ToonzComposer

Create custom videos from text and keyframe images

checking backend… local engine

Prompt never burned into the video

Prompt strength
β€”
Prompting guide — what actually works for image→video

Formula: subject + action + camera move + lighting + style + motion quality

Example: β€œa claymation baker lifts a tray, slow dolly right, warm dawn light through the window, stop-motion feel, smooth 24fps motion”

Camera words the models obey: push in, pull back, dolly left/right, pan up, orbit, crane down, handheld follow, slow zoom out, static wide.

Say how it should move: β€œfluid motion”, β€œgentle drift”, β€œenergetic dash”. Vague prompts give a still image with flicker.

Morph engine tip: describe the journey between the two frames (β€œmorphs into”, β€œtransforms as”, β€œthe camera glides from A to B”).

Keyframes up to 10 Β· drag to reorder

Total 0.0s / 30s max Β· Segment β€”

Text & titles separate from the prompt

Engine & settings

⚑ Turbo / Lightning LoRA πŸš€ Fast local encode (WebCodecs) πŸ’Ύ Auto-save renders to gallery
Cmd/Ctrl + Enter generates Β· paste images with Cmd/Ctrl + V Β· the prompt is not drawn on screen (use the Text section).

Stage

preview Β· 0.0s
0.0 / 0.0s

Render pipeline

Idle0%
    ToonzComposer ready. Β· Live preview is real: it animates your keyframes and segment timings. Β· Local engine encodes in your browser and downloads instantly. Β· AI engines POST to the Space backend, which runs diffusers + ffmpeg.

    Output

    No render yet.

    Gallery saved renders

    Gallery keeps your renders in the Space's storage so you can re-download them later.

    Projects save / load settings + keyframes