Sogni Create, step by step: from your first prompt to images, video and music

The complete walkthrough of the app.sogni.ai web app. How the home screen works, how to generate your first images in seconds, how to pick models and styles, how to compare models side by side, how to create video and music, and how Spark points and the Supernet power all of it.

The fastest way to create with Sogni

Already use Claude, Codex or Hermes? Start here.

If you already have a powerful personal agent, the Sogni Creative Agent Skill is the easiest and most capable way into the platform. Ask in plain English and your agent plans the job, generates the assets, and delivers finished images, video, and music straight to your workspace.

npx setup-sogni-agent-skill Copy

Explore Creative Agent ↗ No agent? Continue with the guide ↓

1 prompt · 30 assets generated in moments

Brand new to AI? Start even simpler

Never made an AI image before? Try Sogni Vibe.

If AI tools are completely new to you, Sogni Vibe is the friendliest place to begin. There is nothing to install and no settings to learn: type one line describing what you want to see, or just say it out loud, and Vibe carries it through four guided steps:

1. 1Intent Wizard · describe your vision
2. 2Visualizer · facet your prompt
3. 3Confirm · review combinations
4. 4Studio · generate & refine

The wizard does the prompt engineering for you, so your idea stays the focus. Images stream in live from several models at once; tap the ones you like and Vibe refines the idea for the next round, until the result matches the picture in your head.

You don't even need an account to try it, since guest mode is free. A free Sogni account unlocks every model, larger batches, higher resolutions, and turning your favorite stills into video.

Try Sogni Vibe ↗ Read the Vibe docs ↗

First, which Sogni should you use?

Sogni is not a single app. It is a family of apps sharing one account, one gallery and one Spark balance, all rendering on the same Sogni Supernet, so whatever you make in one shows up in the others. Here is the lineup, and then the short answer on where to start.

Windows and Linux builds are on the way and have a waitlist on the Super Apps page.

1Create your free account

Head to app.sogni.ai. You can browse freely, but the first time you hit a generate button you will be invited to sign up. It is free, takes under a minute, and comes with free Spark points so you can start creating right away, plus welcome rewards you can claim later (see step 18).

2The home screen

The home page greets you with "Your imagination, unlocked" and a row of tool cards. Every card is a creative mode: click one and you land directly in the right workspace, with no setup needed. The All / Image / Video / Audio filter narrows the cards to one category, and the two icons on the right switch between the carousel and the grid layout.

The left sidebar is your permanent navigation, top to bottom: Home, Create (images), Video, Music, Recent Projects, Wallet, Rewards and Help, which opens docs.sogni.ai. The moon icon at the bottom switches between light and dark theme.

3Every tool at a glance

Switch to the grid layout using the right-hand icon above the cards and the whole toolkit is visible at once. Sogni Create is far more than an image generator: nine creative modes live in one app.

4Models, apps and community

Scroll down the home page and you will find three sections worth knowing before you start creating.

Featured Models is a curated shelf of new open-source and proprietary models: Krea 2 Turbo (fast images with strong aesthetics), Krea 2 Identity Edit (identity-preserving edits), Seedance 2.0 (the top proprietary video model, now in 4K), One Obsession (glossy 2.5D anime), Dark Beast (uncensored fine-tunes), Chroma, Flux Krea, Wan 2.2, LTX-2 and more. The full catalog lives at sogni.ai/models. A green "New" badge marks fresh arrivals, and a Premium badge marks the ones billed separately (see step 9).

More from Sogni collects companion apps built on the same engine: Sogni Infinity (seamless tiles that repeat forever), Sogni Chat (an AI agent that creates images and video inside a conversation), Sogni 360 (cinematic 360 degree orbital video from one photo), Photobooth (dozens of styled portraits from a selfie), Photo Restore (bring scratched and faded photos back to life), Sogni Makeover (try hairstyles and makeup instantly), the Creative Agent Skill (give your own agent Sogni image and video generation as tools) and API Access (render on the Sogni Supernet from your own product via SDK or API).

Made with Sogni is a live feed of community creations, filterable by Images, Videos and Audio. It is the best source of inspiration and ready-made prompts: click any piece to see how it was made.

5The Create workspace

Sidebar > Create, or the "Text to Image" card

This is the heart of the app. The settings panel sits on the left and your results fill the canvas on the right. Everything needed for one generation reads top to bottom, in the order you actually make decisions: prompt, cost and batch, style, model, then the finer settings.

Three ways to write a prompt

1. Type your own. Describe subject, mood, lighting and style. The prompt used throughout this guide shows what the generator can do with color:

2. Let the AI Assistant write it. With the toggle on, the app expands a short idea into a rich prompt automatically. The wand icon beside the box does the same on demand ("Enhance prompt with AI"), and the cross icon clears the prompt.

3. Borrow one. Click Use prompt on any inspiration card below the workspace and its full prompt lands in your box, ready to run or remix.

6Cost, speed and batch explained

The row right under the Imagine button holds the four controls that decide what your generation costs and how fast it arrives. Here is what each one means.

Control	What it means
	Estimated cost of this generation in Spark points, always shown before you click. It updates live as you change batch size, model, resolution or processing speed. The blue info button opens the Generation payment panel with the exact cost, your balance, the SOGNI Token and Spark Points switch, billing history and the Unlimited upgrade.
	Processing speed, in other words which Supernet renders your job. The plain icon is the Relaxed Supernet, which is cheapest. The lightning icon is the Fast Supernet, which is priority processing at a higher price. See the box below.
	Batch size, meaning how many images render from one prompt at once. Batch••• opens Mega Batch for even larger runs. If you have several models selected for comparison, the panel spells out the math for you, for example "1 image per selected model or variation x 2 selections = 2 images total."

The price difference on the same batch, with only the speed toggle changed:

7Styles

The Style section applies a curated aesthetic on top of your prompt. Click it to open a searchable library of more than 130 ready-made styles: Abstract, Anime, Architectural, a full run of art periods (Abstract Expressionism, Art Deco, Art Nouveau, Baroque and more), photographic looks, neon looks and beyond. Every style shows a thumbnail plus the exact keywords it will append, so you always know what it is doing, and styles you have used recently stay pinned at the top under "Recent history".

Once you pick a style, its keywords land in an editable box directly underneath, so you can tweak or delete any of them before generating. For the holographic butterfly prompt in this guide Magic Neon is a natural match; below is Art Period: Art Nouveau with its own recipe loaded.

8Choosing a model

The Model section opens the full library, with search, category filters and a description of every model alongside sample images. Categories run from Next-Gen Diffusion (Krea 2 Turbo, Dark Beast, Chroma, Flux, Qwen Image, Z-Image) through Image Editing & Mixing (Krea 2 Identity Edit, GPT Image 2, Flux 2, Qwen Image Edit) to a deep SDXL shelf with dozens of community fine-tunes.

The default, Krea 2 Turbo, is a great starting point: a distilled, few-step version of the Krea 2 image model, tuned for fast text-to-image and image-to-image with strong aesthetic quality. It runs in roughly 8 steps and accepts a guide image.

9Which models are included and which cost extra

Not every model is billed the same way, and the badge next to the model name tells you which is which.

Open-source models running on the Supernet carry the small green Sogni mark. These are rendered by the decentralized GPU network, they are the cheapest option, and they are the ones covered by a Sogni Unlimited subscription under fair use. This group is large: on the image side it includes Krea 2 Turbo, Chroma, Chroma Flash, Flux [schnell], Flux Krea, Qwen Image, Z-Image, Dark Beast and the entire SDXL shelf, and on the video side it includes LTX 2.3 and Wan 2.2.

Premium commercial models carry a Premium badge. These are proprietary third-party models called through their vendor APIs rather than rendered on the Supernet, so they are billed separately and need Premium Spark, which you top up with a card. In the video picker these are Seedance 2.0 by ByteDance and HappyHorse 1.1 by Alibaba; on the image side the same rule applies to proprietary editors such as GPT Image 2. Every model has a page with prices and API IDs in the model catalog.

10Compare models side by side

This is the feature that sets Sogni apart from most generators. Under the model field there is a + Compare another model / variation button. Add a second model (or a third, or a different version of the same model) and one click renders the same prompt with the same starting seed on all of them, side by side. No more generating separately and squinting between browser tabs.

Here is the result: the identical prompt rendered by Krea 2 Turbo and FLUX.1 Krea [dev]. Each tile is labeled with the model that produced it, so the comparison is unambiguous.

Give each model its own settings

Every compared model has a sliders icon next to its name that opens a Customize panel. There you can override the Prompt, Style, Steps, Sampler and Scheduler for that model alone, and depending on the model also Avoid (the negative prompt) and Guidance. Leave everything on Auto and each model uses its own recommended values, which is almost always what you want. The gray text under each label tells you the range that model actually accepts, and it varies a lot: one engine will list steps 1 to 40, another 20 to 50 with guidance 1 to 7. If any of those names are unfamiliar, step 11 explains what each one does.

11Size, seed, steps and the advanced settings

Below the model sit the precision controls. Scroll the panel to the bottom and you can see all of them at once.

What the Advanced settings actually do

The Advanced section is the one place in the app where the labels assume you already know diffusion vocabulary. You can safely leave every one of these on its default, but here is what each is for, since these are also the settings you can override per model when comparing (step 10).

The LoRA library

Click Browse LoRAs and a library opens that is worth exploring on its own. LoRAs are grouped by what they do: Prompt Control, Art Direction (Aberrant, Amateur, Candid, Mystic X, Realism, Realism Engine) and Character (Age, Breast Size, Skin Tone, Weight, Wetness), with more categories further down. Each one shows who made it, a plain-language description of the effect and a link to its Civitai page.

The important control is Strength. Every LoRA comes with a recommended range, for example +0.25 to +1.2, and the slider runs both ways: a negative value applies the opposite effect and zero does nothing. So a "Realism" LoRA at minus strength pushes the image away from realism. You can attach several at once, the counter at the top of the list tracks how many are active, and the Use this prompt link next to the examples loads the prompt that produced the sample images.

You do not have to guess what a given strength will look like, because the Examples row shows the same scene rendered across the whole slider. Below, the Realism LoRA is displayed at Off, then at -1, -0.5, +0.5 and +1, and you can watch the shot travel from flat illustration at the negative end to a fully photographic frame at the positive end. That is what "a style steering wheel" means in practice.

Once a LoRA is attached it appears in the Create panel with its own strength slider and the recommended range printed underneath, so you can adjust it without reopening the library. The button below turns into Manage LoRAs with a counter such as 1/5, which tells you how many are active and how many the model will accept at once. Inside the library, the attached one is marked with a dot in the list and the button on the selected LoRA changes from Add to Remove.

12Hit Imagine

Click Imagine and the workspace fills with tiles, one per image in your batch. Each tile shows a live step counter, an ETA and the name of the Supernet worker rendering it. You can cancel at any point while it runs.

A few seconds later the batch is done. The neon prompt with the Magic Neon style on Krea 2 Turbo delivered four variants: one idea, four takes on composition and color.

13Working with a finished image

Click an image to open the full-screen viewer. The action bar covers the essentials: Perspective, Enhance, Make a Video and Save, and the side arrows flip through the batch.

The three-dot menu in the corner of the image holds the full toolkit. Every entry is a different way of continuing from the picture you just made.

14Video: models and workflows

Sidebar > Video

The four video models

Open the Model selector in the Video section and you get four engines, each with a different strength.

The workflows

The Workflow type selector at the top of the panel switches between six ways of making a video. The three core ones are text to video, image to video and video to video, and the rest are specialized variants of the last.

Text to Video (T2V). You start from nothing but words. Write a flowing paragraph of four to eight sentences covering scene, action, character and camera movement in cinematic terms, and the model builds the whole shot. This is the most creative and the least predictable of the three: you get total freedom over the world you describe, but no control over what the subject looks like. Use it for establishing shots, abstract or stylized content, and anything where no specific person or product has to appear. The AI Screenwriter toggle rewrites your rough idea in proper cinematic language, and the header links to the LTX-2 Prompting Guide.

Image to Video (I2V). You start from a still, and the model animates it while preserving its look and composition. This is the workhorse for brand and social content because the first frame is already exactly what you approved: generate the perfect image in Create, then animate it. Character, color and framing stay locked, and the prompt only describes the motion. If you want your video to look like a specific thing, start here rather than with text to video.

Video to Video (V2V). You start from an existing clip and change something about it while keeping its timing and movement. In Sogni this splits into purpose-built workflows: Motion Transfer takes the motion from a driving video and applies it to your character or image (a dance, a walk cycle, a gesture, performed by someone else); Replace Subject keeps the original scene, lighting and camera move and swaps only the subject for your character; Reference to Video blends several reference images into one video; and Sound to Video drives the visuals from an audio clip, which is what you want for music-reactive pieces.

The rest of the panel mirrors Create: a prompt box, a Speed / Quality toggle, batch size, the live Spark estimate, Voice Transfer (drop in an audio file to carry a voice into the clip), Video Size, Frames per second (24, 25, 30, 50 or 60), Duration, Steps, Guidance, Sampler, Seed and a Generate Audio switch for models that produce native sound.

15Music: the full studio

Sidebar > Music

The music studio runs on Ace-Step 1.5 XL (with Ace-Step 1.5 also selectable) and it is a proper studio rather than a single prompt box. A full track costs on the order of one Spark, which makes experimenting essentially free.

Song Style presets

Rather than guessing at genre vocabulary, open Pick a style and browse the preset library. Each preset is a genre with its production recipe spelled out, for example Afro-Cuban ("conga drums, timbales, montuno piano, brass section, son clave rhythm, tropical energy"), Afrobeats, Alternative Rock, Amapiano, Ambient Electronic, Bedroom Pop, Cloud Rap or Dance Pop. Selecting one fills the description box with those keywords, which you can then edit freely.

What you can control

The result

Press Compose and the track renders on the Supernet just like an image, with an ETA and the worker name on the tile. When it is done you get an inline player with a waveform, so you can listen immediately, scrub through the track and download it from the three-dot menu without leaving the page.

16Recent Projects

Sidebar > Recent Projects

Every generation, whether image, video or music, lands in Recent Projects, grouped by model and creation time. The info icon shows generation details and the trash icon removes a project.

17Spark, wallet and Sogni Unlimited

Sidebar > Wallet

Generations run on Spark Points, the platform's compute credits. The Wallet page shows your balance (settled, pending credits and pending debits), a SOGNI Token / Spark Points switch, an on-chain wallet on the Base and Etherlink networks for sending and receiving tokens, your full account history line by line, your creator Leaderboard ranking, and a Discord link. Buy Premium Spark tops up by card, and Premium Spark is what unlocks the Premium-badged models from step 9.

If you create a lot, Sogni Unlimited removes per-render costs entirely: unlimited fair-use generation across image, video, music and LLM, from $20 a month, with Unlimited Pro from $50 a month. Full tiers are on the pricing page.

18Rewards and Share & Earn

Sidebar > Rewards, or the "Share & Earn" button in the header

The Rewards page is free Spark. There is a Monthly Boost of 400 Spark (spend some free Spark first, then come back and claim it), a welcome reward of 100 Spark on Etherlink after you confirm your email, card top-ups through Buy Premium Spark, and the Share & Earn referral program. Friends who sign up with your link get bonus Spark, you earn bonus Spark after their first purchase, and you collect Sogni Ambassador Points on every Spark purchase made anywhere in your referral network. Your personal link and a ready-made post are both there to copy.

FAQ