AI Models
One page per generation model — what it's good at, what it costs you in render time, and how to route its output into a campaign.

GPT Image 2.5 — AI Image Generator and Editor
OpenAI's current image model, the one to reach for when the headline, price row, label or UI has to be readable. Generate and edit on the page, with real transparent backgrounds and edits that stay scoped to the instruction.

MiniMax H3 Max — Faster-Than-Real-Time AI Video
fal Research’s post-trained MiniMax H3 model generates a five-second 768P video with native audio in about three seconds. Built for high-volume iteration, stronger prompt adherence, and interactive creative workflows.

Pika AI Video Generator
Free Pika AI video generation for ad creative with Soku AI — turn a prompt or product photo into short-form social video with Pikaffects and Pikaframes, then scale the winners into ad variants and test them across Meta, Google, and TikTok.


MiniMax H3 (Hailuo 3.0) — AI Video Editing
MiniMax H3 ranks #1 on Artificial Analysis' Video Editing leaderboard — feed it a finished clip and relight it, swap the background, add VFX, or re-voice the dialogue while the rest of the shot holds. Real before/after pairs, 30+ copy-ready prompts, and the editing prompt formula.

Seedream AI Image Generator
Free Seedream AI image generation for ad creative with Soku AI — layout-planned images up to 2K, native text rendering in 14 languages, region-precise editing, then deploy straight to Meta, Google, and TikTok campaigns.

Seedance 2.5 by ByteDance — Native 30s AI Video
ByteDance's latest AI video model — a native 30-second single shot with up to 50 multimodal references, 4K output, element-level editing, and native audio. Generate it in Soku AI and turn it into ad creatives.

Seedance 2.0 Mini by ByteDance — Fast, Low-Cost AI Video
ByteDance's distilled Seedance 2.0 video model — ~2× faster than Fast and ~half the cost of standard 2.0, with the same multimodal inputs and native audio at up to 720p. Generate it in Soku AI and turn it into ad creatives.

Seedance 2.1 by ByteDance — AI Video Generator
ByteDance's latest multimodal AI video model — ~20% sharper than 2.0 with smoother motion, stronger consistency, native audio, and up to 2K, 15-second clips. Try it free in Soku AI and turn it into ad creatives.

Luma AI Video Generator
Luma AI (Dream Machine) is Luma Labs' AI video model — Ray3 brings reasoning, native 1080p, and studio-grade HDR. Pair it with Soku AI to turn cinematic clips into ad creative and test them across Meta, Google, and TikTok, free to start.

Imagen AI Image Generator
Imagen is Google's high-fidelity text-to-image model (Imagen 4, Ultra, and Fast). Pair it with Soku AI to turn output into on-brand ad variants and deploy across Meta, Google, and TikTok — free to start.

Ideogram AI Image Generator
Ideogram AI is the image generator known for industry-leading in-image text rendering — legible headlines, logos, and typography. Pair it with Soku AI to turn output into on-brand ad variants and deploy across Meta, Google, and TikTok.

Reve AI Image Generator
Generate ad-ready images for free with Soku AI. Reve AI (Reve Image 1.0) raised the bar on prompt adherence and in-image text — Soku turns AI image generation into a performance creative engine across Meta, Google, and TikTok.