Generate pictures from a description, edit the ones you already have by describing the change, and turn a whole set into a narrated video — without an account, a subscription, or a single upload.
Everything runs on your own hardware, so the work stays yours from the first prompt to the finished video.
Write what you want and Ember renders it. Pick widescreen, vertical, square or photo proportions, and reuse a seed when a set needs to stay consistent.
Bring in up to three reference images and say what should be different. Restyle a photo, swap a background, or move a subject into a new scene. No masks or layers.
No sign-in, no telemetry and no upload. Your prompts, references and finished pictures stay on the machine that made them.
Turn a topic into a production plan, organise the shots into projects, and let the queue render the lot while you do something else.
Write the narration for each shot. Ember records it in a natural voice and holds every image on screen for exactly as long as its line takes to say.
Every picture keeps the prompt, size, seed and model behind it. Search past work and reuse any result as the start of the next one.
Every screen below is the real application, at the size it opens on your desktop.
Three steps from an idea to a finished set.
Download a compact starter set from inside the app, or point Ember at model files you already keep on disk. Your files are read where they are, never copied or altered.
Write a prompt for a single picture, or turn a topic into a full production plan and import it as a project.
Run the queue, then save the images wherever you like — or add narration and export the whole project as a video.
How to plan a narrated video, write prompts that land, and keep your work on your own machine — useful whichever tools you end up using.
Script first, one shot per beat, record before you build. The order that actually works, and the six mistakes that make this format feel cheap.
Prompting Why your AI image prompts don't give you what you picturedWhat goes wrong with adjectives, negatives and long prompts — and the word order that stops your subject getting buried.
Video How to write subtitles people actually readHow much text at once, how to time captions without transcribing, and why Chinese needs a different rule from English.
Privacy What actually happens to your images in cloud AI toolsWhat leaves your computer, the five questions with real answers, and how to check whether a "local" tool really is one.
The things worth knowing before you install.
No. Generation and editing happen entirely on your own computer, and there is no telemetry or analytics. The one exception is the narration used by video export: the spoken voices are fetched over the internet when you export a video. Everything else works with no connection at all.
No. There is no sign-in, no registration and no profile. Install the app and start working.
No. Ember does not meter what you make. Because generation runs on your own hardware, there is nothing to bill per image.
Ember uses the inference engine from stable-diffusion.cpp together with a model, and expects both to be present on your computer. The app does not download the engine itself — every screen tells you exactly what is missing and what to do about it. A graphics card with enough memory makes generation dramatically faster. Video export additionally uses ffmpeg.
Yes. Register any model file already on your disk and Ember reads it in place. Your files are never copied, moved or modified.
Each shot carries the narration spoken over it. That narration is recorded first, and the image is then held on screen for exactly as long as its own line takes to say, so picture and voice line up without any manual timing.
In a folder in your user profile, along with the small database that records the prompt, size and seed of each one. You can export any picture, or a whole selection, to any folder you choose.
Email support@iamdev.me. Questions, bug reports and feature suggestions are all welcome.