Hey Dreamers 🌙 The English side of PixAI. World’s No.1 Anime & Character Generation AI Announcements come from @pixai_official

Joined July 2026
[PixAI × zeta Collaboration] PixAI × zeta Original Character Creation Contest 2026 Now Live! Your OC deserves more than a single picture. Draw them on PixAI, then give them a personality, a story and a voice on zeta. Event period: Oct 23 – Nov 4 (KST) 3 tracks · 76 winners 180,000,000 PixAI Credits + 359,000 zeta Pieces Share the event page and get 5,000 Credits to start creating! Join Now: pixai.art/event/zeta
10
2
62
450,299
We encountered an issue during server maintenance, and the website is currently inaccessible. Our team is working hard to fix it as quickly as possible. We apologize for the inconvenience and appreciate your patience.
5
16
551
We're currently performing server maintenance, and some users may experience login issues. Please log in again to continue as normal. We apologize for any inconvenience and thank you for your patience.
1
20
700
PixAI × zeta Share Your Favorite OCs Find a character card you love in the zeta collab area and tell everyone why it's worth a chat. Total Prize Pool: 36,000,000 PixAI Credits + 84,000 zeta Pieces for 30 Winners. Event Period: Oct 26 – Nov 4 KST How to enter: 1. Pick a character from zeta's collab area. 2. Take a screenshot of its plot details page (featuring PixAI cover art). 3. Post it on X with #PixAIxzeta2026 and the zeta plot link. Optional: Tell us why you love the character, share your favorite chat moments, or show off different looks of your own OC! Judging Criteria: zeta judges each post on how well it shows the card's charm, not on likes or views. Prizes: 1st Prize (3 winners) : 6,000,000 PixAI Credits + 10,000 zeta Pieces per winner, plus a feature on zeta's event results page. 2nd Prize (9 winners): 1,000,000 PixAI Credits + 3,000 zeta Pieces per winner. 3rd Prize (18 winners): 500,000 PixAI Credits + 1,500 zeta Pieces per winner.
4
10
510
The whole Tsubaki.3 pipeline, in one picture. Shared text-to-image pretraining and mid-training come first. Then the path forks. ・Text-to-image track: post-training → distillation → post-distillation RL ・In-context track: continued pretraining on image-conditioned data, then its own distillation The full model is Tsubaki.3. The distilled one is Flash. Thank you for reading along for 28 days. The full report, Tagger 1.0, and open roles are all here: pixai.art/tsubaki-3/report
【What Can #PixAITsubakiV3 Actually Do?】 Explore the capabilities of Tsubaki.3 through PixAI's Tech Report! While general-purpose illustration generation has matured, generating anime art still comes with significant challenges like diverse styles, multiple characters, complex poses, and artistic camera angles. Tsubaki.3 was built to address these challenges as a multimodal foundation model, supporting both text-to-image and in-context generation, and designed specifically for the anime creator community.
1
9
512
PixAI Music Jam — the Rekka Katakiri episode is now up as a standalone video. youtu.be/8xbzgK4Xybo "Dawn Gala" was written for New Year. And yet, "Bright and full of energy? That's summer, surely." Her colours: "energetic red," "sky blue," and "auspicious gold." #PixAIMusicJam #PixAI4th
5
407
What Is Tsubaki.3 Flash? Tsubaki.3 Flash is the distilled version of Tsubaki.3. Distillation trains a student model to generate with far fewer network evaluations than its multi-step teacher. The relevant background here is DMD and DMDR. Distillation tends to lose fine texture and small detail. So the Flash student receives one more stage: reward-guided alignment driven by our in-house aesthetic scorer. Full model and Flash are separate offerings. The text-to-image evaluation in the report covers the full model.
【What Can #PixAITsubakiV3 Actually Do?】 Explore the capabilities of Tsubaki.3 through PixAI's Tech Report! While general-purpose illustration generation has matured, generating anime art still comes with significant challenges like diverse styles, multiple characters, complex poses, and artistic camera angles. Tsubaki.3 was built to address these challenges as a multimodal foundation model, supporting both text-to-image and in-context generation, and designed specifically for the anime creator community.
1
1
11
670
Notice: PixAI iOS App Temporarily Unavailable on the App Store Dear PixAI users, The PixAI iOS app is currently unavailable for search and download on the App Store. We are actively working with Apple to restore it as soon as possible. We sincerely apologize for the inconvenience. Please note: If you already have the app installed, you can continue to use it and sign in as usual (see below for Sign in with Apple). Generation features and your account data are not affected. “Sign in with Apple” is temporarily unavailable. All other sign-in methods work normally. If Sign in with Apple is the only sign-in method linked to your account, please contact us through in-app support and we will help you sign in. In-app purchases (membership subscriptions, credits, etc.) are temporarily unavailable, and existing memberships cannot be renewed in the app for now. Until the app is back on the App Store, please do not delete PixAI from your device. Once deleted, it cannot be downloaded again. We also recommend turning off “Offload Unused Apps” in Settings → App Store so that iOS does not remove PixAI automatically. We will let you know as soon as the app is available again. If you have any questions, please contact us through in-app support. Thank you for your understanding and support! The PixAI Team
5
912
Tagger 1.0 vs v0.9: +10.4 Points on Characters We compared Tagger 1.0 with our previous release, pixai-tagger v0.9, under one controlled protocol and a shared vocabulary of 8,407 General tags and 2,099 Character tags. ・General micro-F1: 0.598 → 0.666 (+6.8 pts) ・Character micro-F1: 0.820 → 0.924 (+10.4 pts) Macro-F1 and mAP improve in both categories. On the 80,531-image held-out split, Character macro-F1 is 0.875 and Copyright macro-F1 is 0.778. General and Meta tags remain weaker on rare tags. Treat them as a strong first pass before human review.
【What Can #PixAITsubakiV3 Actually Do?】 Explore the capabilities of Tsubaki.3 through PixAI's Tech Report! While general-purpose illustration generation has matured, generating anime art still comes with significant challenges like diverse styles, multiple characters, complex poses, and artistic camera angles. Tsubaki.3 was built to address these challenges as a multimodal foundation model, supporting both text-to-image and in-context generation, and designed specifically for the anime creator community.
1
5
452
They put my face on a cake and then ate it. I have questions 🥲
もち、今日のお仕事は、PixAIの4歳のお誕生日を全力でお祝いすること!! 普段はお仕事に励むメンバーたちも、この日はすっかりパーティーモード! もち的には、こういうお仕事なら毎日でも大歓迎です(?)🍰 せっかくなので、当日の様子を皆さんにもおすそ分け📸(なにやら不思議な飲み物で乾杯している人もいたような…… これからも皆さんと一緒に、もっともっと楽しいことができますように!💜 #PixAI #PixAI4周年
2
6
429
PixAI Music Jam — Yui Sakakibara episode is now up as a standalone video. youtu.be/UVWow3qs-mk "The lyrics arrived written entirely in hiragana. I thought it was a bug." Somewhere in the lyric was the word “nasu” — eggplant. It really does become a picture. The only episode with two PixAI creators, and two completely different images. #PixAIMusicJam #PixAI4th
5
7
494
Invite rewards are now 50,000 credits! 🎉 Thanks for being part of PixAI! As a thank-you, invite reward just goes up from 20,000 to 50,000 credits, and both you and your friend get the reward. How it works 1. Copy your referral link and send it to a friend. 2. Your friend signs up through the link, generates at least once, and logs in on 2 different days. 3. You both get 50,000 credits within 24 hours. If you've been enjoying PixAI, share it with your friends and create together with them! Invite a friend now:pixai.art/user-referral
3
1
18
1,285
One painting, three new looks. Prompts in the thread How to use 👇 1. Use your artwork from PixAI or upload your artwork as a reference image 2. paste one of the prompts below Works best with Tsubaki.3
Captain mio reporting for duty.
1
7
24
763
halftone print Convert Image1 into a monochrome halftone print. Render every tone as a regular grid of round black dots on white paper, with large dense dots in the shadows and tiny sparse dots in the highlights, evenly covering the whole image including the background. Keep everything else in Image1 exactly as it is: the character's face, hair, outfit, pose and framing. Newspaper-style halftone illustration, pure black ink on white paper, a fine 45-degree dot screen. Same framing and eye-level angle as Image1. Even flat studio lighting.
1
173
manga illustration Redraw Image1 as a professional black-and-white manga illustration. Trace the character with crisp, confident ink lines of varied weight, fill the darkest areas of the hair and clothing with solid black, and place fine screentone only in the shadow areas. Keep everything else in Image1 exactly as it is: the character's face, expression, hair style, clothing and pose. The background is clean untouched white paper. Delicate clean line art, high contrast, Japanese manga screentone style. Try it here →pixai.art
1
129