Slopify

Slopify turns a prompt and a handful of keywords into a narrated slideshow video. It researches the topic, writes the article, narrates it, generates the imagery and cuts the result together with ffmpeg. Everything runs on your machine against your own provider keys, behind a local web UI.

npx @gentbajko/slopify@latest

The app serves its UI on http://127.0.0.1:6969 and opens a browser at it. Node 26 or newer, nothing else to install.

How to use it

1. Run the command and open the tab it prints.
2. Settings holds your provider API keys and your voices. They live in ~/.slopify/slopify.db and go to the provider you picked, nowhere else.
3. Prompts is where the article, image and thumbnail prompts are written once, with {{keywords}} where the subject goes.
4. Play is the cue sheet: pick the prompts, fill in the keywords, choose a voice and a format, press the key. Any stage you would rather do yourself, you upload instead, and that stage is skipped.
5. The project page shows the six stages as they run and holds the results: the article, the narration, every image, the thumbnail and the mp4, each one downloadable on its own or as a zip.

Options

Every option is a flag or an environment variable of the same name. A flag beats the variable, the variable beats the default.

Flag	Variable	Default	Does
--port	SLOPIFY_PORT	6969	The port the app listens on. An integer from 1 to 65535; anything else refuses to start.
--host	SLOPIFY_HOST	127.0.0.1	The address it binds to.
--data-dir	SLOPIFY_DATA_DIR	~/.slopify	Where slopify.db, projects/, staging/ and logs/ live. Relative paths are resolved against the working directory.
--no-open	SLOPIFY_NO_OPEN	the browser opens	Keeps the browser shut. As a variable, any value other than empty, 0 or false counts as set.
no flag	SLOPIFY_FFMPEG	the bundled binary	Path to an ffmpeg to render with.

There is no login. Binding to anything but 127.0.0.1 prints a warning at startup and means it: whoever reaches that port controls the app and the keys in it.

SLOPIFY_PORT=5000 SLOPIFY_NO_OPEN=1 npx @gentbajko/slopify@latest

ffmpeg and its licence

The video is cut by ffmpeg. Slopify installs one through ffmpeg-static, which fetches a platform binary at npm install time, and never falls back to an ffmpeg on your PATH: the binary that ships is the binary that is tested. Point SLOPIFY_FFMPEG at another build to override it.

SLOPIFY_FFMPEG=/usr/bin/ffmpeg npx @gentbajko/slopify@latest

That binary is a separate program, run as a child process with an argument array. It is licensed under the GPL-3.0-or-later. Slopify does not link against it, does not embed it, and does not distribute it inside the slopify package; ffmpeg-static downloads it to node_modules/ffmpeg-static/ at install time, and its licence text and the location of its corresponding source ship there beside it. Slopify's own code is MIT and stays MIT. Anyone redistributing the downloaded binary takes on the GPL's obligations for it, including offering that corresponding source.

Telemetry

Slopify sends anonymous counters to a collector: installs, projects created, stages completed, images and videos made, audio seconds, and provider token counts. Never your keys, prompts, keywords, titles, article text, filenames, or anything about your machine beyond a random id created the first time you run it. The counters on slopify.stream are the sum of those events. Deleting the data directory makes a fresh install with a new id.

Supporting the project

Slopify is free and always will be. If it is worth something to you:

• Patreon
• Buy Me a Coffee

Working on it

Requires Node 26 or newer, which is what CI runs.

npm install
git config core.hooksPath .githooks   # once per clone, wires the pre-commit hook
npm run lint
npm run typecheck
npm test
npm run build

core.hooksPath is local repository configuration and cannot be committed, so a fresh clone has to set it by hand or the pre-commit check will not run.

The workspace holds four packages. packages/app is the CLI and the HTTP server, packages/web the SPA it serves, packages/site the marketing page, and packages/collector the telemetry Worker.

To run the collector against a real local D1 database:

cd packages/collector
npm run schema:local        # applies schema.sql to the local D1 copy
npm run dev                 # wrangler dev on http://127.0.0.1:8787

The marketing page talks to http://127.0.0.1:8787 whenever it is served from a loopback origin, so a local collector is all it takes to see the counters move. wrangler dev drops the connection a few seconds after a browser starts polling it (wrangler 4.128.0, Error in ProxyController: Network connection lost.); curl is unaffected, and serving the same responses from any small local server is enough to look at the page.

Deploying to Cloudflare

Both packages/site and packages/collector deploy with wrangler. Nothing here needs a secret: the collector's only binding is its D1 database, and the page has no server side at all.

Before the first deploy, the zone slopify.stream has to be on the Cloudflare account, because both wrangler.jsonc files claim a hostname on it. The app posts to https://collector.slopify.stream and the collector answers CORS for https://slopify.stream only, so both names are part of the build.

npm install
npx wrangler login                                   # once, opens a browser

Then the collector, in order:

cd packages/collector
npx wrangler d1 create slopify-collector             # prints database_id

That command prints something like database_id = "b1f0…". Paste it into packages/collector/wrangler.jsonc in place of D1_DATABASE_ID_PLACEHOLDER, then:

npx wrangler d1 execute slopify-collector --remote --file=schema.sql
npx wrangler deploy
curl https://collector.slopify.stream/aggregates     # {"aggregates":{...}} with zeroes

Then the site:

cd ../site
npx wrangler deploy

After that first time, both deploys are one command from the repository root:

npm run deploy:check      # builds and validates both, offline, no account needed
npm run deploy            # collector first, then the site

The order is not arbitrary: the site's counters read from the collector, so deploying the collector first means the page never goes live pointing at nothing.

There is deliberately no npm-publish script. .github/workflows/release.yml publishes on a v* tag and signs a provenance attestation with the workflow's own OIDC identity, which a publish from a laptop cannot do. Releasing is npm version, a commit, a tag and a push.

The database id is the one value that cannot be committed ahead of time: it is generated by d1 create and is specific to one Cloudflare account. Everything else in both configuration files is final.

Licence

MIT. See LICENSE. The ffmpeg binary fetched at install time is a separate GPL-3.0-or-later program, as described above. The bundled Barlow and Barlow Condensed faces under packages/site/public/assets/fonts/ are SIL Open Font License 1.1; the licence text ships with them as OFL.txt.