Mediafy Studio Tutorial: Your First Image in 5 Minutes

You type a cool logo for my coffee shop into an AI image tool, hit go, and get a muddy brown swirl with garbled letters. You try again. Same thing, slightly different swirl. The problem isn't the tool. A six-word prompt leaves the model to guess the style, the subject, the lighting and the text, and it guesses badly.

This Mediafy Studio tutorial walks you through a first image in about five minutes, and shows how to avoid that mistake from the start.

Before you start: what you need

• A Google account. Mediafy Studio signs you in with Google, at /signin.
• Some credits. Mediafy runs on credits, with no free credits; a monthly plan is optional. One credit is CA$0.10. Packs are 50 credits for CA$5, 110 for CA$10, or 300 for CA$25. You pay by card through Square.

A small pack is plenty for this walkthrough. A quick draft can cost as little as 2 credits.

Step 1: Open the Studio and choose Image

In the Studio you pick the type of output first: Image, Video or Audio. Choose Image.

Step 2: Pick a model

Next you choose a model. The lineup changes over time, but image models such as Nano Banana 2, Seedream and GPT Image are offered.

Each model has its own credit cost, and that cost shows on the Generate button before you run anything. So you're never surprised. For example, a Nano Banana image is 3 credits.

A sensible way to choose for a first run:

1. Use a cheap, fast model to test your idea.
2. Once the composition is right, rerun the same prompt on a stronger model.
3. If your image needs readable text, such as a sign or a logo, check the result rather than assuming it is right. Ideogram is generally known for handling lettering, but it is not available in Mediafy Studio.

Drafting cheap first is a habit worth keeping. It's the easiest way to stretch a CA$5 pack.

Step 3: Write a prompt that gives the model something to work with

A good prompt covers five things. You don't need fancy words, just specifics.

• Subject: what is in the picture
• Setting: where it is
• Style: photo, illustration, 3D render, flat vector
• Light and mood: soft morning light, moody neon, overcast
• Framing: close-up, wide shot, overhead

Compare these.

Weak: coffee shop logo

Better: Flat vector logo for a small neighbourhood coffee shop, a steaming cup inside a circle, warm brown and cream palette, clean lines, white background

And for a photo:

Overhead photo of a flat white with latte art on a wooden table, soft window light from the left, shallow depth of field, a croissant blurred in the background

Notice what the better prompts do: they name the medium, the colours and the background. They also say what to leave plain. The model no longer has to invent those choices.

Use Enhance when you're stuck

If you only have a short idea, write it, then press Enhance. It rewrites a short prompt into a fuller one. Read what comes back before you run it. Delete anything that doesn't match your picture, and add what's missing, like a colour or a camera angle.

Stuck for ideas? Use Discover or Chat

• Discover has ready-made example prompts with a Try this button. It's a fast way to see what a good prompt looks like and then edit it.
• Mediafy Chat is an assistant for brainstorming ideas and writing prompts. Describe your project and ask for three prompt options.

Step 4: Check the cost and generate

Look at the Generate button. It shows the credit cost for the model you chose. If you're happy, click it.

If a generation fails, the credits are refunded automatically, so a failed run won't cost you.

Step 5: Read the result and iterate

Don't judge the first image as a pass or fail. Look at what is wrong and change one thing.

• Wrong style? Name the medium more clearly: watercolour illustration instead of artistic.
• Cluttered? Add simple background or plenty of empty space.
• Wrong colours? Name them. Teal and orange beats colourful.
• Garbled text? Put the exact words in quotes, keep them short, and try a text-friendly model.
• Odd hands or faces? Change the framing, or try another model. Different models fail in different places.

Change one variable per attempt. If you rewrite the whole prompt each time, you can't tell what helped.

Common first-timer mistakes

1. Prompts that are too short. Three words give the model too many choices.
2. Prompts that are too long and contradictory. Minimalist, ornate, photorealistic cartoon fights itself. Pick one direction.
3. Starting on the most expensive model. Draft cheap, finish better.
4. Changing everything at once. Edit one thing so you learn what each change does.
5. Not saving what works. Good prompts are worth keeping.

Keep your work organised

Every result lands in Assets, so you can find it again. When a prompt works, save it in My Prompts and reuse it as a template. Swap the subject, keep the style lines, and your images start to look like a set.

What to try next

Once you're comfortable with a single image, there's plenty more to explore:

• Seeds are saved sets of reference images that help keep a consistent look. They're not custom model training, just references you reuse.
• Characters let you design a face once and reuse it. Type @ in the Studio to bring in a saved character or Seed.
• Flows is a node canvas that chains steps, such as image, then video, then voice-over. Magic can build a flow from a plain description.
• UGC Factory takes a character and a product, writes a short script, makes a selfie-style image, lip-syncs it into a talking video and adds subtitles.

If you want a structured path through prompting and AI content workflows, the AI Creator Lab course is worth a look.

Your five-minute checklist

1. Sign in with Google and add a small credit pack.
2. Choose Image, then a cheap model for drafting.
3. Write a prompt with subject, setting, style, light and framing.
4. Use Enhance if it's short, and edit the result.
5. Check the credit cost on Generate, then run it.
6. Change one thing at a time until it's right.
7. Save the winning prompt.

That's the whole loop. Everything else in the studio builds on it. Ready to make your first image? Sign in at /signin.