Magnific (formerly Freepik), explained — and the other AI media providers we support

OpenAI, Claude, and Gemini are text AI. They write captions. They don't make images, video, or music.

For the visual side of social content, we support three providers: Magnific (formerly Freepik), fal.ai, and Google Gemini Nano Banana (with Gemini Veo for video). You can configure all three, pick a default, and the per-day Regenerate-with picker on the campaign review step lets you swap providers per day if one didn't land.

Here's the short version of each.

Magnific (formerly Freepik)

Freepik publicly rebranded to Magnific on 2026-04-28. The API, your existing key, and the underlying models all still work — the brand name changed but the product didn't. We surface the new name in the dashboard with a "(formerly Freepik)" parenthetical so existing customers don't get lost.

What it does: image, short video, and music generation under one API and one credit pool.

Models include:

Nano Banana Pro — strong for flat illustration and editorial scenes
Mystic — Magnific's house model, photographic-style content
Flux — high-quality general image generation
Seedream — clean contemporary aesthetic
Kling — short video (5-15 seconds) from a still + text prompt
Music — short instrumental clips (10-30 seconds)

When to pick it: you want one provider that does everything (image, video, music) under one billing relationship. Particularly strong for cinematic, photoreal aesthetics and for the all-in-one workflow.

Pricing: credit-based, prepaid. ~5-10 credits per image, ~30-50 per short video, ~10 per music clip. A 30-post campaign with AI images runs $5-15. Top up $10-20 when you're low.

Key signup: magnific.com/api (api.freepik.com still works through the deprecation window). Add billing, copy the key, paste into Settings → AI → Magnific key.

fal.ai

A model-routing platform. fal.ai gives you access to 1,000+ specialized models from different research labs and creators through one API and one billing relationship.

What it does: image and video generation, with a model picker.

When to pick it: you want stylized work (illustration, anime, watercolor, painterly, etc.), or you want to test multiple models before committing, or you want bleeding-edge community models that the bigger providers don't carry yet.

Pricing: per-model, generally $0.01-0.10 per image. Comparable to Magnific in aggregate for typical author volumes.

Key signup: fal.ai → dashboard → API keys. Paste into Settings → AI → fal.ai key.

Google Gemini Nano Banana (image) + Gemini Veo (video)

Google's flagship image and video models. Same Google AI Studio account that handles your Gemini text key also gives you image and video access on the same key.

What it does: image generation (Nano Banana) and short video generation (Veo).

When to pick it: you want the most generous free tier of the three providers — most authors never hit paid usage on Gemini. Also a good pick if you're already deep in Google Workspace and want one billing relationship.

Pricing: substantial monthly free quota; paid tier kicks in only at heavy usage.

Key signup: aistudio.google.com/apikey. Same key works for text + image + video. Paste into Settings → AI → Gemini key.

Picking your default

Quick decision tree:

Want one provider that does everything (image + video + music)? → Magnific.
Want to test specific stylized models? → fal.ai.
Want the cheapest on-ramp to AI media? → Gemini.
Authors who want both consistency and the option to swap? → Configure all three. Set Magnific as default for image; the per-day picker lets you swap to fal.ai or Gemini for any specific day in a campaign.

There's no penalty for keeping all three keys active. You only pay for what you actually generate.

What "imagery style" does

In Settings → AI there's an optional "imagery style" field. This is the standing instruction your image provider gets for every image it generates for you, regardless of which provider is active.

Good examples:

Flat illustration style, warm editorial palette (terracotta, cream, navy), minimal composition, no photorealism
Photographic, soft-focus morning light, shallow depth of field, autumnal palette, no text, no logos
Watercolor style, coastal New England tones, hand-drawn sketch quality, literary-cozy aesthetic

The style field tells your provider "every image I generate should look like this, regardless of the subject." This is what gives a 30-post campaign visual cohesion even when you swap providers per day.

What to review on the output

All three providers produce consistently good images. Not flawless. Things you still have to review:

Wrong era. If you asked for a contemporary scene, sometimes you'll get something subtly Victorian. Catch it, regenerate (or Regenerate-with a different provider for that one day).

Wrong mood. A horror-cover-aesthetic prompt can come back cozier than you wanted, or vice versa. Regenerate with stronger adjectives, or swap to a model that handles that aesthetic better.

Text artifacts. AI image models hallucinate text. Even when you ask for "no text," occasional fake-latin letters sneak in. The "no text, no typography, no watermarks" prompt prefix helps, but review anyway.

Character implications. If your book is diverse, make sure the generated characters reflect that. Left unspecified, AI models default to white/Western. Specify in the prompt if that matters.

Where this fits in the workflow

When you generate an AI campaign via /dashboard/create, every post that needs an image runs your default provider in the background. The prompt for each image is derived from:

The caption being generated
Your imagery style guide
Sensible defaults for the platform (vertical for TikTok/Reels, square for Instagram)

You see the generated image in the campaign review screen. If you don't like it:

Regenerate — same provider, same prompt, new attempt
Regenerate with… — pick a different provider for just this day; useful when one provider keeps producing the wrong vibe
Upload your own — drop in your own image instead

For carousels, every slide can be regenerated, swapped, or uploaded individually. Pick per-slide vs single_repeated mode in the campaign builder.

What to do next

If you're starting fresh, Gemini's free tier is the cheapest first step.
If you want all-in-one and don't mind credit-based billing, Magnific is the most popular pick.
For text AI context (separate from this), OpenAI vs Claude vs Gemini is the companion post.
For the full campaign workflow with all of this in place, 14 days of content shows text + media AI working together.