AI Colorization: DeOldify & Stable Diffusion

Black-and-White Image Colorization with AI

Have an archive of black-and-white photos that need colorization, but manual coloring takes hours per image? We solve this with neural networks: DeOldify and Stable Diffusion img2img. They automatically assign colors based on context—sky blue, grass green, skin tones. No manual markup, minutes per batch. We processed over 50,000 frames for three museums and two film archives, averaging 0.4 seconds per image on an NVIDIA A100.

How AI Colorization Works

The neural network analyzes brightness, textures, and objects in the photo, then predicts the most likely color for each pixel. DeOldify uses a generative adversarial network trained on millions of black-and-white and color image pairs. Stable Diffusion in img2img mode adds control via text prompts—you can specify era, region, or even sky hue. We adapt both models to your source images—for historical portraits we increase the weight of skin in the loss function.

DeOldify — The Classic Coloring Approach

from deoldify import device
from deoldify.device_id import DeviceId
from deoldify.visualize import get_image_colorizer
import PIL.Image as Image
import io

device.set(device=DeviceId.GPU0)

colorizer = get_image_colorizer(artistic=True)  # artistic=True — more saturated colors

def colorize_image(image_bytes: bytes, render_factor: int = 35) -> bytes:
    """
    render_factor: 7-45, higher = more saturated, slower
    """
    input_image = Image.open(io.BytesIO(image_bytes)).convert("L").convert("RGB")

    # Save temporarily
    import tempfile, os
    with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as f:
        input_image.save(f.name)
        temp_path = f.name

    result = colorizer.get_transformed_image(temp_path, render_factor=render_factor)
    os.unlink(temp_path)

    buf = io.BytesIO()
    result.save(buf, format="JPEG", quality=95)
    return buf.getvalue()

Stable Diffusion img2img Approach

from diffusers import StableDiffusionImg2ImgPipeline
import torch

pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16
).to("cuda")

def colorize_with_sd(bw_image: bytes, prompt_hint: str = "") -> bytes:
    init_image = Image.open(io.BytesIO(bw_image)).convert("RGB")

    prompt = f"colorized photograph, natural colors, realistic{', ' + prompt_hint if prompt_hint else ''}"

    result = pipe(
        prompt=prompt,
        image=init_image,
        strength=0.5,  # Low strength preserves structure
        guidance_scale=8.0,
        num_inference_steps=30
    ).images[0]

    buf = io.BytesIO()
    result.save(buf, format="JPEG", quality=95)
    return buf.getvalue()

Comparison of Approaches

DeOldify Parameter Table

Why DeOldify is Better for Archival Photos

For archival photos, DeOldify produces the most natural colors—it is trained specifically on historical images and doesn't require manual prompts. Stable Diffusion img2img wins when you need to tailor to a specific era or artistic style. In our projects, we combine both: DeOldify for bulk colorization, SD for complex scenes.

Problems We Solve

• Loss of temporal consistency in video. DeOldify with temporal consistency enabled prevents colors from jumping between frames. For long videos we use GPU buffering and a sliding window.
• Uneven brightness and noise. Preprocessing (denoising with OpenCV, CLAHE for contrast) improves result quality by 20-30%.
• Unrealistic skin tones. We use models fine-tuned on early 20th-century portraits—datasets from the Museum of Modern Art are available.

How We Do It: A Case Study

For a museum with an archive of 5,000 black-and-white photos, we deployed a DeOldify-based pipeline with batch processing on an NVIDIA A100. Average time per photo was 0.3 seconds. For 10% of difficult frames (night, high contrast) we used SD img2img with a prompt describing the era. Result: all photos colorized in 2 days, colors natural, the museum used them for a digital exhibition. Time savings compared to manual colorization: 500x.

Process of Work

1. Analysis and Preparation — assess source quality, denoise, crop.
2. Pipeline Design — choose model (DeOldify / SD / combination), tune parameters.
3. Implementation — integrate API, batch processing, temporal consistency for video.
4. Testing — validate on a sample, A/B comparison, fine-tune render_factor / strength.
5. Deployment — containerize (Docker + FastAPI), schedule automatic runs.

What's Included in the Work

• Model and API documentation
• Access to the deployed service with dashboard
• Training for your engineers on pipeline usage
• Quality guarantee on first 100 frames (free revisions)
• 3 months of support after launch

Timeline and Cost

Integration timeline ranges from 1 to 3 days depending on volume. Cost is calculated individually: we assess the project, select the optimal GPU configuration and model. To get a consultation for your project, contact us—we'll design the optimal pipeline and estimate the cost. You can also order a test colorization of 10 frames—results within a day.

Our team has 10+ years in Computer Vision and 40+ projects in colorization and image restoration. We work with both one-time orders and large archives. Leave a request—we'll send examples and help bring your archival materials to life.

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.