❓FAQs

Sogni is a creative platform built on a decentralized GPU network — the Sogni Supernet — with privacy and ownership at the core. On top of the network, Sogni Intelligence provides an OpenAI-compatible LLM service that can generate and edit images, video, music, and text. The Supernet and Intelligence are surfaced through 10+ SuperApps (Pocket, Studio, Create, Photobooth, Chat, 360, Makeover, Restore, Voice, Vibe, Gen) and the Sogni SDK for developers who want to build their own.

Pick what fits the moment:

• Just want to chat with AI? chat.sogni.ai — a conversational creative assistant.
• iPhone or iPad? Sogni Pocket on the App Store.
• Android? Sogni Create for Android.
• Mac desktop? Sogni Studio (App Store) or Sogni Studio Pro (from sogni.ai/download) — Pro adds Worker Mode and the full Web3 wallet.
• Any browser, nothing to install? app.sogni.ai and the SuperApps: photobooth.sogni.ai, 360.sogni.ai, makeover.sogni.ai, restore.sogni.ai, and others linked from sogni.ai/super-apps.
• Restoring old photos? Sogni Restore for iOS or restore.sogni.ai.
• Building your own app? The Sogni SDK and the Sogni Intelligence API.

Yes. Sogni is built with privacy and ownership at the core:

• With on-device processing, your prompts and images never leave your device.
• With Supernet processing, prompts and assets are encrypted and routed through permissionless workers who never see or store your content. Your wallet is only used to verify identity and reward activity.
• Your wallet is yours. You alone hold the private keys — Sogni never has access to your funds, account, or data.

@@HTML_BLOCK_0@@

Yes. Sign up at app.sogni.ai with just an email. Inside any Sogni app you can purchase Premium Spark Points with Apple Pay or any credit/debit card via Stripe — supported in 195 countries and 135 currencies. A Sogni wallet is created for you automatically; you never have to handle keys, gas, or chains. You can connect or import a wallet later if you want to use SOGNI tokens directly.

Jobs are priced in GPU-time. One Spark Point equals one base render job (an SDXL model at 20 steps), pegged to $0.005 of GPU-time on the Supernet — so 1 Spark = 1 base render, every time.

• Free Spark Points — earned from the Monthly Boost (400 Spark/month), referrals, email verification, and quests. 1× leaderboard multiplier.
• Premium Spark Points — see About Premium Spark Points. Holders are boosted to Premium Tier with access to premium licensed models (like Flux.1 Kontext and Flux.1 Krea), queue priority, full worker-targeting control, and a 10× Artist leaderboard multiplier.
• SOGNI tokens — fully transferable ERC-20 on Base and Etherlink. Pay-as-you-go for renders, claim worker earnings, stake for the Staking Leaderboard, and vote in future governance. 10× leaderboard multiplier when used to pay.

Sogni is pay-as-you-go — no subscriptions required. Optional Unlimited Plan subscriptions are rolling out.

See SOGNI Token vs Spark Points for the full breakdown.

Share your referral code (your Sogni username), or send a link in the format https://app.sogni.ai/create?code=<your-username>. After a friend signs up with your code and confirms their email, they get 25 Spark Points. After their first Spark purchase, you get 25 Spark Points too.

Referrals also feed the Ambassador Leaderboard — every Sogni user is enrolled automatically. Points accrue from purchases by you, your referrals, and the people they refer, with the leaderboard distributing one of the largest slices of each 6-week seasonal airdrop pool.

See Referral Program / Sogni Ambassador Rewards.

The Creative Assistant is the agentic AI layer in Sogni Studio, Sogni Pocket, and the Sogni Creative Agent. Tell it what you want and it picks models, writes prompts, runs multi-step workflows, and edits or analyzes media. It understands the Sogni UI, Supernet, your assets, and your safety filter settings.

• Agent Mode — chat-first. Tell it what you want and it assembles the workflow for you.
• Guided Mode — chat with controls. The assistant helps; you decide which settings to commit to.
• Manual Mode — the classic Sogni workflow, untouched.

The same brain is available at chat.sogni.ai.

• Try the Creative Assistant in Agent or Guided Mode — it picks the right model and settings from your description.
• Pick a model that matches the result. Flux Krea and Flux.2 for photoreal; Flux.1 Kontext for plain-English edits; Qwen Image Edit for fast precise tweaks; Chroma v46/v48 for stylized work. The Model Explorer has descriptions and sample settings.
• Iterate with edits instead of regenerating from scratch. Use Kontext for "make this red," PaintOver for "change only this region," and Auto-Refresh to stack edits fast.
• Adjust Steps and Guidance Scale. Higher steps add detail; a lower Guidance Scale (5–6) softens results for photoreal work.
• Use a Guide Image or ControlNet to give the model a reference.
• Check the Safe Content Filter in the Advanced tab.

Tutorial videos: youtube.com/@SogniAI. Prompt-writing tutorial.

Personas let you save people — yourself, family, friends, pets — once, and refer to them by name in any prompt. You can also define how they should sound in LTX-2.3 audio and add backstory the assistant uses when building scripts.

For one-off photoreal portraits without setting up a Persona:

• Use Flux Krea or another photoreal model.
• Add face-related keywords ("perfect face," "highly detailed face").
• Lower the Guidance Scale to 5–6 for softer lighting and skin.
• Turn off automatic upscaling in the Advanced tab — over-smoothing can flatten skin texture.

Tutorial: Creating Photorealistic Portraits in Sogni AI.

• Kontext — natural-language edits on a reference image. Best for "change the shirt to red," group-photo merges, two-image blends, and identity-preserving edits. Runs on Fast Supernet as a premium licensed model.
• PaintOver — paint directly over the region you want changed. Best for targeted edits where prompts aren't precise enough. Works alongside Guide Images, ControlNet, and Kontext.
• ControlNet — classic Stable Diffusion guidance with 14 CN models (Canny, Depth, InPaint, Instant-ID, Instruct Pixel2Pixel, LineArt, LineArt Anime, M-LSD, Normal Bae, OpenPose, Scribble, Segmentation, Shuffle, Soft Edge, Tile Resample) and preprocessors (Vision Face/Pose, Sketch, Depth Map, Segmentation, Subject/Background Mask, Invert). Best for pose transfer, depth maps, and structural reuse.

See ControlNet Models.

• Wan 2.2 — the default text-to-video and image-to-video engine.
• LTX-2.3 — cinematic video up to 2160p (4K) with synchronized audio (dialogue, SFX, music) in a single pass, roughly 5× faster per frame than Wan 2.2. Includes Depth, Canny, and Pose IC-LoRA controls.
• Motion Transfer — Animate-Move (apply a video's motion to your character) and Animate-Replace (keep the original motion, swap the subject). You can record movements directly from your Mac's webcam.
• Sound to Video — drive image motion from an audio track (music videos, lip-sync, ambient scenes).
• Change Perspective — re-render an image from any camera angle while keeping subject and style consistent. 96 distance/orientation/elevation combinations.
• Sogni 360 — turn a single image into a smooth 360° orbit video.
• Clip Mixer — combine clips with Fast Combine, High Quality, or Share-safe export modes.

Seedance 2.0 is on the way.

Images — most models generate at 1024 × 1024 px (or another aspect ratio at the same megapixel count). With upscaling enabled you can boost to 2048 × 2048 px. Sogni Studio offers three upscalers:

• Z-Image Upscale (Supernet) — best for diffusion-generated images, with strong structure and detail retention.
• Real-ESRGAN (on-device) — fine photographic detail.
• Real-ESRGAN Anime (on-device) — illustrations and line-art.

Video — Wan 2.2 outputs standard HD resolutions; LTX-2.3 supports up to 2160p / 4K with embedded synchronized audio.

Yes, for on-device processing in Sogni Studio. Import any Stable Diffusion model that has been converted to Core ML — downloaded from the web or trained by you — via File → Import S.D. Core ML model. See Importing Stable Diffusion Models. The Supernet manages its own model catalog and does not accept end-user uploads.

• Sogni Pocket — swipe up to save to your Sogni Gallery, or tap the share button → Save Image (Photos) / Save to Files.
• Sogni Studio / Studio Pro — File → Save, the save button in the main window, drag-and-drop from the gallery, or right-click → Save Image / Save a copy.
• Sogni Create (web) — download from the result card. The Supernet supports both JPG (5–10× smaller, great for batch reviews) and PNG (lossless); toggle in the Advanced menu.
• SuperApps (Photobooth, 360, Makeover, Restore, …) — built-in download and share controls on each result.

The filter detects images that may contain potentially disturbing content — graphic violence, nudity, sexual, vulgar, or NSFW content. If detected, the image isn't displayed and you'll see an error instead. You must be 18+ to turn the filter off.

Sogni Intelligence is an OpenAI-compatible LLM service running on the Sogni Supernet. The default model is Qwen 3.6 35B-A3B (qwen3.6-35b-a3b-gguf-iq4xs) with a 262K-token context window, reasoning, and tool calling.

• Base URL: https://api.sogni.ai (requests go to /v1/...).
• Pricing: roughly $0.30 per 1M input tokens and $0.90 per 1M output tokens, billed in Spark or SOGNI.

To get a key:

1. Sign up at app.sogni.ai.
2. Open dashboard.sogni.ai/api-key.
3. Use the eye icon to reveal it, copy it, and send it as Authorization: Bearer YOUR_API_KEY.

The same key works with the Sogni SDK and Fast Worker. See the Sogni Intelligence Introduction and the LLM API reference.

Yes — three paths:

• Sogni Intelligence API (/v1/chat/completions) — OpenAI-compatible. Any client that supports a custom OpenAI endpoint works: Open WebUI, OpenClaw, Hermes Agent, the OpenAI SDK, and agentic coding clients. Sogni auto-injects six built-in media tools (sogni_generate_image, sogni_edit_image, sogni_generate_video, sogni_sound_to_video, sogni_video_to_video, sogni_generate_music) so the model can produce media inline.
• Sogni Client SDK (@sogni-ai/sogni-client on npm) — TypeScript/JavaScript SDK for Node and browser environments with WebSocket-based job processing for images, video, LLM chat, and tool calling.
• n8n — drop-in nodes for no-code automations. See Sogni n8n Plugin.

• Sogni Fast Worker (image/video) — Docker-based, runs on Windows or Linux with a supported Nvidia GPU (≥16 GB VRAM, including RTX 5090/4090, H100, RTX Pro 6000; full list in the Fast Worker docs). Install from api.sogni.ai/v1/download/worker/windows or api.sogni.ai/v1/download/worker/linux. A high-end RTX 5090 can complete up to ~10,000 image jobs per day at peak demand.
• Sogni LLM Worker (Intelligence) — powers Qwen 3.6 and other LLMs. Requires ≥24 GB VRAM. Earnings track token throughput at roughly $0.30/M input tokens and $0.90/M output tokens. Use the same Fast Worker installer.
• Worker Mode in Sogni Studio Pro (macOS) — a Mac-friendly entry point for sharing idle compute.

For requirements, configuration, and the full Fast Worker FAQ, see the Fast Worker section.

Yes. Whether you're running Worker Mode in Sogni Studio Pro (macOS), the Sogni Fast Worker (Windows/Linux Docker), or the Sogni LLM Worker, your machine processes encrypted jobs in a sandboxed environment, isolated from the rest of your system. Your personal files, prompts, and data are never exposed. You can pause or stop working at any time, and earnings accumulate in your Sogni account.

• Web (any browser) — app.sogni.ai plus all SuperApps (chat, photobooth, 360, makeover, restore, and others).
• macOS — Sogni Studio (App Store) and Sogni Studio Pro (sogni.ai/download). Pro adds Worker Mode and full Web3 wallet features.
• iOS / iPadOS — Sogni Pocket and Sogni Restore on the App Store.
• Android — Sogni Create for Android. Sogni Photobooth for Android is rolling out.
• Windows / Linux — Sogni Fast Worker and Sogni LLM Worker via Docker.

See sogni.ai/download for current install links.

Native macOS / iOS apps support both on-device and Supernet processing:

• macOS Sonoma 14.0+ — Apple silicon; 16 GB RAM recommended.
• iOS 17+ — iPhone 13 Pro or later, or iPad with Apple silicon.

Older devices may be unable to compile the Core ML on-device models, but can still use the Supernet through any browser or native app.

Fast Worker — Windows or Linux with Docker and a supported Nvidia GPU (≥16 GB VRAM). LLM Worker — ≥24 GB VRAM.

Web and SuperApps — any modern browser, no install required.

@@HTML_BLOCK_1@@

SOGNI is an ERC-20 token on Base, with availability extended to Etherlink. The Supernet uses Etherlink as the default layer for faster, lower-cost transactions. Workers can bridge their Worker License NFT from Base to Etherlink with nft.sogni.ai and earn a +2% bonus on settlements while it's on Etherlink. Bridging is powered by LayerZero V2 and typically takes 2–5 minutes.

See Etherlink NFT Bonus and SOGNI Token vs Spark Points.