Skip to content
OpenPromo
Get started

Open source · MIT license

Open-Source AI Video Generator

Reels, promo videos and tutorials built from your website, with AI voiceover, music and captions. Rendered on your own server, with code you can read.

Get startedSelf-host with Docker

Examples

Open-source video generator examples

Stills from a reel and a promo video made for Crumb & Co, a fictional bakery. In the videos, captions appear word by word over each scene.

  • Story / reel for Crumb & Co, a fictional bakery, in the bold look: "Sourdough baked before sunrise" with an "Order for pickup" button
  • Story / reel for Crumb & Co, a fictional bakery, in the bold look: "Pick your loaves in two taps" with the bakery’s order page in a device frame
  • Story / reel for Crumb & Co, a fictional bakery, in the clean look: "Why Crumb & Co" with three numbered reasons
  • Story / reel for Crumb & Co, a fictional bakery, in the clean look: "Fresh bread, every morning" with an "Order for pickup" button
  • Landscape video for Crumb & Co, a fictional bakery, in the clean look: "Sourdough baked before sunrise" with an "Order for pickup" button
  • Landscape video for Crumb & Co, a fictional bakery, in the clean look: "Pick your loaves in two taps" with the bakery’s order page in a device frame
  • Landscape video for Crumb & Co, a fictional bakery, in the clean look: "36h" in large type over "Slow fermentation"
  • Landscape video for Crumb & Co, a fictional bakery, in the bold look: "Fresh bread, every morning" with an "Order for pickup" button
A 32-second reel made by OpenPromo for Crumb & Co, a fictional bakery: voiceover, music and word-by-word captions.

What it makes

AI video for reels, promos and tutorials

  • Reels

    9:16 videos for Instagram Reels, TikTok and YouTube Shorts, with a hook in the first two seconds and word-by-word captions.

    AI reel generator

  • Promo videos

    16:9, 1:1 or 9:16 videos about your business, written from your selling points and shown with your own screens.

    URL to video

  • Tutorials

    Getting-started videos with one step per scene, from your product’s own screenshots, narrated.

    Product demo video maker

How it renders

How the video generator renders

No hidden model makes the whole video. Each step is plain, open code you can run and change.

  1. 1

    A storyboard, not a black box

    Each video is a list of scenes: layout, on-screen text, picture, narration and duration. You can edit every field.

  2. 2

    Checks before rendering

    A missing hook or call to action, voiceover pace, reading time, hype words and figures not found on your website are flagged.

  3. 3

    Scenes as HTML and CSS

    Ten scene layouts are rendered as web pages in headless Chrome, with your brand fonts and colors.

  4. 4

    Video with Hyperframes and FFmpeg

    Hyperframes (Apache-2.0) turns the scenes into frames, and FFmpeg adds voice, music and captions and writes the MP4.

  5. 5

    Files you keep

    Each render gives an MP4, a poster thumbnail and an SRT caption file, saved in your library.

Where AI fits

AI video layers, never AI text

Garbled lettering is the usual giveaway of AI video. OpenPromo keeps text out of the models’ hands.

What AI makes

  • Reads your website and writes the script
  • Voiceover for each scene
  • A music bed
  • Optional backdrop pictures
  • Optional B-roll clips

What AI never does

  • Draws your text: headlines and captions are real text in your fonts
  • Invents figures: claims not found on your site are flagged
  • Changes your logo or colors

Self-host

Self-host the video generator

The Docker image includes Chrome, FFmpeg and the fonts it needs. One command starts everything.

  • Docker and about 2 vCPU and 4 GB of memory for video rendering
  • Media files in a Docker volume (MEDIA_DIR), backed up like any folder
  • One render at a time by default; raise RENDER_CONCURRENCY on bigger servers

Self-hosting guide

Terminal
git clone https://github.com/duckduckdeploy/openpromo.git openpromo
cd openpromo
cp .env.example .env
# In .env, set two secrets:
#   AUTH_SECRET=     any 32+ random characters
#   ENCRYPTION_KEY=  output of: openssl rand -base64 32
docker compose up -d --build
# Dashboard http://localhost:3200 · site :3101 · API and MCP :4100

What it costs

AI video generator cost per video

OpenPromo charges nothing. AI calls go to your own YepAPI key, at YepAPI’s prices at the time of writing. A 30-second reel with voice and music is usually $0.15 to $0.50.

What video actions cost
ActionPrice
Write a videoScenes, on-screen text and narration$0.02 – $0.08
VoiceoverPer sceneabout $0.01
Music bedPer video$0.08
AI pictureOnly if used$0.04 – $0.20 each
AI video clipOnly if used, per secondabout $0.11
RenderingOn your server or the hosted app$0

Generated pieces are kept, so rendering again after an edit only pays for what is new. Costs and limits

Your AI key

Start with $5 of free AI credit

The AI work in OpenPromo (reading your site, writing, voices, music and AI pictures) runs on your own YepAPI key. Sign up at YepAPI and get $5 in free credit, for a limited time. That is enough to try every tool.

  1. 1

    Sign up at YepAPI

    New accounts get $5 in free credit, for a limited time.

  2. 2

    Copy your API key

    It is on your YepAPI dashboard once you sign in.

  3. 3

    Paste it into OpenPromo

    In Settings. Set a monthly spending limit there too.

Sign up at YepAPISetup guide

The free credit is YepAPI’s limited-time offer and may end. After it, you pay YepAPI per use at its price. OpenPromo logs every call with its cost, and a monthly limit caps what you spend. See pricing

FAQ

Open-source AI video generator FAQ

Is OpenPromo really open source?

Yes. The app is MIT licensed, and the video renderer it uses, Hyperframes, is Apache-2.0. Everything runs on your server: headless Chrome, FFmpeg, Postgres and the app itself.

Is it an open-source video generator without AI?

Partly. The layouts, text and rendering are plain templates that need no AI. AI writes the script and makes the voice and music, and optional pictures and clips, through your own YepAPI key.

Which AI models does it use?

Models for writing, voice, music, pictures and clips are called through YepAPI with your own key. Each call is logged with its price in Settings and usage.

How long does a render take?

It depends on the length of the video and your server’s CPU. Renders run in the background with a progress bar, so you can keep working.

Your brand, your server, your media.

Start online in a minute, or run OpenPromo on your own server. It is open source either way.

Get startedSelf-host it