Custom AI Digital Designer Development with ComfyUI & LoRA

A designer manually processes 5–20 images per day. If you need a thousand banners, that's weeks. Moreover, manual work is limited by the human factor: fatigue, errors, brand guideline inconsistencies. An AI designer driven by ComfyUI yields 50–500 variants per day. But without proper tuning, the model generates garbage: artifacts, wrong colors, distorted logos. We configure a pipeline that reliably reproduces the brand identity, including MLOps – latency and GPU usage monitoring, auto-scaling. A neural network-based digital designer is not just a generator; it's a full-fledged AI agent that can work in tandem with your CMS. Such a neural network for design takes over routine visual generation, freeing up the creative team for strategic tasks.

How does an AI designer boost productivity?

The difference between manual labor and AI agents is tens of times. An AI designer works 10–25 times faster than a human while maintaining a consistent style. It handles the routine: retargeting banners, blog covers, social media posts. Humans control quality and creativity – AI generates variants in seconds.

Parallel generation on GPU allows running multiple workflows simultaneously. For e-commerce, that's catalogs with thousands of items. For SMM, it's daily posts without overloading the team.

How to ensure a consistent brand style?

The main problem with AI generation is style drift. The solution is LoRA fine-tuning on branded images. Step-by-step process:

1. Collect 20–50 of your reference works.
2. Train the model in 30–60 minutes on GPU (500–2000 steps).
3. Integrate LoRA weights into the ComfyUI pipeline.
4. Add post-processing: logo overlay, watermarks, color correction.

Result: generation exactly matching guidelines – colors, fonts, composition.

We additionally overlay logos and watermarks via post-processing. Each banner is checked for palette compliance and artifacts.

What's included in AI designer development?

Final metrics: generation speed 50–500 images per day, 5–10x reduction in cost-per-image, 95%+ style consistency.

Typical mistakes when implementing an AI designer

• Insufficient references: LoRA requires 20–50 high-quality samples. Fewer – the style doesn't stick.
• Ignoring latency: batch processing without optimization leads to p99 latency > 10 seconds.
• No monitoring: without CLIP-score and SSIM metrics, you won't see quality drift.
• Poor prompt engineering: forgetting negative prompts for artifacts.

Why choose us?

We are a team of AI/ML engineers with 5+ years of experience in Computer Vision and NLP. We have 50+ projects in design automation for e-commerce, media, and brands. We don't just deploy a model – we integrate a pipeline into your stack: AWS, GCP, or on-premise servers.

We guarantee results: if the model outputs defects, we refine it free of charge. Transparent quality reporting: PSNR, SSIM, CLIP-score – we show numbers, not just images.

How to start?

Send us a link to your reference visuals – we'll prepare a demo generation of 10 variants in 1 day. Get a specialist consultation on your project. Order AI designer development today – first month of support free of charge. Contact us for a demo.

Stack of tools

For mass generation, we apply MLOps practices: log CLIP-score, PSNR, SSIM metrics; automatically restart workflows when quality drops. This ensures visual stability at volumes from 1000 images per day. Neural network visual automation reduces content delivery time by 10–25 times.

This article is based on open technologies: Stable Diffusion and ComfyUI.

Automation via ComfyUI API

import httpx
import json
import uuid

class AIDesignerAgent:
    def __init__(self, comfyui_url: str = "http://localhost:8188"):
        self.base_url = comfyui_url

    async def generate_banner(
        self,
        prompt: str,
        brand_colors: list[str],
        width: int = 1200,
        height: int = 628,  # OG-image size
        style: str = "modern corporate"
    ) -> bytes:
        workflow = self.build_sdxl_workflow(
            prompt=f"{prompt}, {style}, brand colors {' '.join(brand_colors)}, professional design",
            negative_prompt="low quality, blurry, text errors, watermark",
            width=width,
            height=height
        )

        client_id = str(uuid.uuid4())
        async with httpx.AsyncClient() as client:
            response = await client.post(
                f"{self.base_url}/prompt",
                json={"prompt": workflow, "client_id": client_id}
            )
            prompt_id = response.json()["prompt_id"]
            return await self.wait_for_result(client_id, prompt_id)

    def build_sdxl_workflow(self, prompt: str, negative_prompt: str, width: int, height: int) -> dict:
        return {
            "4": {"class_type": "CheckpointLoaderSimple", "inputs": {"ckpt_name": "sd_xl_base_1.0.safetensors"}},
            "6": {"class_type": "CLIPTextEncode", "inputs": {"text": prompt, "clip": ["4", 1]}},
            "7": {"class_type": "CLIPTextEncode", "inputs": {"text": negative_prompt, "clip": ["4", 1]}},
            "5": {"class_type": "EmptyLatentImage", "inputs": {"width": width, "height": height, "batch_size": 1}},
            "3": {"class_type": "KSampler", "inputs": {
                "model": ["4", 0], "positive": ["6", 0], "negative": ["7", 0],
                "latent_image": ["5", 0], "seed": 42, "steps": 30, "cfg": 7.5,
                "sampler_name": "euler", "scheduler": "normal", "denoise": 1.0
            }},
            "8": {"class_type": "VAEDecode", "inputs": {"samples": ["3", 0], "vae": ["4", 2]}},
            "9": {"class_type": "SaveImage", "inputs": {"images": ["8", 0], "filename_prefix": "banner"}}
        }

Branded generation (LoRA)

async def generate_with_brand_lora(
    prompt: str,
    lora_path: str,  # trained LoRA on brand images
    lora_strength: float = 0.8
) -> bytes:
    """Generation in a specific brand style via LoRA"""
    workflow = self.build_lora_workflow(prompt, lora_path, lora_strength)
    return await self.run_workflow(workflow)

LoRA training for brand: 20–50 images in brand style, 500–2000 steps, 30–60 minutes on GPU.

Mass generation for e-commerce

async def generate_product_images(
    products: list[dict],
    background_style: str = "white studio"
) -> list[dict]:
    """Generate images for product catalog"""
    results = []

    for product in products:
        prompt = f"{product['name']}, {product['category']}, {background_style}, commercial photography style"

        image = await self.generate_banner(
            prompt=prompt,
            brand_colors=[],
            width=800,
            height=800
        )

        results.append({
            "product_id": product["id"],
            "image": image,
            "prompt_used": prompt
        })

    return results

Post-processing

from PIL import Image
import io

def add_brand_overlay(
    image_bytes: bytes,
    logo_path: str,
    watermark_text: str = None
) -> bytes:
    img = Image.open(io.BytesIO(image_bytes)).convert("RGBA")
    logo = Image.open(logo_path).convert("RGBA")

    # Scale logo to 15% of image width
    logo_width = img.width // 7
    logo_height = int(logo.height * (logo_width / logo.width))
    logo = logo.resize((logo_width, logo_height), Image.LANCZOS)

    # Place in bottom right corner
    position = (img.width - logo_width - 20, img.height - logo_height - 20)
    img.paste(logo, position, logo)

    output = io.BytesIO()
    img.save(output, format="PNG")
    return output.getvalue()

Timeline: setting up an AI designer with basic formats (banners, posts) – 1–2 weeks. LoRA training for brand + CMS integration – additional 1–2 weeks.

Generative AI Development: From Prompt to Production API

We often receive a task "generate a product image" — on the surface it seems simple. But behind this lies a choice between dozens of models, configuring the inference pipeline, manually solving consistency issues, integrating into the product backend, and answering why the model generates hands with six fingers in staging but not in production. Let's break down the directions we work with.

Image Generation: From Prompt to Production API

The current landscape includes FLUX.1 [dev/schnell/pro] from Black Forest Labs and Stable Diffusion 3.5. FLUX.1 [schnell] takes 4 steps instead of 20–50 for SDXL — 5–12 times faster — while maintaining higher quality. On an A100 80GB — 1.2–1.8 s per 1024×1024 image at batch_size=4.

A typical deployment issue: FLUX.1 [dev] requires 24+ GB VRAM in fp16. On A10G 24GB it fits tightly; at batch_size>1 — OOM. Solution: torch_dtype=torch.bfloat16 + enable_model_cpu_offload() from diffusers, or quantization via bitsandbytes to NF4 — minimal quality drop, memory consumption drops to 12–14 GB.

ControlNet and IP-Adapter are key tools for production tasks where controllability is needed. ControlNet with Canny/Depth/Pose maps provides structural control. IP-Adapter (especially IP-Adapter-FaceID) allows transferring character identity to generations — this is the foundation for personalized content. More about ControlNet can be found on Wikipedia.

Case study: e-commerce photography. A retailer with 8000 SKUs needed lifestyle photos for each product. Pipeline: product segmentation (Segment Anything Model 2) → background removal → inpainting with FLUX.1 [dev] using product image as IP-Adapter reference → upscale via RealESRGAN_x4plus. The generation cost is negligible compared to professional photography, providing huge savings. Throughput — 200 images/hour on 2× A100. Our extensive experience from 30+ projects ensures we select the optimal model for your task — an evaluation can be obtained upfront.

Why Is Model Selection Only Half the Battle?

Fine-tuning for a Specific Style or Character

Dreambooth and LoRA are the standard for adapting to a specific visual style or object. LoRA trains in 2–4 hours on 20–30 reference images on a single A100. Rank 16–32 is usually sufficient for style; rank 64+ is needed for precise face reproduction.

A common mistake: training LoRA too long — the model overfits to references, losing the ability to vary. Sign: at cfg_scale=7, all images look like copy-paste of references. Solved by early stopping (usually 1500–2000 steps for 20 images) and prior_preservation_loss.

For deeper customization — full fine-tuning via diffusers + accelerate with FSDP on multiple GPUs. But that already takes 40–80 hours of training and requires a truly large dataset (1000+ images).

Comparison of Image Generation Approaches

Video Generation: Which Models Are Best?

Open-weight video models still lag behind commercial ones in stability and length. Wan2.1 is the best choice for self-hosting: 14B parameters, runs on 2× A100, delivers acceptable quality for short clips.

The main pain of video generation is temporal consistency: the character changes clothing color at the third second, objects "drift." Partial solution — generation with motion_bucket_id and noise_aug_strength in Stable Video Diffusion, or using I2V (image-to-video) instead of pure text-to-video. As noted in VideoPoet research, consistency is achieved by training on long sequences.

AnimateDiff remains a working tool for short loops and motion effects on top of SD/FLUX. Not Sora, but deployable locally and predictable.

Music and Audio Generation

AudioCraft from Meta (MusicGen + AudioGen) is a production-ready stack for music generation. musicgen-large (3.3B) generates 30 s of music in ~8 s on A100. Control via text prompt and melody conditioning — you can specify a melody by humming.

Stable Audio Open from Stability AI is an alternative with length up to 47 s, better structural control (intro/verse/chorus). Deployment is similar: diffusers + FastAPI.

For voice-over and dubbing — ElevenLabs API or self-hosted XTTS v2 (see Speech AI service). For sound design and foley — AudioGen.

3D Generation: Current Practical State

3D generation has not yet reached the same maturity as 2D. But for specific tasks, tools are already working:

TripoSG and Shap-E — text/image-to-3D. Shap-E from OpenAI generates simple 3D meshes in seconds, but geometry is rough. TripoSG gives more detailed results but requires post-processing (remeshing, UV unwrapping).

Wonder3D and Zero123++ — 3D reconstruction from a single image. They work by generating multi-views (6–8 views) and then 3D reconstruction via NeuS or instant-ngp.

Gaussian Splatting (3DGS) — not generation, but reconstruction from a series of photos/videos. For product cards and real estate it's already production: 50–200 photos → 3DGS model in 15–30 min on RTX 4090 → interactive 3D viewer in browser.

What Infrastructure Is Needed for Generative AI Deployment?

Critical for generative models:

• Task queue — Celery + Redis or Ray Serve. Synchronous HTTP for image generation is unacceptable with >5 concurrent requests.
• Caching — similar prompts yield similar results. Semantic cache via embeddings (faiss + sentence-transformers) can reduce GPU load by 20–40%.
• Quality monitoring — CLIP score for text-image alignment, FID for evaluating generation distribution. Integrate into MLflow or Weights & Biases.
• Storage — generated images immediately to S3/MinIO, not on the inference server disk.

What's Included in the Deliverables

We take the project turnkey — from model selection to deployment and monitoring. The result includes:

• Model (or API integration) with performance benchmarks (latency p99, throughput).
• Pipeline documentation (prompt engineering guide, model card, dependency versions).
• Integration with your backend (REST/gRPC, queues).
• Configured monitoring (dashboards, alerts for quality drift).
• Training workshop for the team (2–4 hours).
• Warranty support for 3 months after launch — as part of our quality certificate.

We have completed 30+ projects in generative AI — this gives us the right to guarantee results.

How Is the Generative AI Development Process Structured?

1. Analysis (1–2 days): audit of current architecture, clarification of use case, selection of models and success metrics. We evaluate the project free of charge.
2. Proof of Concept (1–3 weeks): quick prototype on your data — to see real quality, not blog demos.
3. Design (1–2 weeks): pipeline architecture, infrastructure (GPU cluster/API), A/B testing plan.
4. Implementation and fine-tuning (4–12 weeks): development, LoRA/full fine-tuning, integration with queue and cache.
5. Testing (1–2 weeks): load tests, metric validation, edge-case verification (negative scenarios).
6. Deployment and monitoring (1–2 weeks): production deployment, monitoring setup, documentation.

Timeline Estimates

Integration of a ready API (DALL·E 3, Midjourney API, Stability API) — 1–2 weeks. Self-hosted pipeline with fine-tuning — 6–12 weeks. Full platform with UI, queues and monitoring — 3–6 months. The specific cost is calculated individually after analyzing your scenario.

Contact us — order a consultation, and we will select the optimal architecture for your project. Get a preliminary cost and timeline estimate for free.