Skip to main content
Singularity Node is the desktop app for running AI on your own computer and, optionally, serving the Singularity Layer Grid. It bundles local chat, voice, and image generation with the node service described in Node setup. This page covers version 1.11.

Image & Video

The Image & Video tab makes images on your own Mac (Apple silicon). Everything is generated and stored locally.
  • Create. Pick a model, type a prompt, and the image appears in a chat-style thread.
  • Edit with words. Edit any picture by describing the change. Drag a picture in from Finder or paste one to start from it.
  • Upscale. 2x, 4x, 2K, 4K, or 8K.
  • Remove background. Cut-out and upscale can run while another image is being made.
  • Save as. PNG, JPG, HEIC, or transparent PNG.
  • Model pills. Choose the model and aspect ratio. Under advanced: images at once (1 to 4), steps, and seed.
  • Stop. Stopping a generation keeps your prompt and marks it “Stopped”.
The first use asks you to accept responsibility for local generation.

Library

Every image is kept in a Library on your Mac. It has search, filters, and favourites (star). You can select and delete several images at once. Hover actions on each image: edit, upscale, remove background, and save.

Models

Models are listed by what your Mac can run, and new models can arrive without an app update. Choose Keep the model ready (also in Settings) to keep it loaded so the next image starts right away. Windows and Linux with NVIDIA GPUs are coming soon. Video is next.

Serve images on the Grid

Any Mac that can run the image model can serve pictures on the Grid. The model’s own memory requirement is the only limit. Providers keep 80% of what a job earns. The default price is 0.0015 USD per megapixel-step, so a 1024x1024 image is about 0.0063 USD. You can set your own price. Open Pricing, pick the image model, and move the slider between half and double the default (0.5× to 2×). The app shows the price in dollars, such as about 0.0079 USD for a 1024x1024 picture at 1.25×, and what you keep. Reset goes back to the default. Changes can be made once a minute. Speech and transcription models can be priced the same way with the CLI or from the operator dashboard in Cloud. Buyers use the Grid image API, and all generation is subject to the content rules. Provider setup and earnings work as described in Earnings.

Recents

Recents holds your text and image chats. Image chats show an image icon and reopen where you left off.
  • Pin chats to keep them at the top.
  • Groups. Use Move to group or New group to organise chats into your own groups, and collapse groups you are not using.
  • Reorder. Drag chats to reorder them or move them between groups.

Pictures in chat

Vision models (Qwen2.5-VL 3B) accept pictures. Attach one with the picture button, paste it, or drop it into the chat. The picture is sent downscaled, and saved chats keep only a thumbnail.

Models and voice

Switching models runs in the background. Progress shows in the model pill, and a status indicator appears on every tab and on the Overview. Voice models added in 1.11: MOSS-TTS Nano, LongCat-AudioDiT 1B, VoxCPM2 2B, and Rumik 3B. Rumik 3B covers Hindi and 21 more Indian languages. It is for non-commercial personal use only and is never served to the Grid. Soprano 80M now appears in the model list as intended. For dictation, see Speech-to-text transcription.

Window behaviour

Closing the window hides it. Click the Dock icon to bring it back.