Open-source video, written in Flutter

You write widgets.
Fluvie shoots the film.

A real Flutter widget tree goes in. A real MP4 comes out. You say what the video is. Fluvie works out when everything happens, frame by frame. No timeline to scrub by hand. No video editor required.

It is code. It runs anywhere. It is a real video file.

lesson_01.dart Dart

import 'package:flutter/material.dart' hide Animation; // Fluvie's wins
import 'package:fluvie/fluvie.dart';

Video lesson01Video() => Video(
  size: VideoSize.square,
  scenes: [
    Scene(
      duration: 4.seconds,
      background: Background.gradient([...]),
      children: [
        Text('Hello, Fluvie')
          .animate([Animation.fadeIn(), Animation.pop()]),
      ],
    ),
  ],
);

hello.mp4 1080 × 1080 · 30fps

Hello, Fluvie

frame 9 of 16

$ fluvie render ./lib/hello.dart --out hello.mp4

no display needed

Roll your first video Open the live demo

02 · THE ONE IDEA

You bring the what.
Fluvie brings the when.

Not a recorder. A renderer. A screen recorder captures whatever happened to be on screen. Fluvie computes every frame from your description, so the timing and composition can be replayed from the same code.

The stopwatch way

0ms title.fadeIn(at: 0)

650 subtitle.show(at: 650)

900 chart.grow(at: 900)

?! move one, fix them all

The Fluvie way

 show the title

 after the title lands, float the subtitle

 then grow the chart

✓ move one, the rest re-flow

01

Describe, do not direct

02

The frame is the only clock

03

Replayable composition and timing

03 · THE CLEVER BIT

Triggers, not timecodes.

Say "after the title lands, float the subtitle." Fluvie schedules it. Move the title earlier and everything downstream follows. You never type a frame number. Try it: drag the title's start and watch which side keeps up.

Inspector

title

subtitle

subtitle*

Move the title's start

On the Fluvie side, the subtitle is anchored after the title, so it followed automatically.

04 · THE PAYOFF

Write one. Render a thousand.

One template plus a data list renders one video per row. The frame cache keys on content, so an unchanged frame is read, not redrawn. Change one number and only the frames that moved render again.

"A thousand data-driven videos, rendered like one."

rows = [ Q1, Q2, Q3, Q4 ]

$2.4M

$3.1M

$3.0M

$4.2M

One template, one row per card. The cache shares the work across renders.

05 · THE CALL SHEET

From fluvie init to a real MP4.

Five steps, copy as you go. Each command ticks off when you grab it.

Prereqs: Flutter 3.44+. That is the list.

You do not need FFmpeg. The first render downloads a pinned build and caches it.

$ flutter --version

1
Install the renderer

The CLI that scaffolds, previews, and writes the file.
```
dart pub global activate fluvie_cli
```
2
Scaffold the project

A composition, an assets folder, a pubspec. Works in an existing Flutter app.
```
fluvie init --dir my_promo
cd my_promo && flutter pub get
```

Write the video

One centered scene, a teal background, a title that fades in and pops. The import hides the names shared with Flutter.

lib/example_video.dart

import 'package:flutter/material.dart' hide Animation, Clip, Image, Tween;
import 'package:fluvie/fluvie.dart';

Video build() => Video(
  size: VideoSize.square,
  scenes: [
    Scene.centered(
      duration: 4.seconds,
      background: Background.color(const Color(0xFF0E9E8E)),
      child: const Text(
        'Hello, Fluvie',
        style: TextStyle(fontSize: 64, color: Colors.white),
      ).animate([Animation.fadeIn(), Animation.pop()]),
    ),
  ],
);

4
Watch it live

Hot reload. Edit the file, save, watch it redraw. No app to run.
```
fluvie preview ./lib/example_video.dart
```
5
Render it

When it looks right, write the MP4. No display needed.
```
fluvie render ./lib/example_video.dart
```

Change the data, run the render, ship the file.

Your video is a build artifact now. That is the whole point.

Read the full guide Create from local assets Skip setup, scrub it live ↗

06 · THE REEL WALL

13 runnable lessons, each in one readable file.

Explore text and motion, scene transitions, charts, code, visualizers and captions. These illustrated cards link to the demo; the runnable Dart sources live in the gallery.

Scrub any of these live

07 · THE CREW

A small crew. One does the magic.

Ten MIT packages for authoring, rendering and tooling. Most people only ever need fluvie. The rest are there when you scale.

08 · OPEN SOURCE

MIT. Inspect it, fork it, host it, keep it.

No per-render Fluvie license fee. The MIT source lets you inspect how frames are prepared and drawn, fork it, and host it yourself.

MIT

license

packages

lessons

Fluvie license fee

Read the source Self-host the render API

Inspectable
The timing and capture implementations are open to inspection, with source-linked tests.
Auditable
CLI receipts record source inputs, tools, output hashes and probed media facts for comparison.
No lock-in
Readable Dart, ordinary media assets and exported files. Keep the inputs and re-render them.
Self-hostable
Run the render API in your own Docker. Choose your own storage and the network access allowed to media and model providers.
Free, in the open
MIT, with a public roadmap and issues. The lints and lessons came from real use.

09 · PICK YOUR SEAT

Same Video, four front doors.

Scrub it live in the browser. Render it from the CLI. Stand up your own render server. Or let an assistant shoot it for you. These paths share the authoring model and timing engine; each backend has its own capabilities.

For the cold visitor

open https://demo.fluvie.dev

Open the live demo

One composition, several rendering paths.

Compare the default rendering backends

Capability	desktop	browser	mobile
Formats	gif, imageSequence, mp4, transparent	gif, imageSequence, mp4, transparent	mp4
Video codecs	h264, h265	h264, h265	h264, h265
Audio	yes	yes	yes
Target bitrate	yes	yes	yes
CRF control	yes	yes	no
Encoder presets	yes	yes	no

Codec availability and host requirements still depend on your environment. Unsupported requests fail before capture. Custom renderers declare their own capabilities.

Read the requirements and regression evidence

10 · EASY MODE

Or just say what you want.

fluvie_ai turns a plain sentence into a real VideoSpec. In the Playground, the AI Assistant writes editable Flutter-style Dart into the editor, so you can read it and refine it with another sentence. The model runs on the server, so the default flow keeps the provider key on that server. The MCP server in fluvie_server lets your assistant do the same from a chat. This is the optional easy mode, not the only mode.

1 · the prompt

"A 10-second product promo, brand teal, a counter rising to 10,000."

2 · editable Flutter code

Video promo() => Video(
  size: const VideoSize(320, 320),
  fps: 12,
  poster: 3.seconds,
  scenes: [
    Scene.centered(
      duration: 10.seconds,
      background: Background.color(const Color(0xFF0E9E8E)),
      child: Counter(
        to: 10000,
        reveal: 3.seconds,
        style: const TextStyle(fontSize: 48, color: Colors.white),
      ),
    ),
  ],
);

3 · actual Fluvie render