Open source · Images + video · Runs on your machine

Your local-first AI studio.

One studio for every image and video model. Your characters and results stay on your disk. Make stills with Codex, Gemini, OpenAI or ComfyUI, animate them with Kling, Veo or Seedance, and switch providers without moving a thing.

Paste into Terminal · macOS & Linux

$ git clone https://github.com/vishwakulkarni/PoseForge.git && cd PoseForge && npm install && npm run dev✓ Ready on http://localhost:3000

Try the live demo Star on GitHub

No PoseForge account
Apache 2.0
macOS & Linux

Identity portrait, overhead pose reference, and identity-preserving final result shown together

Identity portrait, seated pose reference, and identity-preserving final result shown together

Identity portrait, side-lunge pose reference, and identity-preserving final result shown together

PoseForge identity-preserving pose transfer showing an NRI family identity, walking pose reference, and final generated image

Works with

Codex CLI
ComfyUI
OpenAI
Gemini
fal.ai
Replicate
Higgsfield
Dreamina

engines, one interface: 9
video models: 6
people per frame: 4
bundled pose references: 16
provider cost with local ComfyUI: $0
PoseForge accounts needed: 0

One interface · every provider

Switch the engine. Keep everything else.

Your characters, poses, projects and history belong to you, not to a provider's account. Change engines in one click and carry on where you left off.

Characters live on your disk
Build a character once. Every engine uses the same saved references. Nothing to export, re-upload or rebuild.
Every result stays yours
Images, clips, prompts, cost and latency land in one local history, whichever provider made them.
Switch without starting over
If a provider raises prices or changes its model, pick another from the dropdown. Your project keeps going.

ComfyUI keeps generation fully offline. Other engines receive only the references and prompt for the image you're making. See the data flow

~/PoseForge · studio projectlocal

A saved character reference — **Character 01**storage/characters

Engine

engine → Codex CLI (Local CLI)character ✓ reused, nothing re-uploadedproject ✓ same settings, no setuphistory ✓ every take on your disk

Local history

+ every take

Video, too

From still frame to moving shot.

Send any frame you made to a video model, on the same canvas and with the same characters. Every clip is saved to your local history next to the image it came from.

Image → videoAnimate any frame you made, straight from its node.
Text → videoDescribe the shot and let the model block it out.
First & last framePin the start and end; the model fills the motion.
Reference → videoGuide a shot with reference images, clips and audio.

Kling 2.5 Turbo Pro
Kling 2.1 Master
Veo 3.1 Fast
Seedance 2.5
Seedance 2.0
Seedance 2.0 Mini

Through fal.ai, Higgsfield and Dreamina.

Clip 01 · 5s

Clip 02 · 5s

Clip 03 · 5s

Inside the studio

Everything around the image matters too.

Sources, direction, cost and history live in one local workspace, whichever engine you use.

Multi-person

Up to four people in one frame

Compose families and groups, with identity strength set person by person.

Person 1 · high
Person 2 · high
Person 3 · medium
Person 4 · high

Creative direction

Direct like a photographer

Lens, aperture, light, framing, wardrobe, expression, grain and finish.

35mm
f/1.8
Golden hour
3/4 framing
Film grain

Pose library

16 poses built in

Upload any reference or split a pose sheet into single poses. Everything you add is saved for next time.

Variations

Up to six takes per run

Compare takes side by side and keep the one that works.

Metrics

Know what every image cost

Spend, tokens, success rate and median latency for each engine. Export to CSV or JSON.

Lifetime spend
Tokens used
Success rate
Median latency
CSV · JSON

ID photos

Passport and visa photos

Crop U.S. and Indian passport, visa and OCI photos locally, with the official guidance alongside.

Recipes

Reuse a look

Save camera, light and styling as a recipe and apply it to the next shoot.

Any photo

Straight from your phone

JPG, PNG, HEIC and HEIF are welcome. No conversion step.

How it works

Three steps from portrait to finished frame.

Keep the person and change the pose, camera, light and mood. Chain image and video steps on one canvas, even when each step uses a different engine.

01
Pick your characters
Choose one to four people from your local library, or add a portrait from your camera roll.
02
Direct the pose
Borrow body language from any reference, then set lens, light, framing, wardrobe and finish.
03
Generate, then animate
Run on any image engine, keep the best take, then turn it into a clip with Kling, Veo or Seedance.

The Advanced Studio canvas: prompts feed Codex CLI image generators, whose frames feed Kling video generators, all in one project — Advanced StudioCodex CLI images → Kling video, one projectOpen the canvas

Local-first by design

One command. Your own AI studio.

Paste it into Terminal. PoseForge downloads, installs, sets up its own database and starts. No account, no Docker, no cloud gallery.