Weave

// a concept in · a narrated video out

Explain anything, one scene at a time.

Describe an idea. Weave writes the script, draws every scene, narrates it, and encodes the video in your browser. Every cut lands where the narration does — and any scene you choose can move, as a short generated clip in place of the held still.

Create your accountHow it worksWatch the tour

5 of 5 scenes ready · 1 in motion · encoding in this tab

narrator voices
30
formats — landscape, portrait, square
3
music styles, or describe your own
5
per music clip
$0.04

// watch

Ninety seconds with Lumi.

A concept and a cost estimate, an editable script, every scene generated at once, cuts timed to the narration, voices, labels, motion and music — then an MP4 with subtitles.

Watch on its own page

Weave in ninety seconds, with Lumi · 1:33

// how it works

Four steps, two of them yours.

01you

It writes the script

Give it a concept. It returns a scene list — one narration paragraph, one image description and one on-screen label per scene. You edit all of it before anything is paid for.

02

It draws every scene

One image per scene, generated against a shared style reference so a dozen pictures read as one video instead of a dozen unrelated ones — in the same click as the narration below. Regenerate any single scene you do not like, or set it in motion as a few-second clip generated from its own still.

03you

It records the narration

A voice you pick — auditioned first — reads each scene. The audio it returns decides how long that scene is held on screen, so the cuts land on the narration rather than on a guess. Re-record any single scene.

04

Your browser makes the video

The images, the narration and an optional music bed are encoded into an MP4 in this tab — the server never holds a frame of it — and the file is yours the moment it finishes. A copy under 10 MB is kept as usage telemetry; the privacy page says exactly what.

// what you get

Built to be reviewed, not trusted.

01

Legible labels, always

Image models spell badly, so they are told to draw no text at all. On-screen terms are composited afterwards in a real typeface — which makes a typo a text edit rather than a paid regeneration.

02

Two review gates

The script costs about a cent; the pictures and narration cost most of the total. So you approve the words before anything is generated, and then every image and every take is reviewable — and regenerable one at a time — before the render.

03

Generated background music

An optional instrumental bed, generated to suit the tone and ducked under the narration. Pick a style or describe your own; vocals are always excluded.

04

Motion, scene by scene

Any still can become a short clip generated from itself — four, six or eight seconds, slowed and then held to cover exactly its narration. Optional per scene, priced precisely before the click, and the most expensive thing here, so it is never on by default.

05

Your key, your bill

Generation runs on your own Gemini API key, encrypted at rest. You see an estimate before the expensive step, and the real spend afterwards from recorded token counts.

// what it costs

An estimate first. The real number after.

Every generation runs on your own Gemini key, so the bill is Google's and the figure here is the same one the app shows before you approve the images. Afterwards it reports what was actually spent, from the token counts each call returned — not from this estimate.

A 2-minute video, 13 scenes, music on

≈ $1.04

Script
$0.01
Images
$0.87
Narration
$0.08
Music
$0.08

Estimate at today's list prices for the recommended image model. Images carry a SynthID watermark; the script is a flat allowance of about a cent.

With every scene in motion, at $0.05 a second of clip — chosen scene by scene, before the click≈ $6.24

// the narrators

30 voices. Audition before you commit.

Every voice reads the same line so they can be compared fairly, and a sample costs a fraction of a cent. The voice locks once a scene is narrated — one video, one narrator.

// get started

Bring a concept and a key. Leave with a video.

Create your account