fal.ai Official Blender Extension

AI-powered 3D, materials, rendering, video, and audio — directly inside Blender.

Quick Start

Install the addon (see Installation) - supports Blender 4.2 and up on all platforms
Enable fal.ai — AI Generation Suite in Edit → Preferences → Add-ons
Set your fal.ai API key in the addon preferences
Open the sidebar (N) in either the 3D Viewport or the Video Sequence Editor and click the fal.ai tab
Pick a workflow, configure its parameters, and click Generate

Jobs run in the background — keep working and they'll appear in the Active Jobs sub-panel. Multiple jobs can run at once.

The fal.ai Blender extension preferences window.

3D Panel Workflows

Available in the 3D Viewport sidebar under the fal.ai tab.

3D Generation

fal_extension_3d_generation.mp4

Generates 3D models from image or text and imports the resulting mesh at the 3D cursor as GLB (falls back to OBJ + MTL + texture when needed).

Text-to-3D

Generate a textured 3D mesh from a prompt. Exposes endpoint-specific controls (face budget, symmetry, pose hints, art style, quad vs triangle topology, etc.) conditionally based on which endpoint you pick.

Endpoint	Notes
Meshy v6 Preview	Fast geometry + texture, 100–300k face budget, pose / symmetry hints
Hunyuan 3D v3.1 Pro	High-fidelity, 40k–1.5M faces, optional geometry-only output
Hunyuan 3D v3.1 Rapid	Quick turnaround, geometry-only toggle
Tripo P1	Low-poly friendly, 48–20k face budget
Tripo H3.1	Quad topology option, real-world auto-sizing, separate texture-seed

Image-to-3D

Same endpoints, but conditioned on a source image (file on disk or current render result). Prompt is still available as an optional guide.

Endpoint	Notes
Meshy v6 Preview	Pose / symmetry hints, texture prompt
Hunyuan 3D v3.1 Pro	Geometry-only option
Hunyuan 3D v3.1 Rapid	Geometry-only option
Tripo P1	Low-poly
Tripo H3.1	Orientation + texture-alignment controls for image-to-3D

Materials

fal_extension_materials.mp4

Produces a full Principled BSDF material (base color, roughness, metalness, normal, displacement) applied to the selected object.

Text-to-Material

Generate a complete tiling PBR material from a text prompt.

Endpoint	Notes
PATINA Material	Generates base color + all PBR maps from a prompt

Image-to-Maps

Estimate PBR maps (roughness, normal, displacement, metalness) from an existing base-color image. Useful when you already have a texture and want the rest of the stack.

Endpoint	Notes
PATINA	Estimates PBR maps from a supplied base-color image

There's also a Tiling Texture sub-mode if you just want a seamless base color from a prompt without the PBR stack.

Image Rendering

Each mode does a quick technical render of the scene (depth, edges, sketch, or a normal render), then hands that off to fal as conditioning for an AI image.

Note: All videos in this section have processing time truncated.

Depth-to-Image

fal_extension_depth_render.mp4

Renders a Mist depth pass, then generates an AI image guided by scene depth. Great for preserving spatial layout while restyling.

Endpoint	Notes
Z-Image Turbo (ControlNet)	Fast, general-purpose
FLUX.1 [dev] (ControlNet)	Higher quality, slower

Edge-to-Image

fal_extension_edge_render.mp4

Runs Canny edge detection on a scene render and uses the edges as structural guidance for the AI image.

Endpoint	Notes
Z-Image Turbo (ControlNet)	Fast
FLUX.1 [dev] (ControlNet)	Higher quality

Sketch-to-Image

fal_extension_render_sketch.mp4

Renders a Freestyle line drawing (optionally with object-name labels overlaid) and reimagines it as a finished image. Label overlays let the model ground parts of the sketch to specific concepts.

Endpoint	Notes
Nano Banana	Google's image-edit
Nano Banana Pro	Google's image-edit, higher quality
Nano Banana 2	Google's image-edit, newer
GPT Image 1.5	OpenAI's image-edit
GPT Image 2	OpenAI's latest image-edit
Seedream 4.5	ByteDance's image-edit
Seedream 5 Lite	ByteDance's image-edit, lite variant

Render-to-Image (Refine)

fal_extension_render_refine.mp4

Renders the scene normally and refines the result with img2img or an edit model. A strength slider controls how far the model drifts from the input render.

Endpoint	Notes
Nano Banana	Image-edit style
Nano Banana Pro	Image-edit style, higher quality
Nano Banana 2	Image-edit style, newer
GPT Image 1.5	OpenAI's image-edit
GPT Image 2	OpenAI's latest image-edit
Z-Image Turbo	Img2img with strength control, fast
FLUX.1 [dev]	Img2img with strength control
FLUX.2 Klein 9B	Img2img with strength control

Video Rendering

fal_extension_depth_video.mp4

Same controller as image rendering, but the mode switcher is set to Video. Renders an animation sequence as depth or edges, then generates a video conditioned on that sequence.

Note: The above video has processing time truncated.

Depth-to-Video

Exports a depth animation across the scene's frame range and generates a depth-guided video. Optionally supply a first-frame reference image.

Endpoint	Notes
LTX-2 19B	Open-weight, fast
LTX-2 19B Distilled	Open-weight, faster
LTX 2.3 22B Ref V2V	Newer LTX with reference-image support
LTX 2.3 22B Distilled Ref V2V	Distilled, with reference-image support
Wan-VACE 14B	Good at natural motion
Wan Fun 2.2 A14B	Stylized

Edge-to-Video

Canny edges are computed per frame (in parallel threads) and fed to the model as structural conditioning.

Endpoint	Notes
LTX-2 19B	Open-weight
LTX-2 19B Distilled	Open-weight, faster
LTX 2.3 22B Ref V2V	Supports a reference image
LTX 2.3 22B Distilled Ref V2V	Distilled, with reference image

Video results import into the Video Sequence Editor. The addon auto-disables scene.render.use_sequencer so F12 keeps rendering the 3D view, and pops up a notification when the clip lands.

Video Editor Workflows

Available in the Video Sequence Editor sidebar under the fal.ai tab. Results drop in as VSE strips at the current frame and are scaled to the scene's target resolution.

Audio

fal_extension_audio.mp4

Text-to-Speech (Presets)

Generate speech from text using a named preset voice.

Endpoint	Notes
ElevenLabs TTS Turbo v2.5	Broad voice catalog
ElevenLabs v3	Broad voice catalog, newer
MiniMax Speech Turbo
MiniMax Speech 2.8 HD	Higher quality
Kokoro	Open-weight
xAI TTS
Gemini Flash TTS
Inworld TTS

Text-to-Speech (Voice Clone)

Same catalog (for endpoints that support it) but conditioned on a reference audio file — the model clones that voice.

Endpoint	Notes
MiniMax Voice Clone	Used by MiniMax Speech Turbo and 2.8 HD

Text-to-SFX

Generate sound effects from a text description.

Endpoint	Notes
ElevenLabs Sound Effects v2
CassetteAI SFX

Text-to-Music

Generate music from a text prompt with duration control.

Endpoint	Notes
ElevenLabs Music
MiniMax Music 2.6
Stable Audio 2.5
CassetteAI Music

Video

fal_extension_video.mp4

Before submitting a video job you'll see a confirm dialog with the effective size, duration, and cost-relevant settings.

Note: The above video has processing time truncated.

Text-to-Video

Generate a video clip directly from a prompt.

Endpoint	Notes
LTX-2 19B	Open-weight
LTX-2 19B Distilled	Open-weight, faster
LTX 2.3 22B	Newer LTX
LTX 2.3 22B Distilled	Newer LTX, faster
Seedance 2.0	ByteDance
Seedance 2.0 Fast	ByteDance, faster
Kling v3 Standard
Kling v3 Pro	Higher quality
Veo 3.1	Google
Veo 3.1 Fast	Google, faster
Wan 2.2
Wan 2.2 Turbo	Faster
Wan 2.7	Newer
Sora 2	OpenAI

Image-to-Video

Animate a still image. Same catalog of video endpoints (those that support image conditioning).

Endpoint	Notes
LTX-2 19B	Open-weight
LTX-2 19B Distilled	Open-weight, faster
LTX 2.3 22B	Newer LTX
LTX 2.3 22B Distilled	Newer LTX, faster
Seedance 2.0	ByteDance
Seedance 2.0 Fast	ByteDance, faster
Kling v3 Standard
Kling v3 Pro	Higher quality
Veo 3.1	Google
Veo 3.1 Fast	Google, faster
Wan 2.2
Wan 2.2 Turbo	Faster
Wan 2.7	Newer

Shared Workflows

These appear in both the 3D Viewport sidebar and the VSE sidebar with shared config.

Upscale Image

Upscale an image from a file on disk, the current render result, or a texture slot on the active object.

Endpoint	Notes
SeedVR2 9B	Highest quality
AuraSR	Fast 4x
Clarity Upscaler	Detail-preserving

Upscale Video

Upscale a video file on disk or from the VSE.

Endpoint	Notes
SeedVR2 9B
Topaz Video

Preferences

Edit → Preferences → Add-ons → fal.ai — AI Generation Suite:

API Key — your fal.ai key (required)
Output Directory — where generated files are saved
Auto-Import — whether results are imported into the active scene automatically

Advanced Parameters

Every workflow has a collapsible Advanced Parameters section that lets you set arbitrary key/value pairs on the request. Useful when fal ships a new endpoint option before the addon's UI catches up — just add it by name. Values are typed (string / int / float / bool / JSON).

Requirements

Blender 4.2+ (works on 4.x and 5.x)
A fal.ai API key

Installation

From Release

Download the latest .zip from Releases
Open Blender, then either a.) use Edit → Preferences → Add-ons → Install from Disk, then select the .zip file, or b.) drag and drop the .zip into the interface anywhere
Set your API key in the addon preferences

Architecture

controllers/          Blender-side workflows (UI, operators, properties)
  base.py             FalController base class and registration
  operators.py        FalOperator base class (background job submission)
  ui.py               FalControllerPanel (declarative field layout)
  advanced_params.py  Key/value override UIList shared across controllers
  audio/              TTS (preset + clone), SFX, Music (VSE)
  generate_3d/        Text/Image-to-3D (3D Viewport)
  material/           Text-to-Material, Image-to-Maps, Tiling (3D Viewport)
  render/             Depth / Edge / Sketch / Refine modes + Video (3D Viewport)
  upscale/            Image and video upscaling (shared: 3D + VSE)
  video/              Text/Image-to-Video (VSE)

models/               fal.ai endpoint definitions and parameter builders
  base.py             FalModel, VisualFalModel, AudioFalModel, VideoFalModel
  audio_generation/   Speech, SFX, music
  image_generation/   Depth / edge / sketch / refine endpoints
  image_processing/   Image upscaling
  material_generation/ Material, PBR estimation, tiling
  mesh_generation/    Text-to-3D, image-to-3D
  video_generation/   Text/image-to-video, depth/edge video
  video_processing/   Video upscaling

app.py                Panel registration for 3D and VSE sidebars
job_queue.py          Async FalJob / JobManager (threads + bpy timer)
importers.py          GLB/OBJ import, texture application, VSE strip helpers
preferences.py        API key, output dir, auto-import
utils.py              Upload/download, compositor snapshot/restore, fonts

Design notes:

Controller / Model separation. Controllers handle Blender-side UI, operators, and scene interaction. Models define the fal.ai endpoint URL and translate UI params into the endpoint's schema. Adding a new endpoint is usually a single new FalModel subclass.
Declarative panels. FalControllerPanel drives field layout, conditional visibility, and grouping from lists and lambdas — no manual draw() per workflow. Conditionals key off the selected endpoint's ui_parameter_map, so fields show up only when the current endpoint actually supports them.
Async job queue. API calls run in background threads via fal_client.subscribe(). A bpy.app.timers loop polls for completion, keeping Blender's UI responsive.
Dual-context panels. Controllers declare panel_3d, panel_vse, or both and auto-register in the appropriate editor sidebar.

Tests

cd fal-blender
./tests/run_tests.sh

Model tests don't require Blender; they exercise the FalModel.parameters() builders and the per-endpoint ui_parameter_map forwarding.

Planned

Image-to-Grease Pencil-to-Curves (vector workflow)
Multi-Image-to-3D pipeline (multi-angle capture to mesh)
Real-time neural rendering preview
LoRA support for consistent styling

License

Built with love by Benjamin Paine for fal.ai

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
.github/workflows		.github/workflows
assets		assets
controllers		controllers
models		models
scripts		scripts
tests		tests
vendor		vendor
.gitignore		.gitignore
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
TESTING.md		TESTING.md
__init__.py		__init__.py
app.py		app.py
blender_manifest.toml.template		blender_manifest.toml.template
branding.py		branding.py
importers.py		importers.py
job_queue.py		job_queue.py
preferences.py		preferences.py
render_material.py		render_material.py
requirements.txt		requirements.txt
utils.py		utils.py

Folders and files

Latest commit

History

Repository files navigation

fal.ai Official Blender Extension

Quick Start

3D Panel Workflows

3D Generation

Text-to-3D

Image-to-3D

Materials

Text-to-Material

Image-to-Maps

Image Rendering

Depth-to-Image

Edge-to-Image

Sketch-to-Image

Render-to-Image (Refine)

Video Rendering

Depth-to-Video

Edge-to-Video

Video Editor Workflows

Audio

Text-to-Speech (Presets)

Text-to-Speech (Voice Clone)

Text-to-SFX

Text-to-Music

Video

Text-to-Video

Image-to-Video

Shared Workflows

Upscale Image

Upscale Video

Preferences

Advanced Parameters

Requirements

Installation

From Release

Architecture

Tests

Planned

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 1

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages