Sogni Pocket
Sogni Pocket is the native iPhone and iPad app for the Sogni Supernet. Generate images and short videos from your phone, swipe through results, and pick up where you left off on web or Studio — it's one account, one gallery, one balance across every Sogni Create surface.
#Get Sogni Pocket on the App Store →
#What it does
- Mobile-first creation. Type a prompt, pick a style, and render — the full text-to-image and image-to-video toolkit, sized for one-handed use.
- Camera roll as input. Pull any photo from your library (or shoot one fresh) into image-to-image, ControlNet, Paintover, or video generation.
- Swipe to browse. Swipe gestures move you through results — left for the next render, right for history, up to keep, down to discard.
- Same Supernet as everywhere else. Jobs run on the distributed Sogni worker network, so a phone renders SDXL, SD3, Flux, and video models without straining the device.
- Background queue. Long batches keep rendering when you switch apps — results land in the gallery, ready to review later.
- On-device fallback. Optional CoreML pipeline for offline generation when you'd rather not spend Spark.
#What it ships with
- SwipeFlow — pre-rendered batches so swiping never waits on a queue.
- Paintover — mask directly on an image to inpaint specific regions.
- Guide Image — feed in a reference image to steer composition and color.
- ControlNet — sketch, pose, depth, and face capture for fine-grained structural guidance.
- Image-to-video, sound-to-video, and motion transfer for short clips.
#Account and balance
Sign in with your existing Sogni account and your Spark balance, gallery, prompts, and favorites are already there. New users can create an account in-app from the same flow. See Downloading Sogni Pocket for setup steps.
#Workflows
Idea sketch on the go. Start a SwipeFlow batch in a style category, swipe through 20 renders during a coffee break, favorite the keepers — pick them back up at the desktop later.
Selfie to portrait. Open the camera roll, pick a recent photo, set it as a Guide Image at lower strength, and generate stylized variants without losing the subject.
Targeted edit. Load a generated image, tap Paintover, mask the area you want to change (hair, clothing, background), and re-render only that region.
Reference-driven render. Drop a pose or sketch into ControlNet to lock structure, then iterate on style and lighting through the prompt.
#Tips
- Use SwipeFlow with Fast Supernet. It pre-generates ahead so the swipe gesture stays instant; minimum batch size is 8.
- Lock the seed when you find a winner. Then shuffle it once if quality degrades on the next iteration.
- Save big batches for the background queue. Anything over 16 images auto-saves to the gallery without blocking the UI.
- Switch surfaces freely. Start in Pocket, finish on web or Studio — the gallery syncs both ways.