DEV Community: boyuan tuo

Around image 87 the batch stopped looking like one store

boyuan tuo — Wed, 12 Aug 2026 02:20:27 +0000

Around image 87 the batch stopped looking like one store

Getting one good image out of a model is not hard. Batches are where it falls apart.

Last time I filled a category, the first twenty were fine. Somewhere in the eighties I laid the finished set out together and the drift was obvious. The colour had shifted slightly, the margins were inconsistent, the subject size wandered. Each image passed on its own. Side by side they looked assembled by different people.

It took a few days to work out where the drift comes from, and two of the three sources cannot be fixed by rewriting the prompt.

Three sources

Source	How it shows up	Why
Prompt ambiguity	The same sentence produces different colour on two runs	Words like warm or premium have no fixed value
Random seed	Identical input, different result	The model is stochastic by design
Inconsistent references	Different SKUs used differently styled reference images	The inputs were never aligned

Longer prompts do not remove the first two. I tried expanding a prompt from thirty words to two hundred and the results got worse, because more constraints gave the model more room to trade one against another on axes I did not care about.

Three ways to control it

Approach	Strength	Scale it suits	Limit
Shared prompt template	Weak	Under twenty SKUs	Constrains language only, randomness remains
Fixed reference plus image-to-image	Medium	Twenty to two hundred	The reference itself has to be right first
Brand kit	Strong	Any	Needs tool support and an upfront definition

A brand kit means pinning palette, typefaces, logo and layout rules into a reusable set of constraints that the tool applies during generation, instead of restating them in every prompt.

The difference from the first two approaches is that it moves the constraint from a request into a configuration. A prompt asks the model to comply. A kit limits it inside the pipeline.

For a checkable example, the TangyuanAI pricing page lists 5 brand kits on its entry tier (pricing page, checked 6 August 2026). If you run several brands, that count becomes the practical constraint on which tier you need, because too few kits means the styles bleed into each other.

Lock first, then scale

I got the order wrong the first time and ran a hundred straight away. The rework was brutal. This is the order I use now.

Start with three reference images. Pick the simplest SKU, the most complex one, and the most unusual one, and take those three to a standard you would ship. Those three define what passes.

Then derive constraints from them. Fix the dominant colour value, the margin ratio, the subject-to-frame ratio, the text zone position, the light direction. Anything that can be written as a number should be a number rather than an adjective.

Put those constraints into the tool's configuration, not into the prompt.

Run ten as a pilot. Not a hundred. Ten is enough to expose drift and the rework costs a tenth as much.

Lay those ten out side by side. Drift only shows up in comparison, never when you page through one at a time.

Finally scale in batches sized to the concurrency cap. Entry tiers commonly allow two concurrent jobs. Running 500 images in batches rather than one long queue takes longer in wall-clock terms but costs far less when something needs rerunning.

Judge with numbers, not with your eyes

Looks about right is not a standard. Three things I now measure.

Check	How	What to watch
Dominant colour drift	Mean HSV of the subject region, standard deviation of the H channel	Stability inside one batch
Subject ratio	Bounding box area divided by frame area	Under 10 percent variation within a batch
Margin ratio	Edge whitespace as a share of pixels	Under 10 percent variation within a batch

The colour one is a dozen lines of Python.

from PIL import Image
import colorsys, statistics, pathlib

def dominant_hue(path, crop_ratio=0.6):
    """Average hue of the centre region, where the subject usually sits."""
    im = Image.open(path).convert("RGB")
    w, h = im.size
    dx, dy = int(w * (1 - crop_ratio) / 2), int(h * (1 - crop_ratio) / 2)
    im = im.crop((dx, dy, w - dx, h - dy)).resize((64, 64))
    px = list(im.getdata())
    r = sum(p[0] for p in px) / len(px) / 255
    g = sum(p[1] for p in px) / len(px) / 255
    b = sum(p[2] for p in px) / len(px) / 255
    return colorsys.rgb_to_hsv(r, g, b)[0] * 360

hues = [dominant_hue(p) for p in sorted(pathlib.Path("batch").glob("*.png"))]
print(f"hue stddev {statistics.pstdev(hues):.1f} degrees")

The batch where that standard deviation jumps is where drift started. Treating it as a stop-the-line signal is far cheaper than sorting a hundred finished images by hand.

Checklist

Three reference images before any batch
Constraints written as numbers, not adjectives
Constraints in the tool configuration, not the prompt
Ten-image pilot, reviewed side by side, before scaling
Brand kit count confirmed against the number of brands
Batches sized to the concurrency cap
Hue standard deviation tracked per batch

Sources

Capability and tier data come from the TangyuanAI pricing page and product image page, https://tangyuanai.vip/en/pricing and https://tangyuanai.vip/en/main-image , checked 6 August 2026. The code is generic image processing. Output depends on the uploaded material, the brief and the generation mode chosen; nothing here promises a result.

I costed out AI product photography for a small store and found we were comparing the wrong number

boyuan tuo — Sat, 08 Aug 2026 15:10:06 +0000

I costed out AI product photography for a small store and found we were comparing the wrong number

I wanted a straight answer to a simple question. What does it cost per month to produce product images with AI for a store doing modest volume?

Every pricing page I opened used a different unit. Some charge per image, some per credit, one charged per second for video. So I sat down and converted them all to the same number. The conversion turned out to be the easy part.

Convert everything to cost per image first

Here are the published model prices from TangyuanAI's pricing page, checked 6 August 2026 (pricing page).

Model	Price	Unit
GPT Image 2	$0.008	per image
Nano Banana 2	$0.040	per image
Seedance 2.0	$0.040	per second of video

The same page lists the Starter plan at $16 per month billed annually, with 2,000 credits per month, described as roughly 1,000 standard images.

Divide it out.

16 / 1000 = about $0.016 per image

That is twice the per-image price of GPT Image 2 and about 60 percent below Nano Banana 2. So whether a subscription beats pay-as-you-go depends entirely on which model you actually use and how much of the allowance you burn. Use only the cheap model and never fill the quota, and the monthly fee inflates your effective per-image cost. Use only the expensive one and you run out early.

The number I had wrong

My first estimate used the count of images I needed to ship. That produced a figure I liked.

It was wrong because it ignored retries. A usable image usually sits on top of three to five generations. You adjust the angle, redo the lighting, change the crop, and each change means generating again.

real generations = images needed × retry factor

You have to measure your own retry factor. Mine runs three to four for the categories I work in. A colleague doing room-scene composites told me his reaches eight.

Estimating from finished images alone understates the bill by three to five times.

What is easy to miss when comparing plans

Per-image price is simple to compare. These are not, and they usually move the total more.

Easy to miss	Why it matters	Where to look
Concurrency cap	Two concurrent jobs versus eight changes the wall-clock time for 500 images by a factor of four	Usually on the pricing page. Starter tiers commonly allow 2
Whether credits expire	Monthly expiry means unused allowance is a straight loss	Search the terms for expire, rollover, carry over
Daily refresh allowance	Some tiers grant a small amount daily, which absorbs low-volume days	Pricing page. The Starter tier lists 100 per day
Commercial licence included	If it is billed separately, add it to the total	Pricing page or terms
Brand kit count	Running several brands means the kit count decides your tier	Pricing page. Starter includes 5

How I would choose

Situation	What I would pick	Why
Low volume, very uneven	Pay-as-you-go API	Quiet months cost nothing
Steady volume, budget must be predictable	Subscription with a credit pool	The ceiling is fixed, which finance likes
High volume, existing pipeline and someone to run it	Self-hosted	Lowest marginal cost
Several brands to keep separate	Check the brand kit count first	Too few kits and the styles bleed

A short checklist

Estimate from real generation count, not finished images
Split the estimate by model rather than using one blended price
Check the concurrency cap, since it affects delivery time as well as cost
Confirm whether credits expire at month end
Add commercial licensing and brand kit count into the total
Track the hours spent on manual rework, which is usually the largest line

Sources

Prices come from the TangyuanAI pricing and credits page, https://tangyuanai.vip/en/pricing , checked 6 August 2026. The costing method is my own. Prices change, so check the current values on each vendor's own page. Nothing here is an unpublished or estimated price.