A 3D scan of a person copies your body from a plain webcam, live, and you are painted out of your own room so it stands where you stand. Add face tracking, hand signs and captions, then send it all to OBS, or straight to Twitch and YouTube Live.
Scan by Erant (superspl.at), CC BY 4.0.
Scan by Erant (superspl.at), CC BY 4.0.
No motion capture suit and no rigging by hand. Four steps, all in one graph, and the two demos that do it come with the app.
Open a Gaussian splat scan of a person, saved as .ply, .spz, .splat or .sog. Crop off the floor and any stray splats around the feet.
Splat rig fit stands a skeleton inside the scan from its shape alone, with fingers, toes and face joints if you want them. Drag any joint into place with the Skeleton tool in the viewport.
Body mesh reads your whole body from the webcam every frame. The motion goes onto the scan's skeleton, so it copies your arms, legs, spine and head with its own proportions. Stand in frame and press Calibrate once.
Person segmentation finds you in the webcam picture and AI inpaint fills the gap from the room around you. The avatar lands where you stand, and when the camera moves the fill holds still.
Import a VRM, glTF or FBX character and move it from the webcam the same way. Or send your body and face to VSeeFace, Warudo or VNyan over VMC and keep the app you use now.
The jobs a week of streaming asks for, in one app. Each one links to a short how-to.
Take the green out, clean the spill off your edges and put any room behind you. No green screen? Background removal cuts you out of a plain room.
A name bar, a title or a badge, laid over the camera and faded in when you want them.
Give each key an action, an OSC message, a Python snippet or a web request, so one press switches a scene, a look or a sound. Tally lights the keys.
Blendshapes reads 52 face values off the webcam, smile, blink, jaw and brows, and moves a character's face shapes with them.
Thumbs up, open palm, a fist or a peace sign each arrive as a named trigger. Wire one to a scene change, a sound or a look.
Your voice turns into words on screen as you speak, transcribed on your own machine.
Put clips in order on the timeline, trim them, add fades and a title, and export a short.
Crop one picture square, 4:5 and 9:16, give each its own line of text and save all three.
Cut the long pauses and stumbles, even out the level and save the episode.
Your mic and your music come in as live values. Your voice can swell a shape, each beat can glitch the camera, and the loudness of a song can push particles across the frame. Any setting in the graph can follow the sound.
The same mic can run through your own plugin effects on the way out, or into a vocoder that turns your voice into a synth.
Keep OBS for your scenes and alerts and hand it the avatar, or go live from Kinematic itself. Either way, everything runs on your own machine.
Go live →Run the graph with no window and send its picture out over NDI:
$ kinematic run avatar.kin --out ndi://Avatar
Name an NDI sender and pick it up in OBS as an NDI source. On a Mac, Syphon hands the picture to OBS on the same machine.
RTMP out encodes H.264 video and AAC sound and streams to Twitch, YouTube or Facebook Live with your stream key, no other app in between.
Write the same picture and sound to an MP4, MOV, MKV or WebM file as it goes out, and split a long stream into parts.
Screen capture grabs a whole monitor, one window or a region of the screen, so a game or a browser can sit under your avatar.
Keep your webcam, your mic, your Stream Deck and OBS. Kinematic sits beside them on the same machine or across the network.
Each tutorial starts from an empty file and takes under fifteen minutes. The demos come with the app: open one in the Demos browser, press Play, then take it apart on the canvas.
Tutorial · 12 minCut a short video
Tutorial · 12 minTidy a podcast episode
Tutorial · 12 minMake a visual that pulses to your music
Tutorial · 8 minMake a voice sound like an old radio
Tutorial · 12 minMake a poster from a photo
Tutorial · 12 minPaste in a Shadertoy and give it sliders
Rigged person infillYou vanish from the webcam picture and the scanned man stands in your place, with the room painted back in.In the Demos browser
Rigged person liveThe same scanned man, moved by your whole body on the webcam.In the Demos browser
Blendshapes avatarJaw, smiles, blinks and brows come off your face as named values, ready to drive an avatar.In the Demos browser
Gesture triggerMake a hand sign, fire a cue: thumbs up, open palm, a fist and more.In the Demos browser
Googly eyesCartoon googly eyes pinned to your face, following your eyes on the live camera.In the Demos browser
Face particle fountainGlowing sparks spray from your eyes and nose and follow your face as you move.In the Demos browserThe first two demos use a scan by Erant (superspl.at), CC BY 4.0.
Free. No account, no license key and no watermark on your stream. macOS, Windows, Linux and Raspberry Pi.