How can we help?

• Linux: the AppImage is one file that runs on most distributions from 2022 on (glibc 2.35 or later). The .tar.xz has the same files to unpack wherever you like.
• Windows: the installer, for Windows 10 and 11 on x64, or the .zip to unpack wherever you like.
• macOS: the universal app, for Apple silicon and Intel Macs on macOS 13.5 or later.

The downloads page picks the right one for the computer you are on.

It needs to be marked as a program first: chmod +x Reelwright-*.AppImage, or tick “Allow executing file as program” in its properties. If it still will not start, your system may be missing FUSE, which AppImages use to mount themselves. Run it with --appimage-extract-and-run, or use the .tar.xz instead.

SmartScreen warns about apps it has not seen often, and Reelwright's builds are not signed yet. Check the download's SHA-256 against the one on the downloads page, then choose More info and Run anyway.

Until releases are notarized by Apple, Gatekeeper stops the first launch. Open System Settings, then Privacy & Security, and choose Open Anyway next to Reelwright Studio. If macOS says the app “is damaged”, it is the download's quarantine flag: remove it with

xattr -dr com.apple.quarantine "/Applications/Reelwright Studio.app"

Yes: reelwright comes with Studio and does everything Studio does, from validate to render. In the .tar.xz it sits beside Studio; from the AppImage, put cli first, as in ./Reelwright-0.13.0-x86_64.AppImage cli help. On Windows the installer puts it on your PATH (unless you untick that); on a Mac it is inside the app, "/Applications/Reelwright Studio.app/Contents/MacOS/reelwright". The command line reference lists every command.

Reelwright draws with Godot's Forward+ renderer: Vulkan on Linux and Windows, Metal on macOS. Any graphics card or integrated GPU with a working Vulkan driver will do, and every Mac that runs macOS 13 has Metal.

• Update your graphics driver. On Linux, install your distribution's Vulkan drivers (on Debian and Ubuntu, mesa-vulkan-drivers) or your GPU vendor's driver, and check that vulkaninfo --summary lists your GPU.
• Run reelwright doctor. Its engine line names the renderer and GPU Reelwright found, or says what stopped it, with how to fix it.
• On a Linux machine with no desktop session, Studio has no display to draw on; rendering from the command line still works under xvfb-run.

On Linux, yes: Mesa's software Vulkan driver (lavapipe) draws exactly the same frames, only more slowly. Reelwright's own tests render that way.

Wherever you keep them. “Try the demo show” and “New show…” make theirs in a Reelwright folder in your Documents (in your home folder if you have no Documents folder), and Studio says where each one went. You can open a show folder from anywhere. A show is a folder of plain files:

• specs for its theme, style, characters, sets, props and episodes (JSON), and its screenplays (.fountain);
• renders/, where finished episodes go;
• .reelwright/, caches Reelwright can make again (reelwright clean empties them).

Each show is its own git repository, and every approved change is a commit in it. Studio keeps its settings in ~/.local/share/Reelwright on Linux, ~/Library/Application Support/Reelwright on macOS and %APPDATA%\Reelwright on Windows.

In Studio, Settings (the last place in the rail) has Export as Show Pack…: the show's files in one zip, without its renders or caches. Or copy its folder, or push its git repository somewhere. Nothing about a show lives outside its folder except the shared voice model, which Reelwright downloads again if it is missing.

Open it and go to Settings, or choose Show settings… in the show menu. Its title, logline, cover picture, style, theme and colors, the size full renders are made at, and how Claude helps with it are all there. Every change there is undone with Ctrl+Z. Duplicate makes a copy to try things on; Archive takes it off your list (it waits under Archived, and Restore brings it back); Delete asks you to type its name, then puts its folder in the trash. On reelwright.diy/app it is the same, and only the show's owner can do these.

Yes. Each release reads every earlier version of each kind of document and writes the newest. Opening a show changes nothing on disk until you save something.

No. Writing, directing, voicing, playing, editing on the timeline and rendering all work without it, as do New character, New set and New episode. So does re-directing a moment: a note such as “angrier” goes into the script for the built-in director. Claude adds the showrunner chat, the writers' room, and Claude's own direction, memory and looks.

With the Claude you already have, and no API key. Open Settings › Claude in Studio (Ctrl+,, the gear at the foot of the rail, or Connect Claude… wherever it is offered): it adds Reelwright to the Claude Code on your computer, or gives you the address to add to claude.ai or the Claude app as a custom connector. Your Claude then directs, writes, keeps the show's memory and finds looks on your own plan, and what it does shows in Studio's chat as it does it. Studio's own chat runs on Claude Code when it is signed in with your plan; with no key and no helper, it hands what you ask to your own Claude.

Yes, a key is an option. In Settings › Claude, paste an Anthropic API key from your Anthropic Console account. Reelwright checks it with Anthropic (free), then saves it in a file only your user can read: ~/.config/reelwright/anthropic-api-key. It is sent only to Anthropic, and never into a problem report. Settings shows it by its last four characters, with Test, Replace key and Remove; reelwright key forget removes it too.

A key is optional there too. Open Settings › Claude (Ctrl+,, Settings… in the show menu, or Connect Claude… in the chat) and paste your own Anthropic API key. Reelwright checks it with Anthropic (free), then keeps it encrypted on the service for your account only. Nobody is shown it again, not even the service's admins: Settings shows only its last four characters, what you spent this month, Test and Remove. Only what you ask for uses it, to talk to Anthropic, in your own shows and in those shared with you; nobody else in a shared show can use it. The web edition never holds a Claude subscription's sign-in; to use your subscription, connect your own Claude to it instead (below). Whoever runs the service may set you a monthly spending cap.

In the web edition, yes, with anyone who has an account on the same service. Open the show, choose People… in the show menu (or in the show's own settings, under World), and add them by their email as an editor (they can do everything you can but share it) or a viewer (they can watch, read and play, and change nothing). You can change their role or remove them there at any time; someone you remove is cut off at once. You both work in the same studio, so an edit by one is there for the other, the initials of whoever else is in the show appear under its status at the foot of the sidebar, and every change in History says who made it. Ctrl+Z takes back only your own changes. Each person's requests to Claude go through their own Claude, or their own key. A show shared with you is listed under Shared with you in the show menu, with its owner's name.

Watch and read: the dashboard, the cast and sets, every script, the timeline and its playback, renders, History and memory. Their Studio says View only at the foot of the sidebar and leaves out everything that would change the show or spend money: no new episodes or imported scripts, no rendering, voicing or recording, no chat or Write with Claude, no approving proposals or memory updates, no locks. The service refuses those too, whatever the page sends. Ask the show's owner to make you an editor to work on it.

Reelwright costs nothing. Your own Claude works on your Claude plan, as anything else you ask it. With an API key, you pay Anthropic directly, at their API prices, for what the showrunner uses; Reelwright adds nothing on top. The chat shows what each turn on a key cost, and each show has budgets, per request, per episode and per month, that stop it from spending past them.

Yes, and it is the first way. On the desktop, Studio adds Reelwright to Claude Code (reelwright mcp serves a show's tools to it), and Studio's own chat runs on Claude Code signed in with your plan. Claude's direction, memory, looks and writing work through it with no key, and reelwright direct directs on it from the command line too.

Yes. Your own Claude, Claude Code or the Claude app on your Pro or Max plan, connects to Reelwright's MCP server at https://reelwright.diy/app/mcp and works on your shows from there. In Claude Code, run claude mcp add --transport http reelwright https://reelwright.diy/app/mcp, then claude mcp login reelwright to sign in; in claude.ai or Claude Desktop, open Customize › Connectors, choose Add custom connector with that address, and if asked, Sign in now. Your browser opens Reelwright to sign in and asks you to allow it. Claude works as you, with your role in each show, and what it proposes appears in Studio for you to approve. Settings › Claude in the web edition has the command and the address to copy, a token for setups with no browser, and your connections, each of which you can disconnect. Reelwright never sees your Claude sign-in. See Use your Claude subscription with Reelwright on the web.

Yes, if you turn it on, through your own computer: in Settings › Claude, Your Claude subscription, make a token and turn on Also use my plan for Studio's chat (experimental), then run reelwright claude bridge --service https://reelwright.diy/app on your computer with the token in REELWRIGHT_TOKEN. Your chat messages then go to the Claude Code on your computer, signed in with your plan, and its answers stream back into Studio. It is off by default because Anthropic's rules ask for its agreement before a product offers this to everyone. If the bridge is not running, the chat says so.

Yes. Writing, directing, playing and rendering need no network. Reelwright goes online only to download its voice model once, to talk to Claude if you connected it, and for the daily update check, which you can switch off.

The first time a show's lines are voiced, Reelwright fetches the Kokoro voice model (about 330 MB) and the small runtime it runs in, checks their checksums, and keeps them in your user cache. It says so while it does. Until then, lines are timed from their text, so you can direct and play an episode before any voice is downloaded.

Studio checks reelwright.diy at most once a day and says when a newer version is out, with a link to what changed. It never downloads or installs anything until you ask; switch the check off on the welcome screen. You can also follow the release feed or the changelog.

On Linux, with Studio started from its AppImage, press Update on the rail or the welcome page. Studio downloads the new AppImage beside the one you run and checks it against Reelwright's signed release feed, and Restart to update reopens it on the new version with the show you had open. If the new version has not opened within three minutes, the old one is put back and started.

On Windows and macOS, and from the .tar.xz, download the new version and use it in place of the old one. Your shows are separate from the app, so they are not touched.