Slopcamera

A multimedia studio for your coding agent. Slopcamera lets your coding agent make images, diagrams, animation, 3D scenes, and video from source files it can keep revising. It is a free, open-source Bun CLI, TypeScript SDK, Agent Skill, and MCP server for Codex, Claude Code, and other coding agents.

Your agent writes the scene, diagram, or edit as a short source file. Slopcamera checks it, renders it with the right engine on your machine, and keeps the inputs, so the next change is an edit to that file and a new render. Combine authored scenes, generated media, and your own footage in one project.

Install · Techniques · First diagram · Docs · Examples

What you can make with Slopcamera

• Films and animation. Author motion graphics, illustrated scenes, and native Blender shots. Control camera, materials, lighting, and timing.
• Diagrams as code. Write diagram JSON and get an editable tldraw file, light and dark SVGs, and PNGs after a strict check.
• Visual explanations. Make educational films with Manim, editable labels, mathematics, narration, and graphics.
• Images and designs. Create illustrations, trace raster artwork to SVG, or revise a parametric architectural model.
• Edits of your footage. Compose shots, sound, overlays, and captions, then export horizontal, vertical, square, or portrait versions.

The techniques catalog connects each job to its supported commands and requirements. Slopcamera is free and open source. Local editing and rendering need no Slopcamera account. Generation uses your own Vercel AI Gateway account, or prepaid Hraness Credits for prompt-only hosted images; model usage is billed separately.

Install Slopcamera

Install Bun 1.3.14 or newer, then install the Slopcamera v3.10.3 release and its matching Agent Skill:

bun add --global https://github.com/hraness/slopcamera/releases/download/v3.10.3/hraness-slopcamera-3.10.3.tgz
slopcamera skill install --target agents

The agents target works for agents that read ~/.agents/skills. Use --target claude for Claude Code, or omit the target for Codex. Add --scope project inside a repository for a project-only install. Start a new agent session so it loads the skill, then ask for a result:

Make a short illustrated film of a late-night tram whose destination changes to the moon. Give it warm windows, a rainy city, and a quiet ending. Start with a frame and a motion test so we can direct it before the final render. Keep the source and sound editable.

Setup guides: Claude Code · Codex · Other agents · MCP clients. Native engines such as Blender install separately; the capability reference lists what each technique needs.

On macOS, start with your first animation for a supplied scene you can render and revise without a model account. It needs the browser and FFmpeg runtimes reported by slopcamera doctor --json. For a first result on macOS, Linux, or Windows with no browser or paid model, follow your first diagram. The directing guide explains how to ask for useful changes to composition, timing, camera, and sound.

Build from source

Install Git as well, then use a new checkout:

git clone --branch main https://github.com/hraness/slopcamera.git slopcamera-source
cd slopcamera-source
git rev-parse HEAD > ../slopcamera-source-commit.txt
bun install --frozen-lockfile --ignore-scripts
bun run build:sdk
bun run build:desktop:cli

Keep the checkout and recorded commit. In this shell, define the command against that exact build:

export SLOPCAMERA_SOURCE_ROOT="$PWD"
slopcamera() { bun "$SLOPCAMERA_SOURCE_ROOT/apps/desktop/dist/cli/main.js" "$@"; }
slopcamera --help
slopcamera doctor --json
slopcamera skill install --target agents

The skill comes from the same checkout as the CLI. In a later shell, restore the checkout path and function. The source-install guide covers workspace placement and resuming long runs.

CLI updates in Slopcamera 3.10.0 and newer are enabled by default for supported Bun and npm global installations. See update controls to keep a version or check for a release.

First project: make a film, then direct it again

Last tram to the moon is a complete first project: original illustrated artwork, a twelve-second journey, and a visible revision. The tutorial renders locally without a model account or downloaded artwork.

The two supplied render requests use the same scene. The first sets the city and departure; the second enlarges the moon while keeping the journey, palette, and timing. Ask your agent:

Make the moon dominate the destination. Preserve the copper tram and rainy city. Keep both versions so I can compare their endings.

Follow the animation tutorial, then remix the showcase: a bottled storm, a paper ocean, a dancing laundromat, a band of harmonics, and three edits of the same eclipse footage. Each piece has source, production instructions, and a suggested creative revision.

For a smaller first task, make a diagram and export its editable tldraw document, light and dark SVGs, and PNGs.

Techniques: diagrams, motion graphics, 3D, and video

Each technique pairs a short source file the agent writes with commands that render and check it. The techniques reference lists them all with their requirements.

Technique	The agent writes	Slopcamera renders	Example
Diagrams	Diagram JSON	.tldr, light and dark SVG and PNG, after a strict check	Pipeline
Vector tracing	A path to a raster image	SVG traced locally with VTracer (macOS and Linux)	Color and duotone
Motion graphics	An HTML scene in one of seven profiles: plain, Motion, p5, Two, Paper Shaders, Three.js, vgpu (WGSL)	H.264 MP4 with optional local audio	Kinetic type
Music-timed motion	An HTML scene, a local track, and its tempo in BPM	Motion timed to the declared tempo; beats are not detected from the audio	Island Pulse
3D scenes	Scene JSON with parts, media, and named cameras	Stills, contact sheets, and video, with orbit, dolly, crane, and rail moves	Camera orbit
Parametric design	Named dimensions and constraints	Geometry and a scene you can re-render at new sizes	Pavilion
Native films	A Blender, CadQuery, or Manim program from a starter	Frames and video from your installed engines	Product film
Explainer video	A Manim scene, captions, and a presenter	A lesson video with typeset math	Geometry lesson
Footage edits	Cuts, captions, reframes, and color as project decisions	16:9, 9:16, 1:1, and 4:5 exports from one edit	Four formats

Slopcamera also composes vector icon scenes with the bundled icon.place library and reads Soundfish scores or MIDI files into beat grids for music-timed scenes, both locally and without a model. Neither renders audio.

Model-backed generation of images, video, and speech is available through your own provider access, and cinematic effects, particles, and character performance can be planned from source. The gallery has no rendered examples of these yet.

Blender, CadQuery, and Manim for coding agents

Slopcamera can direct Blender for sets, materials, lighting, skinned characters, cloth, and liquid caches; CadQuery for parametric solids and STEP; and Manim Community for mathematical animation. Seven editable starters include a product, character, shaded street, cloth, liquid, CAD bracket, and educational presenter. Open Rain, bottled, then change its weather, materials, and camera. For 2D motion, try the illustrated tram film. The portrait geometry lesson adds mathematical typesetting, a presenter, and captions.

Native source runs as the current user. Blender and Python environments install separately, and native source execution requires explicit current-user trust. See Native film studio for the starters and engine requirements.

Generate images, video, and speech

Discover image, video, speech, and transcription models through your own Vercel AI Gateway access, with AI_GATEWAY_API_KEY set in the local process environment. Generate images from text and references, add a voiceover, transcribe sound, or create short video shots. Availability and pricing come from the live model catalog. Without a Gateway key, prepaid Hraness Credits run prompt-only image generation through the hosted API (slopcamera ai image generate --hosted). See Generate media and Direct short generated clips.

Standalone image files can also use your own Vertex express, Google Gemini, or OpenAI key with slopcamera image generate --provider …. The scrolling pixel landscape guide compares continuous landscapes or linked image sections and produces a transparent pixel mask that can take a site's tint.

Edit footage and deliver every aspect ratio

Import existing footage or recording bundles. Remove pauses and filler words, align sound, reframe speakers, zoom into screen actions, and add captions, graphics, color, and audio treatment. Preview a candidate, then export clean and captioned versions in each format from the same edit.

Edit my product demo: cut the pauses, zoom into each important click, keep the speaker framed, add captions and logo.svg, and show a preview before export.

Slopcamera edits recordings you already have; it does not record the screen, camera, or microphone. Standalone media imports need an existing project; html render, studio assemble, or direct assemble can create one. Read PRIVACY.md before editing sensitive material. Eight built-in workflows package common jobs such as social cuts and talking-head cleanup:

slopcamera workflows list --json
slopcamera workflows show social-variants --json

Start with Edit a video or Run workflows.

How Slopcamera works

Each kind of work keeps its own source file. A native scene holds a rig or simulation; a portable scene holds geometry, cameras, and media placement; a diagram holds its shapes and labels; a video project holds cuts and delivery settings. Slopcamera connects them through files: a rendered diagram can appear on a screen in a 3D scene, and native frames can become a clip in a video project.

1. Prepare the sources. Import footage and assets, or author a scene, diagram, or native program from a starter.
2. Direct the result. Name cameras, shots, timing, composition, and output settings. Generate missing media only when the job calls for it.
3. Review a render. Inspect contact frames, motion, captions, sound, and continuity, then select the version to keep.
4. Deliver and revise. Export the formats you need. The sources and project decisions stay in place for the next change.

The same system is available four ways: the Bun CLI, the TypeScript SDK, the Agent Skill, and an MCP server (slopcamera mcp) with a smaller fixed toolset.

Instructions for coding agents

Read local project instructions and inspect sources before changing them. Use slopcamera --help, slopcamera doctor --json, slopcamera operations list --json, and the installed skill to discover the exact local commands. Agree on output requirements, preview substantial changes, inspect the resulting files, and report their paths. A successful plan does not show that provider access, native trust, or model quality is in place.

Slopcamera vs Remotion and HyperFrames

Details as of 28 September 2026.

	Agent writes	License	Cloud rendering	Choose it when
Slopcamera	Diagram JSON, HTML scenes, scene JSON, Blender, CadQuery, or Manim programs, and video edits	MIT	Video renders on your machine; a hosted API checks and renders diagrams and generates images	One agent needs diagrams, 3D, native films, and footage edits in one local project
Remotion	React components	Source-available; free for individuals, non-profits, and organizations of up to 3 people	AWS Lambda in your account	Your team writes React and renders at scale
HyperFrames	HTML, CSS, and JavaScript animation	Apache 2.0	HeyGen-hosted rendering or AWS Lambda	You want HTML motion graphics rendered to MP4

Remotion and HyperFrames have much larger communities. For a one-off image or clip, a hosted app needs no install. Read Slopcamera vs Remotion, Slopcamera vs HyperFrames, Remotion alternatives for coding agents, or Why Slopcamera.

Important limitations

• Runtime support varies. The CLI runs with Bun on macOS, Linux, and Windows. Media, browser, GPU, and native studio profiles need additional software, and vector tracing does not run on Windows. Run slopcamera doctor and see the capability reference.
• Checks do not judge quality. A passing check or finished render shows that the job ran, not that the result looks right. Models may also change a subject's identity, motion, or text; review generated media.
• Interchange keeps a supported subset. Exporting a GLB does not turn a native rig or simulation into an editable scene. An image or video on a plane supplies pixels, not geometry.
• Pixels can differ across machines. Native tools, codecs, GPU drivers, and provider models affect results, so the same source does not promise identical output elsewhere.
• Trusted code is not sandboxed. Native Python and your own Bun workflows run with the current user's access.
• MCP is a subset. Its 21 fixed tools check and render diagrams, plan and audit scenes, compose icon scenes and soundtrack beat grids, and run ten operation codes. It does not expose every CLI command and never changes project state.

Design and trust

There is no Slopcamera account or hosted project database. Video editing and rendering run on your machine; the optional hosted API checks and renders diagrams and generates images. Gateway generation and cloud analysis use credentials from the local process and ask for your acknowledgement before uploading named media. The optional hosted image API runs each paid call in a short-lived workspace against a Hraness Credits device token and keeps no credentials or projects; the token lives in owner-only files and never appears in command arguments or operation records. The website never accepts a Gateway credential.

Native Python requires separate authorization. Custom Bun workflow modules run when loaded, including during check and plan, so review their source first. Both run with the current user's access, including network access. Normal edit operations leave original media unchanged, and important operations record their inputs and outputs.

Read Architecture, SECURITY.md, PRIVACY.md, and NOTICE.md for details. Keeping the source beside the result is the design every Hraness project shares; The thread through hraness follows it across projects, and the ALGAL vision states the bet behind it.

Frequently asked questions

What is Slopcamera?

Slopcamera is an open-source CLI, TypeScript SDK, Agent Skill, and MCP server that lets coding agents make images, diagrams, animation, 3D scenes, and video. The agent writes a short source file for a diagram, HTML scene, 3D scene, native Blender, CadQuery, or Manim program, or video edit, and the slopcamera CLI renders it. A revision is an edit to that file and a new render.

Does Slopcamera work with Claude Code and Codex?

Yes. Install the CLI, then run slopcamera skill install --target claude for Claude Code, or slopcamera skill install for Codex. Other agents that read ~/.agents/skills use --target agents, and MCP clients can connect to slopcamera mcp.

Can a coding agent make AI video with Slopcamera?

Yes, through your own Vercel AI Gateway access: the agent can generate short video shots, images, and voiceover, and combine them with authored scenes and your footage. Local editing and rendering need no model account.

Can my agent drive Blender?

Yes. Slopcamera directs an installed Blender for sets, materials, lighting, skinned characters, cloth, and liquid caches, starting from editable starters. Blender installs separately, and native source runs only with explicit current-user trust.

Is Slopcamera a Remotion alternative?

It depends on the job. Choose Remotion when your team writes React components and renders at scale. Choose Slopcamera when one agent needs diagrams, 3D, native films, and footage edits in one local project. See the comparison above.

What does Slopcamera cost?

Slopcamera is free and open source under the MIT license. Generation uses your own Vercel AI Gateway account, or prepaid Hraness Credits for prompt-only hosted images, and model usage is billed separately.

Documentation

• Learn: Your first diagram · Your first native film.
• Make a result: Edit video · Generate media · Run workflows · Educational video · Plan camera moves, lighting, and effects for a 3D scene · Direct visual styles.
• Look up support: Techniques · Capabilities · SDK entrypoints · Creative tools.
• Understand the system: Architecture · Native studio · Directed scenes · Extension architecture.

The documentation index connects these paths, and slopcamera.com/docs publishes them.

Use the SDK

Run this example with Bun from the source checkout after bun run build:sdk. SDK imports do not start the CLI or read local project state. For example, convert an existing local image into an SVG:

import { vectorizeImage } from "@hraness/slopcamera"

const result = await vectorizeImage("logo.png", { outputPath: "logo.svg" })
console.log(result.receipt.sourceSha256, result.receipt.svgSha256)

Use @hraness/slopcamera/code for declarative workflow graphs, @hraness/slopcamera/workflow for trusted Bun workflows, and @hraness/slopcamera/local/* for the local media engine. See the SDK reference.

Verification

bun install --frozen-lockfile --ignore-scripts
bun run check

The check covers the public SDK, the local runtime, schemas, the Agent Skill, generated entrypoints, the website, deterministic and property tests, and installs of the packed package. Native engines and model providers need their own runs on real hardware and accounts; a local test pass does not cover them. See CONTRIBUTING.md for every check.

Contributing

Read CONTRIBUTING.md and the nearest AGENTS.md before changing a package or runtime boundary. Report vulnerabilities through SECURITY.md.

License

MIT, with third-party notices in NOTICE.md.

Optional support

Every feature works without payment. slopcamera support shows an optional way to support development and opens no browser. support dismiss opts out across the local suite, support snooze pauses the notice for 30 days, and HRANESS_SUPPORT_AUDIENCE=off disables ambient notices. Agents follow the support reference.