AI-Generated Product Images for E-Commerce

Studio photography of a single product costs hundreds of rubles, and for a catalog of 10,000 items, the budget becomes astronomical. Meanwhile, marketplace image requirements keep changing: Wildberries asks for 1000×1000 px, Amazon demands 2000×2000 with 85% padding, Ozon requires a white background without shadows. Reshooting each season or for each platform is unjustifiable. Our AI pipeline automates product image generation, saving up to 80% of the budget and accelerating the process to 2–5 seconds per frame. We use a combination of Stable Diffusion XL, ControlNet, and IP-Adapter to precisely preserve product shape and texture, and LoRA fine-tuning (rank 16) adapts the model to your catalog's unique features.

A typical problem is image incompatibility across platforms: Ozon requires a strict white background, while lifestyle scenes need creative environments. We solve this with a unified algorithm that feeds the appropriate background during generation based on the target platform. The approach requires no manual rework and scales across the entire catalog.

How AI Generates Product Images

At the core is Stable Diffusion XL with inpainting and ControlNet. The model is fine-tuned on your catalog using LoRA (rank 16) to preserve unique product attributes — texture, logo, shape. The entire pipeline runs on PyTorch with Triton Inference Server for low latency. We use dynamic prompts based on product category: for electronics — "modern minimalist desk setup", for clothing — "urban street style", for cosmetics — "marble surface with flowers". This ensures background relevance and boosts click-through rates on marketplaces. Without fine-tuning, the model hallucinates — distorts shapes or overlays foreign objects. LoRA with rank 16 solves this, preserving detail.

Background Removal + Re-Generation

from rembg import remove, new_session
from PIL import Image
from diffusers import StableDiffusionXLInpaintPipeline
import torch
import io
import numpy as np

class ProductImageGenerator:
    def __init__(self):
        self.bg_remover = new_session("isnet-general-use")
        self.pipe = StableDiffusionXLInpaintPipeline.from_pretrained(
            "diffusers/stable-diffusion-xl-1.0-inpainting-0.1",
            torch_dtype=torch.float16
        ).to("cuda")

    def remove_background(self, product_image: bytes) -> tuple[bytes, bytes]:
        """Returns (image without background, background mask)"""
        result = remove(product_image, session=self.bg_remover)
        img_rgba = Image.open(io.BytesIO(result)).convert("RGBA")

        r, g, b, a = img_rgba.split()
        mask = Image.fromarray(255 - np.array(a))

        img_rgb = Image.new("RGB", img_rgba.size, (255, 255, 255))
        img_rgb.paste(img_rgba, mask=img_rgba.split()[3])

        img_buf = io.BytesIO()
        img_rgb.save(img_buf, format="PNG")

        mask_buf = io.BytesIO()
        mask.save(mask_buf, format="PNG")

        return img_buf.getvalue(), mask_buf.getvalue()

    def place_on_background(
        self,
        product_image: bytes,
        background_prompt: str,
        steps: int = 30
    ) -> bytes:
        product_bytes, mask_bytes = self.remove_background(product_image)

        result = self.pipe(
            prompt=background_prompt,
            image=Image.open(io.BytesIO(product_bytes)),
            mask_image=Image.open(io.BytesIO(mask_bytes)),
            num_inference_steps=steps,
            guidance_scale=9.0,
            strength=0.99
        ).images[0]

        buf = io.BytesIO()
        result.save(buf, format="PNG")
        return buf.getvalue()

Lifestyle Scene Generation

PRODUCT_SCENE_PROMPTS = {
    "electronics": [
        "modern minimalist desk setup, natural light, laptop nearby, bokeh background",
        "cozy home office, wooden desk, plants, warm ambient lighting",
    ],
    "clothing": [
        "fashion editorial, urban street style, natural daylight",
        "lifestyle shot, outdoor park, casual summer day",
    ],
    "food": [
        "rustic wooden table, natural light, herbs nearby, shallow depth of field",
        "modern kitchen counter, marble surface, fresh ingredients",
    ],
    "cosmetics": [
        "marble surface with flowers, soft pink background, luxury aesthetic",
        "bathroom counter, morning light, clean minimalist style",
    ]
}

async def generate_product_scenes(
    product_image: bytes,
    product_category: str,
    num_variants: int = 4
) -> list[bytes]:
    prompts = PRODUCT_SCENE_PROMPTS.get(product_category, ["professional studio, white background"])
    results = []

    for prompt in prompts[:num_variants]:
        result = product_gen.place_on_background(
            product_image,
            f"product photography, {prompt}, high quality, 8k"
        )
        results.append(result)

    return results

Color Variant Generation

async def generate_color_variants(
    product_image: bytes,
    colors: list[str],
    product_mask: bytes
) -> dict[str, bytes]:
    """Changes product color via inpainting with mask"""
    results = {}

    for color in colors:
        from diffusers import StableDiffusionXLInpaintPipeline
        result = pipe(
            prompt=f"same product shape, {color} color, same material texture, product photography",
            image=Image.open(io.BytesIO(product_image)),
            mask_image=Image.open(io.BytesIO(product_mask)),
            strength=0.6,
            guidance_scale=10.0
        ).images[0]

        buf = io.BytesIO()
        result.save(buf, format="PNG")
        results[color] = buf.getvalue()

    return results

Why AI Is Cheaper Than Studio Photography

Budget savings reach up to 80%, and the project payback period is 2–3 months for catalogs with over 1000 items. For example, for a retailer with a catalog of several thousand products, we generated tens of thousands of images (main + lifestyle + color variants) in a few days. Studio photography would have taken a month and cost orders of magnitude more.

Marketplace Image Requirements

How We Implement the Solution

1. Catalog audit — analyze current photos, marketplace requirements, data volume.
2. Pipeline setup — fine-tune LoRA model on 20–50 of your products, select prompts.
3. Batch processing — generate all images (backgrounds, scenes, variants).
4. Optimization — conform to marketplace requirements (size, format, background).
5. Integration — connect API to your CMS or upload via FTP.
6. Quality control — automated validation plus spot manual checks.

Typical timelines: catalog up to 5,000 items — 1–2 weeks; up to 50,000 — 3–4 weeks.

What's Included in Our Project

• Pipeline and API documentation.
• Training your team to work with the model.
• 1 month of post-launch support.
• Guaranteed stable quality — fixed metrics and SLA.

Our team has several years of experience in computer vision and generative models, with over 50 e-commerce projects. We use certified open-source solutions and guarantee reproducible results.

Common Implementation Mistakes

• Using one model for all categories without fine-tuning — loss of detail.
• Ignoring marketplace requirements for padding and color profile.
• No A/B testing of AI photos vs studio photos — conversion can drop if the model hallucinates.

We set metrics for each case and conduct comparative analysis before launch. Order a pilot on 100 items — see the quality and speed for yourself. Contact us for a detailed audit of your catalog. Get a consultation on integrating AI generation into your business.

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.