Model Catalog
30+ models — including our own
EGAKU AI hosts 9 frontier models (Sora 2, Veo 3.1, Kling 3, Nano Banana 2, GPT Image 2, Flux Pro, and more) — plus 8 custom-trained models we're building in-house. Own weights = own aesthetic = competitors can't clone it.
The moat
EGAKU Original
Our own base models and LoRAs, trained from scratch by us. This is what nobody else has.
EGAKU Original Base Models
EGAKU α-real 2 (Z-Image)
Own — betaEGAKU · 3 cr · basic+ plan
自前ホストの写実ベース。人物・肌・日本語テキストに強い。無検閲(NSFW+SFW)・商用クリーン(Apache)。外部APIに繋がず EGAKU 自社インフラで生成。
Try it →EGAKU α-real 1 (Chroma)
Own — betaEGAKU · 3 cr · basic+ plan
自前ホストの写実ベース。背景の緻密さ・作風の幅・全身に強い。無検閲(NSFW+SFW)・商用クリーン(Apache)。外部APIに繋がず EGAKU 自社インフラで生成。
Try it →EGAKU α-anime
Own — coming soonEGAKU · TBD cr · basic+ plan
Anime & illustration SDXL base merged from Illustrious / NoobAI / Pony + our own aesthetic. Japanese anime → American comics range in one base.
ETA: In training (Q3 2026)
EGAKU β-video
Own — coming soonEGAKU · TBD cr · pro+ plan
Wan 2.6 / HunyuanVideo base fine-tuned for our aesthetic. Fills the domain gap where Seedance 2 Pro leaves off (Japanese UGC, adult, horror, gore).
ETA: In training (Q4 2026)
EGAKU Original LoRAs
Japanese Realism (LoRA)
Own — coming soonEGAKU · TBD cr · basic+ plan
Japanese-face realism LoRA on EGAKU α-real. Fixes the 'generic Asian face' problem in stock SDXL.
ETA: Q3 2026
Shinkai-style Landscape (LoRA)
Own — coming soonEGAKU · TBD cr · basic+ plan
Anime landscape LoRA on α-anime — Makoto Shinkai / new-wave anime background aesthetic.
ETA: Q3 2026
Western Comics (LoRA)
Own — coming soonEGAKU · TBD cr · basic+ plan
American comics / Marvel-style LoRA on α-anime — brings Western illustration range into the same base.
ETA: Q3 2026
Splatter / Body Horror (LoRA)
Own — coming soonEGAKU · TBD cr · pro+ plan
Practical-effects horror LoRA — Cronenberg body horror, Sam Raimi splatter, Ari Aster dread. Adult-gated.
ETA: Q4 2026
LoRAs roll out as they finish training. Roughly one per week during Q3 2026 ramp.
Frontier Image Models
Nano Banana 2 (Google)
Frontierfal.ai · 8 cr · lite+ plan
Top-tier prompt adherence + kanji text rendering. Google DeepMind.
Try it →GPT Image 2 (OpenAI)
Frontierfal.ai · 8 cr · lite+ plan
OpenAI's latest — strong all-round quality + CJK text.
Try it →Grok Imagine (xAI)
Frontierfal.ai · 8 cr · lite+ plan
xAI's image model — permissive, photorealistic bias.
Try it →Ideogram v3
Frontierfal.ai · 6 cr · lite+ plan
Best text rendering, poster / signage design specialty.
Try it →Frontier Video Models
Sora 2 (OpenAI)
Videofal.ai · premium cr · pro+ plan
OpenAI Sora 2 — longest coherent video, best physics.
Try it →Veo 3.1 (Google)
Videofal.ai · premium cr · pro+ plan
Google Veo 3.1 — video + native audio in one generation.
Try it →Kling 3.0
Videofal.ai · premium cr · pro+ plan
Kuaishou's Kling 3.0 — top character consistency.
Try it →Seedance 2 Pro (ByteDance)
Videofal.ai · premium cr · pro+ plan
1080p image-to-video, strong physics. Used in Try On UGC flow.
Try it →Open Video Models
Wan 2.6 I2V
Videofal.ai (self-host planned) · 20 cr · lite+ plan
Alibaba Wan 2.6 image-to-video. Powers Try On. Self-hosting on RunPod planned once α-real ships.
Try it →Wan 2.1 T2V 14B
Videofal.ai · 18 cr · lite+ plan
Alibaba's prior generation, still strong text-to-video baseline.
Try it →HunyuanVideo I2V 13B
Videofal.ai (self-host planned) · 22 cr · lite+ plan
Tencent HunyuanVideo — open weights, strong physics, NSFW-friendly. Alt aesthetic to Wan.
Try it →LTX Video 2.3
Videofal.ai · 5 cr
Fast, low-VRAM open video. Iteration-friendly. Free tier accessible.
Try it →CogVideoX 1.5 5B
Videofal.ai · 10 cr
SFW anime + light illustration. Permissive commercial license.
Try it →SkyReels V2 (anime)
Videofal.ai · 12 cr · lite+ plan
Anime video specialist. Studio-anime aesthetic out of the box.
Try it →Photorealistic Image Models
Anime & Illustration Models
Adult (18+) Models
Try any of the models above with 5 free generations — no signup required.
Start Free →