.

SECTION 07

Performance & Requirements

AI processing runs locally on your CPU. A dedicated GPU is not required.

Operation Detail Measured Time
Image — Quality 518 0.6–1.4 sec
Image — Fast 518 0.5–1.3 sec
Video Depth Anything 252 / 16:9 0.5 sec/frame
Video Depth Anything 518 / 16:9 3.0 sec/frame

Benchmark measurements were collected in a 2-core test environment. Actual performance varies by hardware, source media, model configuration, and processing settings. These are not guaranteed processing times.

Memory Recommendations

High-resolution video processing can require approximately 4–5 GB of available memory. Detail values above 448 display a memory warning in the add-on.

macOS: On Apple Silicon, direct video decoding requires macOS 14 or newer. Use supported numbered image sequences on older versions.

SECTION 08

Troubleshooting

The add-on does not appear in the sidebar.

Confirm you installed the ZIP that matches your Blender version, operating system, and architecture. Check Blender Preferences and verify that DepthForge is enabled.

My depth map looks incorrect.

Try changing the AI model or adjusting Black Point, White Point, Gamma, Smooth, and Invert. AI depth estimation cannot perfectly interpret every scene.

Video processing is too slow.

Reduce Detail, select a shorter frame range, or use Per Frame processing where appropriate. Processing speed depends primarily on CPU performance and memory availability.

My video cannot be opened.

Try another compatible codec or convert the video into a numbered PNG or JPG sequence.

The mesh is too dense or Blender becomes slow.

Reduce the Geometry Nodes Resolution setting. Higher grid densities can consume significant memory and reduce viewport responsiveness.

The animation timing is incorrect.

Check Start Frame, Frames, scene FPS, and your scene frame range. Enable scene synchronization if appropriate for your workflow.

SECTION 09

Limitations & Licensing

Important Limitations

  • AI models estimate relative depth, not real-world metric distances.
  • DepthForge produces relief surfaces, not fully reconstructed 360-degree objects.
  • Reflective materials, transparent objects, heavy occlusion, and unusual lighting may affect depth quality.
  • Temporal processing improves consistency but does not guarantee flicker-free output for every video.
  • High-resolution video processing can consume substantial CPU time and RAM.

Third-Party Components

Component License
DepthForge Add-on Code GPL-3.0-or-later
Depth Anything V2 Small Apache-2.0
Video Depth Anything Small Apache-2.0
ONNX Runtime MIT
PyAV BSD-3-Clause
Bundled FFmpeg Components See bundled license notices

Refer to the licenses directory and THIRD_PARTY_NOTICES.txt included in the product package for complete license and attribution information.

Privacy & AI Processing

Depth estimation is performed on your own computer using locally bundled AI models. Your source images and videos are not uploaded to an online AI service.

CREATE BEYOND 2D.

Images. Videos. Depth. Geometry.

DEPTHFORGE · VERSION 0.3.0
$14.99$10.49

Have questions about this product?
Login to message

Details
Blender Extension Compatible Yes
Published about 3 hours ago
Blender Version 4.2 - 5.2
Extension Type Add-on
Render Engine Used Cycles, Eevee
License GPL

Voyde Procedural

Visit Website See all items by this creator
Discover more products like this

Bas Relief 2.5d Heightmap ai animation displacement video to 3d Image-to-3D depth-estimation geometry nodes procedural depth map