Product Description AI in India | unoblox

Writing one good product description is easy; writing a consistent, on-brand description for a catalog of a few thousand SKUs by hand is not. unoblox's rupee-billed, OpenAI-compatible endpoint fits this as a straightforward batch job — feed in structured product attributes, get back consistent copy, at a per-token cost you can estimate directly in rupees before you run it across your full catalog.

The bulk-copy problem

Catalogs grow faster than a copywriting team can keep up, and inconsistency creeps in when descriptions are written across different people or different months — different tone, different level of detail, different handling of specifications. A model applied consistently across every SKU, with the same prompt template, produces more uniform output than a growing team of writers ever will, though it still needs a spot-check pass before publishing.

Prompting for consistent output

• Fix the structure: ask for the same sections every time — a one-line hook, a short paragraph, then a bullet list of specifications.
• Feed structured input: pass attributes (material, size, price, category) rather than expecting the model to infer them, so it isn't guessing at facts about your product.
• Set the voice once: describe your brand tone in the system prompt so it's applied uniformly rather than re-explained per request.
• Cap the length: specify a word or character range so descriptions fit your product-page template without manual trimming.

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model":"deepseek-ai/deepseek-v4-flash",
    "messages":[{"role":"system","content":"Write a 60-80 word product description in a friendly, direct tone, ending with a short bullet list of specifications."},{"role":"user","content":"Product: cotton kurta, size S-XL, machine washable, price 899 rupees."}]
  }'

Model picks by catalog size

Model	Input (₹/1M)	Output (₹/1M)	Fit
Qwen3 1.7B	Free (₹0)	Free (₹0)	testing prompt templates before a full catalog run
DeepSeek V4 Flash	₹9.07	₹18.14	large catalogs, thousands of SKUs
Llama 4 Scout	₹10.08	₹30.24	general catalog copy with broader vocabulary
GPT-4.1	₹201.6	₹806.4	flagship or hero product pages where quality matters most

Other catalog models, like GPT-4.1 mini or Mistral Small, work the same way — see /models for live ₹ pricing before a full production run.

SEO and localisation notes

A model can naturally weave in the keywords you specify if you name them in the prompt, but don't expect it to know your actual search-ranking data — feed it the terms you already know matter. For regional-language storefronts, ask directly for the target language in the prompt; review a sample before trusting a full catalog run, since we don't publish per-language quality figures.

Frequently asked questions

Will the descriptions be duplicate-content across similar products? They can be, if your input attributes are nearly identical — vary the prompt slightly per product category, or include a distinguishing detail, to reduce repetition across similar SKUs.

Can this generate descriptions in Hindi or other Indian languages? Yes, ask for the target language directly in your prompt; review output quality on a sample before a full catalog run.

How much would a few thousand product descriptions cost? Estimate it directly: multiply your expected input and output tokens per description by the ₹ rate of your chosen model on /models — every rate is already published in rupees, so the estimate needs no extra step.

Should I review AI-generated descriptions before publishing? Yes — spot-check for factual accuracy against your actual product attributes, since the model only knows what you told it in the prompt.

Can I regenerate a description if I don't like the first result? Yes, each call is independent — rerun with an adjusted prompt or a higher-quality model if the first pass isn't a good fit.

Does this work for marketplace listings, not just my own storefront? Yes — the same API call structure works regardless of where the description ends up; just adjust the prompt to match each marketplace's format and length limits.

Get started in rupees → https://unoblox.ai/sign-in