Swap the entire face in a portrait with Qwen Image 2.1.
Hair, jawline, and head shape included, while keeping the pose and background intact.
@Alibaba_Qwen
Create now ⬇️
One image in. 60 FPS anime video with native audio out.
KatoriKonoe built a single-queue workflow on Floyo: Illustrious SDXL → MiniMax H3 → RIFE interpolation → audio VAE. No post-processing, no separate audio sync.
Character motion holds. Audio matches the scene.
@MiniMax_AI
Try it now ⬇️
Run the workflow: floyo.ai/workflows/minimax-h…
KatoriKonoe's profile: floyo.ai/creators/katorikono…
AI art moved fast this week. 6 updates worth knowing:
1. Nano Banana 2.1 dropped. 14-reference fusion, text rendering that actually works, up to 4K. Google's quiet model just got loud.
2. ChatGPT is bringing ads into image generation. They show up while your image renders.
3. Tencent Hy Image 3.5 Preview. $0.024/image with ComfyUI nodes.
4. Seedream 5.0 Flash. $0.018/image, 15 seconds. A year ago this cost 10x more.
5. Kling 4.0 Flash in early access. Up to 4K, stereo audio, 10-bit HDR.
6. Vidu Q4 Preview just dropped. 15 image refs, 3 audio refs, 4K. Built for production.
Better models. Bigger possibilities. Smarter workflows.
Full info in replies. ⬇️
@Google, @OpenAI, @TencentHunyuan, @BytedanceTalk
1. Nano Banana 2.1 → ai.google.dev/gemini-api/doc…
2. ChatGPT visual ads → techcrunch.com/2026/10/05/op…
3. Tencent Hy Image 3.5 Preview → docs.comfy.org/tutorials/par…
4. Seedream 5.0 Flash → openrouter.ai/bytedance-seed…
5. Kling 4.0 Flash → klingai.com
6. Vidu Q4 Preview → unite.ai/vidu-releases-q4-pr…
Two images are all you need for a virtual try-on.
Qwen Image 2.1 swaps the outfit while keeping the person’s face, pose, and background intact.
Useful for testing different looks, product variations, or clothing concepts, with support for up to 10 reference images.
Workflow below ⬇️
@Alibaba_Qwen
Try it now: floyo.ai/workflows/qwen-imag…
Wan Animate 2 vs Wan 2.1 + SCAIL-2 vs MiniMax H3 for motion transfer.
Same source video. Same character. Completely different strengths.
Which one would you pick?
Full comparison in the comments. ⬇️
@Alibaba_Wan, @Zai_org, @MiniMax_AI
Full comparison: floyo.ai/blog/wan-animate-vs…
One workspace. Any model. LoRA training. API access. Per-team governance over every model and node.
Floyo runs open-source and closed-source models in the same browser-based workspace. Every workflow becomes an API endpoint. Every model is tagged for commercial use and privacy before anyone hits generate.
Admins control what runs where. R&D and production under different rules, same org.
Every creative challenge has a workflow.
floyo.ai
Loved seeing so many creators come together!
100+ creators joined us for a live workshop with Luka Tisler, diving into MiniMax H3.
From LoRAs and ControlNet to frame control and prompting, he covered what works, what doesn’t, and how to get the most out of H3.
Real demos, practical tips, and techniques you can put to use.
A big thank you to Luka and everyone who joined us and made this such a wonderful experience!
The recording and workflow are now live.
Link in comments ⬇️
One prompt. Three deliverables.
→ Newsletter draft with hook, takeaways, and CTA
→ Five subject lines with preview text
→ Responsive HTML email code
Built by one of our community creators on Floyo. Describe your topic, hit run, and your next send is ready.
Try it now: floyo.ai/workflows/ai-newsle…
Your AI assistant can now run Floyo workflows.
Describe what you need in Claude, ChatGPT, Cursor, or Codex. Floyo MCP finds the right workflow, runs it on Floyo GPUs, and delivers the output. You never leave the chat.
→ "Find me a product photo workflow" - searches every workflow on Floyo
→ "Run it with this image" - uploads your file, runs the pipeline
→ "Show me the result" - output lands in your team's storage
No API keys. Sign in once through your browser, done. Your team's workflows, files, and history stay private.
Learn more: docs.floyo.ai/floyo-mcp
This Thursday.
Luka Tisler walks through his full MiniMax H3 workflow live. LoRA step trade-offs, ControlNet, reference frames, and the prompting patterns that hold up in production.
45 minutes. Free. Everyone who joins gets the workflow file and the recording.
Spots are filling up. Register now: luma.com/0nzddhga
One podcast episode. Three content pieces. Under 3 minutes.
Drop in a video podcast recording. Get back:
→ An SEO blog post (800-1200 words, ready to publish)
→ A full social package (LinkedIn post, X thread, short-form captions, pull quotes)
→ A hero image built from your actual podcast frames, not stock art
No prompts to write. Upload, run, done.
Run it on Floyo: floyo.ai/workflows/turn-any-…
If you're publishing 3+ videos a week, thumbnails can become a bottleneck.
This workflow on Floyo turns it into a one-step process. Upload your video and brand assets. Describe the thumbnail in a few words. The workflow's AI prompt engine handles composition and lighting. Nano Banana Pro generates on-brand 16:9 thumbnails ready to upload.
Your logo carries through. Your brand stays consistent.
Try it now: floyo.ai/workflows/video-to-…
7 models dropped recently. Image, video, music, and editing.
GPT Image 2.5 · Gemini Omni 1.1 Flash · LTX-2.5 · MiniMax H3 Max Turbo · Luma UNI-1 · Sonilo V1.1 · FIBO Edit 1.5 Turbo
All seven run on Floyo. Workflow links in comments.
@OpenAI, @Google, @ltx_io, @MiniMax_AI, @LumaLabsAI, @Sonilo_music, @bria_ai_
GPT Image 2.5 - floyo.ai/workflows/gpt-image…
Gemini Omni 1.1 Flash - floyo.ai/workflows/gemini-om…
LTX-2.5 - floyo.ai/workflows/ltx-2-5-f…
MiniMax H3 Max Turbo - floyo.ai/workflows/minimax-h…
Luma UNI-1 - floyo.ai/workflows/luma-uni-…
Sonilo V1.1 - floyo.ai/workflows/sonilo-v1…
FIBO Edit 1.5 Turbo - Workflow updates shortly!
Explore more workflows for every model on our platform.
