You won’t believe some of the real AI requests clients are bringing to me! 😲 From automating bookings to AI‑generated content and intelligent analytics — these examples show what’s possible today.
📖 Dive in here: https://medium.com/@addisonolivia721/what-clients-are-asking-me-to-build-with-ai-031678d0a810
#BuildWithAI #AIProjects #AIClientRequests #AIAutomation #AIChatbots #PredictiveAI #MachineLearning #AIForBusiness #AIContentGeneration #AIWorkflow #IntelligentSystems #AIForHealthcare #NoCodeAI #AIForMarketing #AIForEcommerce #AITrends2026 #AIInsights #AIbots

The images were created with Stable Diffusion 3.5, using the text encoders:
- clip_g
- clip_l
- t5xxl_fp8_e4m3fn

SD 3.5 is particularly strong in concept art, abstract ideas, and complex scenes, as it handles nested prompts very well. The multiple text encoders are needed to accurately capture different aspects of the description: clip_g and clip_l for general visual semantics and style, and t5xxl_fp8_e4m3fn for complex, nested language structures.

Compared to SDXL, SD 3.5 is more experimental and geared toward creative, symbolic compositions, while SDXL is more stable with photorealism and figures.

Included: a screenshot of my ComfyUI workflow, which was used to implement the prompts in SD 3.5.

#AI #StableDiffusion #ComfyUI #GenerativeArt #ConceptArt #DigitalArt #AIArt #CreativeAI #TextToImage #PromptEngineering #OpenSource #Fediverse #ProceduralArt #SurrealArt #SciFiArt #FantasyArt #VisualAI #AIWorkflow

Small making-of of my “True Beauty Is So Painful” piece (listening to “True Beauty Is So Painful” by Oomph! in the background), because “AI art = just pressing a button” is still a thing.

Here I’m showing briefly (15 MB max file upload) my SDXL workflow in ComfyUI, from node structure to model choice to parameters.

LoRAs in this setup are only linked to the positive prompt, because I wanted to fine-tune their weights there specifically, without affecting the negative prompt.

During rendering, I ran in parallel:
- GPU load with radeontop, you can clearly see how on RDNA2 everything (matrix multiplications, convs, etc.) runs over the shaders
- Temperatures & power states briefly shown with corectrl

Peak at 187 W, hotspot briefly at 97 °C
RDNA2 doing RDNA2 things…

Video workflow:
- recorded with OBS
- edited in Kdenlive
- transcoded with VAAPI (H.264)

No cloud, just decisions, iteration and real hardware.
Everything runs on Linux + ComfyUI (FOSS), so anyone can set this up.
No GPU? No problem, you can also run it using PyTorch’s CPU backend, just much slower.

#AIArt #ComfyUI #SDXL #stablediffusion #LoRA #FOSS #Linux #AMD #RDNA2 #GPUComputing #OpenSource #AIWorkflow #OBS #Kdenlive #VAAPI #DigitalArt #MakingOf #AIProcess #NoCloud

Các nhà phát triển AI tại 2026 sẽ dùng "n8n + Llama3.1 + Mistral/Deepgram + Composio" để xử lý song song thay thế chuỗi AI tuần tự. Ưu điểm: Tốc độ 6x, chính xác 94%, chi phí chỉ 1.15 USD/tháng. Ví dụ: Phân tích hợp đồng pháp lý, xử lý giọng nói. Thách thức: Xung đột dữ liệu, mất cân bằng tải. Câu hỏi: Bạn chọn chiến lược "hàng đợi ưu tiên" hay "luân phiên" cho phân bổ tải swarm AI?
#ParallelSwarm #AIOrchestrator #n8nAI #2026Tech #SwarmIntelligence #AIWorkflow #ScaleAI #CongNgheAI #PhanTichDuL

YES SUCCEEDED!!!

Just rendered an image at 944×1152 (slightly above 1024×1024) using Flux1-Schnell-FP8 on my 6700 XT, and it works! (Image 1 is the Real-ESRGAN 2× upscaled version)

Workflow 1: Sampling (Image 2)

Prompt executed → UNet generates the latent

Step 1 (model load + latent generation) took 419 seconds

Output: Latent tensor saved to disk

Workflow 2 : VAE Decode (Image 3)

Latent loaded → VAE decodes the image

Duration: 7.5 seconds

Advantage: UNet doesn’t need to stay in VRAM → VRAM freed, even on 12 GB GPUs

The problem with the stock LoadLatent Node

Dropdown only shows files if they were produced / annotated by a previous SaveLatent Node

Node is designed to pass latents inside a graph, not load arbitrary files from disk

Purpose: prevents accidentally loading wrong files

Workaround (Image 4)

Edited /ComfyUI/nodes.py, class LoadLatent

Hardcoded latent path → Node now loads directly from disk

Result: Workflow 2 runs instantly, UNet can be unloaded

Timing

Step 1 (model load + latent generation): 419 s

Step 2 (VAE decode): 7.5 s

Result: High-res images on a 12 GB RDNA2 GPU are now possible on Flux1-Schnell-FP8 without ComfyUI crashing! (Image 5 is the original output)

This might actually become my new Flux workflow: render quick 512×512 previews first (which works perfectly on RDNA2 GPUs), sort out the good ones, extract the seed from the PNG metadata, and then re-render only the selected images with the same seed using the split workflow at higher resolutions. This way, high-resolution Flux1-Schnell-FP8 renders become possible on 12 GB RDNA2 GPUs D:

Question at the end: Has anyone ever done this before? Because I have no clue xD

#ComfyUI #flux #Flux1SchnellFP8 #FP8 #AMD #RDNA2 #VAE #AIArt #Pixelfed #HighResolution #GPUOptimization #LatentWorkflow #AIWorkflow #AIHacks #RealESRGAN #Upscale #AIExperiment #CreativeAI #DigitalArt #AICommunity #python #linux #opensource #foss