Freepik on X

Most praise arrives as work rather than verdicts: posts credit the tool with skin textures and fabric weave, a ten-keyframe storyboard, a Wan 3.0 prompt with eight camera setups, a Seedream 5 Flash test they call a positive surprise, upscaled event photos, a UI mock-up that became a working dashboard, and a short animated film made in an hour through the GPT plugin. A few simply call it the best, or "addictive". Praise for the model itself is smaller: 4 people credit Magnific One, for built-in art direction, less prompting, and results they like more than the GPT image models.

The disappointment is about identity rather than output. 8 people say the rebrand lost them: they disconnected, stopped using it, say they no longer recognize it or that it "fell off". Level with them, 8 people read the pivot as the smart call, measured against Getty's bankruptcy — visionary, profitable, a company that adapted where a rival litigated. 3 people hold both at once, liking the output while noting artifacts or a price above Nano Banana 2.

Output complaints are specific: renders at the highest resolution come out full of artifacts and still spend credits while lower settings are fine, a bug users say went days unfixed; background removal that leaves a pixelated file; characters that drift between generations; the brand kit in Final overriding a chosen Draft; no portrait ratio; illustration one person calls not quite perfect.

On accounts, an account is said to have been hacked for an unauthorized purchase with no reply for days, assets were deleted from a former Freepik account, and folders synced into the wrong account. One subscriber posted repeatedly about being asked for thousands of euros instead of a normal monthly charge.

On price, people measure against named rivals: higher than Nano Banana 2, cheaper than Seedance 2.5 Pro for video yet still "breaks the bank", and paid-only. 3 people run out of credits, one "credit-broke" after tens of thousands spent on models for a game; 3 say unlimited models quietly started charging credits, and one subscriber says they are cancelling.

Each theme in this section is one person, so this is a list of what gets named, not a ranking.

On image models, one person ran the same prompt in Magnific One and Midjourney and found Midjourney kept the requested style truer while still liking the One result; another calls One their new go-to after preferring it to the GPT image models. Against Nano Banana 2, one person puts the price as "quite high".

On video, a sample is judged cheaper than Seedance 2.5 Pro but still beyond budget, and a game developer is told Astra Light is as good and much faster for the same job. Designers name Canva as the place they start, falling back to Freepik when Canva lacks what they need, while one subscriber weighing a Black Friday renewal is thinking about trying Higgsfield rather than renewing.

The steering questions are about control and cost at once: does a regenerated clip keep the character consistent with the shots around it or force a redo of the scene; does Style hold a creature's design across several shots; how many generations a consistent look took, "the part that eats the budget"; which expression took the most tries; and whether filmmakers can direct the movement itself or whether it arrives automatic.

Plan questions are about what a subscription now includes: why Seedream 5 Pro left the unlimited tier and whether it will come back, why a desktop offer is free for some and not for a Premium subscriber, and what to spend leftover credits on before they expire — one user had 71,000 with two weeks left.

Smaller asks: 3 people cannot find an announced model in the picker or their account; 2 want tutorials, especially for automating big jobs in Spaces; 1 ask what separates similarly named versions. One person asks whether Magnific is the new Freepik, and one whether it can produce several long videos a week.

The asks are small in number and each stands alone, so read them as individual requests rather than a demand.

Two are about directing video rather than extending it: one wants keyframes to shape how a performance plays, since longer scenes make timing and continuity matter more, not less; another wants a warning when default settings will spend a large share of a credit balance on a single run, after a twenty-second video at defaults took 90% of theirs.

The rest are about reach and transparency: one asks for ElevenLabs v4 to be integrated alongside the image and video models; one wants the Magnific model available inside another platform they already work in; and one, under a showcased result, asks simply to see the making of it — how the piece was made, not only the finished output.