/v1/diacritize
Diacritization
Add Persian short vowels and optional Ezafe while keeping the submitted letters, punctuation, spacing, and ZWNJ intact.
- Marked Persian
- Source-indexed operations
- Optional Ezafe evidence
VowelMarks API
Build Persian reading support into your product.
Add source-preserving diacritics, readable Pinglish, word-origin insights, and Persian-to-English translation through one developer interface.
Private developer access. Keys are issued manually.
POST /v1/diacritize
Authorization: Bearer vm_...
{
"text": "کتاب جدید",
"include_ezafe": true
}
// response
{
"original": "کتاب جدید",
"marked": "کِتابِ جَدید",
"operations": [...]
}Current status
The language engines exist. Self-serve API infrastructure does not yet.
This page describes the intended developer product and pilot pricing. There is no public key dashboard, automated billing, or instant activation today. We review each early integration directly before issuing access.
One key, four capabilities
Use only the Persian support your product needs.
Each endpoint returns structured evidence alongside its text result. Translation remains separate from the reading-cue pipeline.
/v1/pinglish
Pinglish
Turn Persian script into pronunciation-first Latin text using either a readable heritage style or a systematic style.
- Readable or systematic output
- Word alignment
- Pronunciation warnings
/v1/word-insights
Word insights
Identify supported word origins and suggest context-sensitive Persian alternatives without rewriting the source sentence.
- Exact source spans
- Evidence labels
- Ranked Persian alternatives
/v1/translate
Translation
Translate Persian into English as a separate result so products can preserve Persian as the primary reading surface.
- Persian to English
- Request ID parity
- Separate source and result
Pilot pricing
A shared balance across every endpoint.
These are the intended early-access rates. Exact limits and terms are confirmed with you before a key or invoice is issued.
How credits work
- 1x
- Diacritization or Pinglish character
- 2x
- Word-insight or translation character
Returned marks, metadata, and English output do not add credits.
Sandbox
$0one-time evaluation
For testing one integration before committing to a paid pilot.
- Included
- 25,000 credits
- Rate limit
- 10 requests / minute
- Additional use
- No overage
Builder
$29per month
For prototypes, small learning tools, and low-volume production use.
- Included
- 1 million credits / month
- Rate limit
- 60 requests / minute
- Additional use
- $29 per additional 1M credits
Growth
$99per month
For established apps adding Persian reading support to regular workflows.
- Included
- 5 million credits / month
- Rate limit
- 180 requests / minute
- Additional use
- $20 per additional 1M credits
Scale
$299per month
For higher-volume platforms that need coordinated capacity planning.
- Included
- 20 million credits / month
- Rate limit
- 600 requests / minute
- Additional use
- $15 per additional 1M credits
Higher volume
Need custom capacity or processing terms?
Tell us about monthly text volume, latency needs, regions, and support requirements.
Why this model
Text volume is more honest than request count.
A 50-character lookup should not cost the same as a 5,000-character article.
One balance lets a product combine marks, Pinglish, origins, and translation as needed.
Billing follows submitted text, not the number of marks or metadata fields returned.
API questions
What early access means.
Can I generate an API key now?
Not through a dashboard yet. Public key management and automated billing have not been built. Early keys are issued manually after we review the use case, expected volume, and data requirements with you.
Why bill by characters instead of requests?
A request might contain one word or several pages. Input-character billing makes those workloads comparable and prevents short requests from subsidizing unusually large ones.
What counts as a character?
The planned meter counts Unicode code points in the submitted input, including spaces and punctuation. Added diacritics and returned metadata do not add to the input count.
Can one key use every endpoint?
Yes. The pilot design uses one shared credit balance across diacritization, Pinglish, word insights, and translation, with no separate feature subscription.
Are failed requests billed?
The intended policy is to charge successful responses, not validation errors or VowelMarks service failures. Provider-attempt and retry details will be written into the pilot terms before access begins.
Is there a production SLA?
Not during manual pilot access. Products with uptime, throughput, regional processing, or contractual support requirements should describe those needs in the access request.
Start with a conversation
Tell us what you want Persian text to do.
Include your product, endpoints, expected text volume, and launch timing. We will reply with fit, current limitations, and pilot terms.