Cloud PrimeroCloud PrimeroInternal tools
Cloud Primero for Aga Khan University

Storyblok AI translation: credit and cost model

Sizing a multilingual rollout across AKUH Pakistan, AKUH East Africa and the AKU-EB microsite against a 9.6 million AI credit allowance, using GPT-4o as the configured model.

10,558,258AI credits consumed by the selected scope
110%of the 9.6M annual allowance
Over budget by 958,258 credits. Either reduce scope, run GPT-4o on AKU's own key, or raise the quota through a contract amendment.
110% drawn
Consumed 10,558,258Allowance 9,600,000

Content inventory

Benchmark set at 5,000 pages across the three properties. Adjust counts and average translatable words per page as the Storyblok content audit firms up. Untick a row to exclude it from AI translation.

OnPage typePagesWords
per page		Words
Global chrome, navigation, footer, SEO meta12,500
Patient information and admissions175,000
Clinical specialty and service pages455,000
Programmes and academic pages, AKU-EB105,000
Doctor and practitioner profiles252,000
Facility and location pages135,000
Health education and condition articles540,000
Institutional, about, leadership72,000
Packages, price lists, test catalogues50,000
Legal, privacy, patient rights36,000
News, events and press archive360,000
Total in scope5,0002,192,500

Words counted are translatable field content only. Numeric tables, doctor schedules, lab test codes, prices and downloadable PDFs are excluded on the basis that they stay untranslated or are handled outside Storyblok.

Target languages

Output token factors reflect how each language tokenises against an English source word. Non-Latin scripts cost materially more per word. Only languages on Storyblok's supported list are shown.

Model and cost assumptions

OpenAI tokenisers average roughly 1.3 to 1.4 tokens per English word.
Covers the system prompt, AI branding rules, style groups and per-field JSON wrapping. Storyblok translates field by field, so short fields carry proportionally more overhead.
Storyblok states credits are a multiple of token cost and vary by model, but does not publish the table. Confirm the GPT-4o factor with the Customer Success Manager before committing to a number in the SOW.
Allows for regenerated translations, terminology corrections and content republished after the first pass.
Contracted allowance on the current plan. Premium and Elite quotas sit in the annual contract, so overage needs a contract amendment rather than an automatic top up.
Published Growth and Growth Plus pack price. Used to express credit consumption in money terms, not as a purchasable option on an enterprise plan.