Your First Hour with RandomSeed

In one hour you can generate your first image, animate it to video, build a moodboard, and run a preset chain. This guide walks through each step and — more importantly — helps you pick the right model for each task so you are not guessing from a list of 35+ options.

Orient Yourself

Open the studio and you will see three things: the spatial canvas in the center, the prompt bar at the bottom, and the mode switcher. The six generation modes are Image, Enhance, Video, Voice, Music, and 3D. Each mode exposes a different set of AI models tuned for that media type.

In the top corner you will find your credit balance and settings. Every generation costs credits — the price depends on the model and is always visible before you run anything. New accounts start with 100 free credits, which is enough to try several models across different modes.

The canvas is infinite and persistent. Everything you generate appears as a node that you can drag, group, and connect. Think of it as a whiteboard that remembers everything between sessions.

Your First Image

Switch to Image mode, pick a model from the dropdown, type a prompt, and hit generate. Your image appears on the canvas in a few seconds. The hard part is not the process — it is choosing from 11 image models. Here is how to think about it.

Which image model should I pick?

If you are unsure, start with FLUX Dev at 6 credits. It is fast enough to iterate on and good enough for most projects. Move up to FLUX Pro 1.1 when you need something client-ready. Use HiDream Fast when you want to test five prompt variations without burning through credits.

Two models have specific strengths worth noting. Recraft V3 is the best choice for design assets — logos, icons, illustrations, anything that needs a clean graphic look. Ideogram v3 is the only model that reliably renders readable text inside an image, which matters for posters, social cards, and branded graphics.

Enhance and Derive

Click any image on the canvas to open the inspector panel. From here you can branch your work in several directions without leaving the workspace.

What you can do from the inspector

• Regenerate — Run the same prompt again with a different seed for variations
• Regenerate with a different model — Keep the prompt, swap the model to compare outputs side by side
• Branch — Use the image as a starting point for image-to-image generation with a new prompt
• Enhance — Apply enhancement tools like background removal, upscaling, or face restoration
• Animate — Send the image directly to a video model

Enhancement tools and when to use them

A common workflow: generate with FLUX Dev, check the result, run Face Restore if needed (1 credit), then Creative Upscale for the final version (12 credits). Total cost: 19 credits for a production-ready image.

Animate to Video

From the inspector, click Animate to send your image to a video model. Image-to-video is where credits scale up — but so does the output quality. Here is how to pick the right video model.

Image-to-video model guide

Seedance 1.5 Pro and Sora 2 Pro are the newest additions to the video lineup. Try them if you want to compare against the established models above.

The practical approach: preview your animation with LTX Video at 10 credits. If the motion direction looks right, re-run with Kling 1.6 Standard (55 credits) or Veo 3 (120 credits) for production quality. This saves you from spending 120 credits on a motion concept that needs a different prompt.

Build a Moodboard

A moodboard anchors your generations to a consistent aesthetic. Without one, each generation starts from zero and outputs can drift in style, color palette, and mood between sessions.

How to set one up

Open the moodboard library and create a new board. Upload 3 to 10 reference images — photographs, illustrations, screenshots, anything that captures the visual direction you are aiming for. These do not need to be your own work. They are references, not style transfer sources.

The Chrome extension

The best references come from your natural browsing. The RandomSeed Chrome extension lets you right-click any image on the web and save it directly to a moodboard — no downloading, no re-uploading. You can also crop-save a region of any page to grab just a color, texture, or composition detail. Everything auto-tags on save.

Saves stay in local storage until you create an account, then sync automatically. Your moodboard grows while you browse, without extra effort.

The six descriptors

RandomSeed analyzes your references across six dimensions: color palette, lighting mood, composition style, texture quality, subject treatment, and overall atmosphere. Together these form a style brief — a compact summary of your aesthetic preferences that you can apply to any generation.

Using your taste profile

Once your board has a synthesized brief, you can pull prompt fragments from it directly into the prompt bar. This means you are not rewriting style instructions every time — the moodboard supplies them. Over time, as you generate and give feedback, the brief refines itself to better match what you actually want.

Moodboards are especially valuable for brand work, content series, or any project where visual consistency matters across multiple sessions and prompts.

Try a Preset Chain

Preset chains are pre-built pipelines that run multiple models in sequence with one click. Instead of manually generating an image, then removing the background, then upscaling, then animating — a preset does all four steps automatically, passing each output to the next model in the chain.

The six presets

• Product Shoot — Upload a product image, remove the background, upscale to 4x, animate to video. Starts from an image upload.
• Animate Shot — Write a prompt, generate an image with FLUX Pro 1.1, animate to video with Kling 1.6. The simplest text-to-video pipeline.
• Launch Pack — Prompt to image (FLUX Dev), remove background, upscale, animate. A complete asset pipeline for product launches and marketing.
• Podcast Kit — Generate cover art with FLUX Pro 1.1 and a voiceover with Kokoro from the same prompt. Two assets, one click.
• Product to 3D — Upload a product image, remove the background, generate a 3D model with Trellis. Useful for e-commerce and product visualization.
• Music Video Kit — Prompt to image (FLUX Pro 1.1), animate to video (Kling 1.6), generate a soundtrack (CassetteAI). Three media types from a single prompt.

Presets are a good way to see how pipelines work before building your own custom chains. Pick the one closest to your use case and run it — you can always swap individual models in the chain afterward.

Voice, Music, and 3D

The remaining three modes round out the studio. Each has a smaller model set, which makes choosing easier.

Voice

Music

3D

The Model Cheat Sheet

Bookmark this table. When you are staring at a model dropdown and unsure what to pick, match your task to the left column.

For the full list of every model, endpoint, and credit cost, see the model guide.

What to Do Next

You have now used most of what RandomSeed offers. From here, the workflows that matter most depend on what you are building:

• For brand consistency — invest time in your moodboard. Upload more references, generate a few batches, give feedback to refine the brief.
• For production pipelines — start building custom chains beyond the presets. Swap models, add enhancement steps, tune the flow for your specific output needs.
• For exploration — try the same prompt across multiple image models to develop an intuition for each model's strengths. The canvas makes side-by-side comparison easy.

Open the studio and start building.