Full OCR

This endpoint takes images or PDFs of identity documents and returns, for each document found, three things:

Classification — which document it is (driver's license (CNH), Brazilian ID card (RG), CRNM, CRLV…), from which country, and which side was seen.
Extraction — the fields written on the document, read by OCR.
Cross-check — when the document carries a CPF (the Brazilian individual taxpayer number), the data read is checked against the Brazilian Federal Revenue Service database, and the response says field by field whether it matched.

That is the difference from the Classifier, which only answers which document is this and reads nothing inside it.

This endpoint does not perform liveness (use Liveness), does not compare the face on the document with a selfie (use Face Match) and does not query the government's biometric database. To add validation in Datavalid, run by SERPRO (the Brazilian federal data processing agency), use Full OCR + Datavalid.

Proof of residence has an endpoint of its own

Full OCR recognizes proofs of residence and extracts the address from them, but the dedicated analysis is better. For proofs of residence, use Proof of residence.

Request

POST /full-ocr/v4

Headers

http

Authorization: ApiKey <your-api-key>

This endpoint also accepts the JWT token, in the format Authorization: Bearer <accessToken>. See JWT Token.

Parameters

All optional, in the query string.

Parameter	Description	Required
federalRevenueNumber	A CPF you already know for the person, to guide the search at the Federal Revenue Service. Accepts formatting. Can be repeated to send more than one CPF.	No
returnsFaceInfo	When `true`, the response carries the `face` object for each document. Default: `false`. Only has an effect on v4 — on v2 and v3 the face always comes.	No
returnsCroppedDocumentBase64	When `true`, each document carries the crop of the image in `documentBase64`. Default: `false`.	No
returnsMultiCroppedDocuments	When `true`, `documentBase64` becomes a list of crops instead of a single value. Depends on `returnsCroppedDocumentBase64`. Default: `false`.	No
returnsCroppedDocumentsInfo	When `true`, each crop becomes an object `{ side, b64 }` instead of a string. Depends on `returnsCroppedDocumentBase64`. Default: `false`.	No
forceLiveTaxData	When `true`, asks for the query to the Federal Revenue Service to be made live. Default: `false`.	No
documentscopy	When `true`, runs the document forensics analysis on CNH and RG. Default: `false`. Only has an effect on v4. See Document forensics.	No
groupAnalysisByField	Groups the document forensics result by field. Only makes sense together with `documentscopy=true`. Default: `false`.	No
analyzeForgery	Adds the tampering analysis to the document forensics result. Only makes sense together with `documentscopy=true`. Default: `false`.	No

🚧 The boolean parameters are only turned on by the string true

Any other value — 1, TRUE, yes, or the parameter with no value — is read as false. And, unlike other endpoints, an unknown query parameter does not raise an error here: it is simply ignored. A returnFaceInfo written without the s does not return 422; it returns 200 without the face.

🚧 federalRevenueNumber is validated by its check digits

A CPF that fails validation brings the whole request down with 422, before any processing. This applies to every value when you send more than one.

Accepted files

Item	Value
Formats	`image/png`, `image/jpeg`, `application/pdf`
Files per call	up to 7 files
Maximum size (multipart)	15 MB per file

The form field names are up to you — they come back in the response, in metadata.filesInfo. Sending more than seven files is rejected with 422.

Each page of a multi-page PDF is analyzed separately. From v3 on, the front and back of the same document found on different pages or in different files are merged into a single data entry.

Besides multipart/form-data, the files can be sent as JSON, in the base64 field, as described in Sending files.

Example request

bash

curl -i -X POST 'https://api-homolog.nxcd.app/full-ocr/v4' \
  --header 'Authorization: ApiKey YOUR_API_KEY' \
  --form 'frente=@./cnh-frente.jpg' \
  --form 'verso=@./cnh-verso.jpg'

Stating the expected CPF, to guide the query to the Federal Revenue Service:

bash

curl -i -X POST 'https://api-homolog.nxcd.app/full-ocr/v4?federalRevenueNumber=123.456.789-09' \
  --header 'Authorization: ApiKey YOUR_API_KEY' \
  --form 'documento=@./cnh.pdf'

Response

This section describes v4, the recommended version. v2 and v3 return a data in a different format — see Versions.

The envelope is the API standard:

Field	Description	Type
id	Unique identifier of the request	`String`
version	API version that served the call	`String`
data	List with one entry per document recognized. Comes back empty when no document was recognized.	`Object[]`
metadata	Metadata of the request	`Object`

The list comes ordered from the best to the worst result

data is ordered internally: documents with Federal Revenue Service data come first, then those with more fields matching, then those with more fields extracted. If you expect a single document, read data[0] — but check first that the list is not empty.

Each item of data is assembled from the blocks below. Not all of them always appear: the conditions are given in each section.

Block	What it is
`classification`	Which document it is, from which country, which side
`extraction` and `enhanced`	What was read from the document, raw and cleaned up
`taxData` and `matches`	What the Federal Revenue Service returned and what matched
`postOfficeData`	Address checked against the Correios (proofs of residence only)
`face`	The face found on the document
`documentBase64`	Crop of the document image
`metadata`	Files sent and processing time. Lives at the root, outside `data[]`.

data[].classification

What the AI recognized in the file, before reading any field.

Field	Description	Type
classification.type	Type of the document: `DriversLicense`, `FederalID`, `Passport`, `Transportation`, `ProofOfResidence`…	`String`
classification.subtype	Model within the type: `Printed`, `Digital`, `Decree2018Paper`, `CRLV-Printed`…	`String`
classification.country	Issuing country, in three letters: `BRA`, `UNI`, `USA`…	`String`
classification.side	Consolidated side of the document: `Front`, `Back` or `FrontAndBack`	`String`
classification.sameImage	`true` when front and back came from the same page of the same file	`Boolean`

🚧 On v4 the side vocabulary changes and sides disappears

On v3, classification.sides is a list with one entry per page classified, and the side of each one uses OnlyFront / OnlyBack / FrontAndBack.

On v4, the list is gone from classification: what is left is a single, already consolidated side, in the Front / Back / FrontAndBack vocabulary. The per-page detail moves to metadata.filesInfo[].details — that is where the confidence and the page of each side recognized now live.

json

{
  "classification": {
    "type": "DriversLicense",
    "subtype": "Printed",
    "country": "BRA",
    "side": "FrontAndBack",
    "sameImage": false
  }
}

data[].extraction and data[].enhanced

These are the fields read from the document, in two versions of the same data:

extraction — the text as it came out of the OCR, untreated. It is what was written on the paper, reading noise included.
enhanced — the same data normalized and corrected. This is the version that goes through the cross-check against the Federal Revenue Service: when the name matches, the CPF confirmed by the Federal Revenue Service is written here, and fields with OCR leftovers are cleaned up.

To fill in a form and to persist data, use enhanced. Use extraction when you need to know what was literally printed.

The two have the same structure, with the field-level differences marked in the table:

Field	Description	Type
schemaName	Identifier of the set of fields used in this extraction. Depends on the type and the model of the document.	`String`
person	The person's data	`Object`
person.taxId	CPF	`String`
person.name	Name	`String`
person.birthdate	Date of birth	`String`
person.parentage	Parents' names on a single line. Exists only in `extraction`.	`String`
person.mothersName	Mother's name. Exists only in `enhanced`, already split out of the parentage line.	`String`
person.fathersName	Father's name. Exists only in `enhanced`.	`String`
otherFields	The remaining fields of the document. The set of keys varies with the type and the model.	`Object`

An empty field is "", not absent

The keys of person and of otherFields are fixed for each document model: they always appear, and whatever was not read comes back as an empty string. You do not need to test whether the key exists — you need to test whether it is empty.

The set of otherFields keys is not the same in extraction and in enhanced — normalization splits some fields in two.

On a CNH, extraction carries driversLicenseCategory, expireAt, firstIssuedAt, formNumberFront, formNumberBack, issuedAt, locale, mopedsLicense, nacionality, naturalness, notes, permission, registerNumber, renach, securityNumber, sourceDocument, sourceDocumentIssuer and state. In enhanced, locale becomes localeCity and localeState, naturalness becomes naturalnessCity and naturalnessState, and the two formNumber* fields become a single formNumber; the rest is the same.

On a standard-model RG, extraction carries documentId, header, issuedAt, naturalness, origin and version. enhanced carries documentId, headerState, issuedAt, naturalnessCity, naturalnessState, originCity, originState, version, illiterate and ageTag.

json

{
  "extraction": {
    "schemaName": "driversLicense",
    "person": {
      "taxId": "123.456.789-09",
      "name": "JOAO DA SILVA SANTOS",
      "birthdate": "01/02/1990",
      "parentage": "ANTONIO SANTOS MARIA DA SILVA SANTOS"
    },
    "otherFields": {
      "driversLicenseCategory": "AB",
      "expireAt": "10/03/2029",
      "firstIssuedAt": "15/04/2010",
      "formNumberFront": "00123456789",
      "formNumberBack": "00123456789",
      "issuedAt": "10/03/2019",
      "locale": "SAO PAULO",
      "naturalness": "SAO PAULO SP",
      "registerNumber": "01234567890",
      "renach": "SP012345678",
      "securityNumber": "12345678901",
      "sourceDocument": "12.345.678-9",
      "sourceDocumentIssuer": "SSP SP",
      "state": "SP",
      "mopedsLicense": "",
      "nacionality": "BRASILEIRA",
      "notes": "",
      "permission": ""
    }
  },
  "enhanced": {
    "schemaName": "driversLicense",
    "person": {
      "taxId": "12345678909",
      "name": "JOAO DA SILVA SANTOS",
      "birthdate": "1990-02-01",
      "mothersName": "MARIA DA SILVA SANTOS",
      "fathersName": "ANTONIO SANTOS"
    },
    "otherFields": {
      "driversLicenseCategory": "AB",
      "expireAt": "10/03/2029",
      "firstIssuedAt": "15/04/2010",
      "formNumber": "00123456789",
      "issuedAt": "2019-03-10",
      "localeCity": "SAO PAULO",
      "localeState": "SP",
      "mopedsLicense": "",
      "nacionality": "BRASILEIRA",
      "notes": "",
      "permission": "",
      "registerNumber": "01234567890",
      "renach": "SP012345678",
      "securityNumber": "12345678901",
      "sourceDocument": "12.345.678-9",
      "sourceDocumentIssuer": "SSP SP",
      "state": "SP",
      "naturalnessCity": "SAO PAULO",
      "naturalnessState": "SP"
    }
  }
}

data[].taxData and data[].matches

This is where the external cross-check lives. taxData is what the Federal Revenue Service returned for the CPF; matches says, field by field, whether what was on the document matches what the Federal Revenue Service has.

Field	Description	Type
taxData.taxId	CPF found at the Federal Revenue Service	`String`
taxData.name	Name on record at the Federal Revenue Service	`String`
taxData.mothersName	Mother's name on record at the Federal Revenue Service	`String`
taxData.birthdate	Date of birth on record at the Federal Revenue Service	`String`
matches.name	`true` when the name read from the document matches the Federal Revenue Service's	`Boolean`
matches.mothersName	`true` when the mother's name matches	`Boolean`
matches.birthdate	`true` when the date of birth matches	`Boolean`

🚧 When the two blocks do not appear

taxData and matches are removed from the response for documents that have nothing to cross-check against the Federal Revenue Service — CRLV and ANTT, for example, and the Chilean identity card.

Beyond that, the query to the Federal Revenue Service depends on the side you sent, because the CPF is not printed on both sides of every document:

CNH — the query happens with front and back, or with the front alone. Sending only the back, it does not happen. And, for a driver's license, the query is only made when the country is BRA.
RG — the query happens with front and back; with the back alone, only on the standard models; with the front alone, only on the 2022 CIN.

When the query does not happen, the two blocks remain in the response, but with empty fields and every matches set to false. If you need the cross-check, send the whole document.

matches.name: false does not always mean "a different name"

A name read with fewer than 8 characters — typical of an OCR that failed — is treated as unchecked: matches.name comes back false without the comparison ever being made. And, when the name does not match, the cross-check stops there: the remaining fields are not used to correct enhanced.

json

{
  "taxData": {
    "taxId": "12345678909",
    "name": "JOAO DA SILVA SANTOS",
    "mothersName": "MARIA DA SILVA SANTOS",
    "birthdate": "1990-02-01"
  },
  "matches": {
    "name": true,
    "mothersName": true,
    "birthdate": true
  }
}

data[].postOfficeData

Only appears when the document recognized is a proof of residence. It carries the address the Brazilian Postal Service (Correios) returned for the ZIP code read, and the result of comparing it with what was written on the proof.

Field	Description	Type
postOfficeData.zipCode	ZIP code according to the Correios	`String`
postOfficeData.address	Street address according to the Correios	`String`
postOfficeData.district	District according to the Correios	`String`
postOfficeData.city	City according to the Correios	`String`
postOfficeData.state	State according to the Correios	`String`
postOfficeData.matches	Object with booleans for `zipCode`, `address`, `district`, `city` and `state`	`Object`

To analyze proofs of residence, prefer the dedicated endpoint, which documents this block in detail.

data[].face

The face found on the document. On v4 it only appears when you send returnsFaceInfo=true, and it is gone when no face was detected.

Field	Description	Type
face.age	Estimated apparent age	`Number`
face.gender.value	`Male` or `Female`	`String`
face.boundingBox	Position of the face in the image, in proportions between 0 and 1 (`top`, `left`, `width`, `height`)	`Object`
face.croppedBase64	Crop of the face in base64	`String`

json

{
  "face": {
    "age": 34,
    "gender": { "value": "Male" },
    "boundingBox": {
      "top": 0.2237228155136108,
      "left": 0.2856607735157013,
      "width": 0.1555942893028259,
      "height": 0.1687679588794708
    },
    "croppedBase64": "iVBORw0KGgoAAAANSUhEUgAAAlgAAAJYCAMAAA...K5CYII="
  }
}

data[].documentBase64

The crop of the document image, isolated from the rest of the photo. It only appears with returnsCroppedDocumentBase64=true, and its format depends on the other two parameters:

Parameters	Format of `documentBase64`
`returnsCroppedDocumentBase64=true`	`String` with the base64
`+ returnsCroppedDocumentsInfo=true`	`Object` with `side` and `b64`
`+ returnsMultiCroppedDocuments=true`	`Array` of the formats above

Document forensics

With documentscopy=true on v4, documents of type CNH and RG gain a data[].documentscopy block with the result of the automatic forensic examination of the fields. groupAnalysisByField=true reorganizes that block by field, and analyzeForgery=true adds a forgeryAnalysis to it with the tampering analysis.

The contents of the block vary with the model of the document and are not described here. If document forensics is your main use case, the dedicated product is Automated Document Forensics.

metadata

Field	Description	Type
metadata.filesInfo	List with one entry per file sent, including those where nothing was recognized	`Object[]`
metadata.filesInfo[].fieldname	Name of the field used to send the file	`String`
metadata.filesInfo[].name	Name of the file sent	`String`
metadata.filesInfo[].size	Size of the file, in bytes	`Number`
metadata.filesInfo[].pages	Number of pages of the file. For images, `1`.	`Number`
metadata.filesInfo[].mimetype	Type of the file sent	`String`
metadata.filesInfo[].encoding	Encoding of the file as sent	`String`
metadata.filesInfo[].sha256	SHA-256 hash of the file sent	`String`
metadata.filesInfo[].details	v4 only. One entry per side recognized in that file.	`Object[]`
metadata.filesInfo[].details[].side	Side recognized: `Front`, `Back` or `FrontAndBack`	`String`
metadata.filesInfo[].details[].page	Page where it was recognized. Starts at 0.	`Number`
metadata.filesInfo[].details[].confidence	Confidence of the classification of that side	`Number`
metadata.timeSpent	Processing time of the request, in milliseconds	`Number`

Full example

Status Code: 200

A driver's license sent in two files, front and back, with the CPF confirmed by the Federal Revenue Service:

json

{
  "id": "13cfec7c-b238-4820-a4d1-5173e4c1418e",
  "version": "v4",
  "data": [
    {
      "extraction": {
        "schemaName": "driversLicense",
        "person": {
          "taxId": "123.456.789-09",
          "name": "JOAO DA SILVA SANTOS",
          "birthdate": "01/02/1990",
          "parentage": "ANTONIO SANTOS MARIA DA SILVA SANTOS"
        },
        "otherFields": {
          "driversLicenseCategory": "AB",
          "expireAt": "10/03/2029",
          "firstIssuedAt": "15/04/2010",
          "formNumber": "00123456789",
          "issuedAt": "10/03/2019",
          "locale": "SAO PAULO",
          "mopedsLicense": "",
          "nacionality": "BRASILEIRA",
          "notes": "",
          "permission": "",
          "registerNumber": "01234567890",
          "renach": "SP012345678",
          "securityNumber": "12345678901",
          "sourceDocument": "12.345.678-9",
          "sourceDocumentIssuer": "SSP SP",
          "state": "SP"
        }
      },
      "enhanced": {
        "schemaName": "driversLicense",
        "person": {
          "taxId": "12345678909",
          "name": "JOAO DA SILVA SANTOS",
          "birthdate": "1990-02-01",
          "mothersName": "MARIA DA SILVA SANTOS",
          "fathersName": "ANTONIO SANTOS"
        },
        "otherFields": {
          "driversLicenseCategory": "AB",
          "expireAt": "10/03/2029",
          "firstIssuedAt": "15/04/2010",
          "formNumber": "00123456789",
          "issuedAt": "2019-03-10",
          "localeCity": "SAO PAULO",
          "localeState": "SP",
          "mopedsLicense": "",
          "nacionality": "BRASILEIRA",
          "notes": "",
          "permission": "",
          "registerNumber": "01234567890",
          "renach": "SP012345678",
          "securityNumber": "12345678901",
          "sourceDocument": "12.345.678-9",
          "sourceDocumentIssuer": "SSP SP",
          "state": "SP",
          "naturalnessCity": "SAO PAULO",
          "naturalnessState": "SP"
        }
      },
      "taxData": {
        "taxId": "12345678909",
        "name": "JOAO DA SILVA SANTOS",
        "mothersName": "MARIA DA SILVA SANTOS",
        "birthdate": "1990-02-01"
      },
      "matches": {
        "name": true,
        "mothersName": true,
        "birthdate": true
      },
      "classification": {
        "type": "DriversLicense",
        "subtype": "Printed",
        "country": "BRA",
        "side": "FrontAndBack",
        "sameImage": false
      }
    }
  ],
  "metadata": {
    "filesInfo": [
      {
        "fieldname": "frente",
        "name": "cnh-frente.jpg",
        "size": 567098,
        "pages": 1,
        "mimetype": "image/jpeg",
        "encoding": "7bit",
        "sha256": "41c2ece339abfcf9ba1e7d5a60666162a24ef34ba95adeb41ed1d9721c91c6b0",
        "details": [
          { "side": "Front", "confidence": 0.99, "page": 0 }
        ]
      },
      {
        "fieldname": "verso",
        "name": "cnh-verso.jpg",
        "size": 512044,
        "pages": 1,
        "mimetype": "image/jpeg",
        "encoding": "7bit",
        "sha256": "2c26b46b68ffc68ff99b453c1d30413413422d706483bfa0f98a5e886266e7ae",
        "details": [
          { "side": "Back", "confidence": 0.97, "page": 0 }
        ]
      }
    ],
    "timeSpent": 9840
  }
}

No document recognized

Status Code: 200

A file with no recognizable document is not an error: the response is 200, data comes back empty, and metadata.filesInfo still lists what was sent.

json

{
  "id": "6a1d1e8c-4a7d-4a5e-9f2c-1b0f2a9c7e11",
  "version": "v4",
  "data": [],
  "metadata": {
    "filesInfo": [
      {
        "fieldname": "documento",
        "name": "borrada.jpg",
        "size": 152064,
        "pages": 1,
        "mimetype": "image/jpeg",
        "encoding": "7bit",
        "sha256": "2c26b46b68ffc68ff99b453c1d30413413422d706483bfa0f98a5e886266e7ae"
      }
    ],
    "timeSpent": 3120
  }
}

Response headers

Header	When it appears
`Nextid-ReqId`	On every response. Carries the request identifier, the same one as the `id` field in the body.

Errors

Code	When it happens
401 Unauthorized	API key or JWT token missing, invalid, or without permission for this endpoint.
404 Not Found	The version given in the URL does not exist. Only `v2`, `v3` and `v4` are accepted.
406 Not Acceptable	The body was sent as JSON, but without the `base64` field.
413 Payload Too Large	One of the files sent as `multipart/form-data` is larger than 15 MB.
415 Unsupported Media Type	The `Content-Type` header is missing or is not supported.
422 Unprocessable Entity	A `federalRevenueNumber` that fails check-digit validation, a file format that is not accepted, or more than seven files.
500 Internal Server Error	Unexpected failure during processing.

Example response with an invalid CPF:

Status Code: 422

json

{
  "id": "13cfec7c-b238-4820-a4d1-5173e4c1418e",
  "error": {
    "statusCode": 422,
    "error": "Unprocessable Entity",
    "message": "Invalid Federal Revenue Number"
  }
}

The error response format is described in HTTP response codes.

Versions

The recommended version is v4, called at /full-ocr/v4. It is the only one that separates extraction from enhanced, that uses the Federal Revenue Service result to correct the extracted data, and where the query parameters take effect.

‼️ Omitting the version hands the call to v2

POST /full-ocr, with no version in the path, is served by v2 — not by v4. v2 has a different response format and ignores every query parameter described in the Request section. Always state the version in the URL.

Version	Status	What changes
v4	Recommended	`extraction` and `enhanced` separated, `taxData`/`matches` at the root of the item, consolidated `classification.side`, `details` inside `filesInfo`. The only one where the query parameters apply.
v3	Available	A single `extraction` object, with no `enhanced`. The Federal Revenue Service cross-check lives in `federalRevenueData`. `classification.sides` as a list.
v2	Legacy	Does not merge front and back: each side becomes a `data` entry of its own. `classification` uses the old names (`CNH`, `RG`). This is what answers when the version is omitted from the URL.

What changes in v3

data[] stops having extraction/enhanced separated and instead has:

Field	Description	Type
extraction	A flat object with the fields read from the document, with no split between raw and cleaned	`Object`
federalRevenueData	The Federal Revenue Service cross-check, now nested	`Object`
federalRevenueData.name	Name at the Federal Revenue Service	`String`
federalRevenueData.federalRevenueNumber	CPF at the Federal Revenue Service	`String`
federalRevenueData.mothersName	Mother's name at the Federal Revenue Service	`String`
federalRevenueData.birthdate	Date of birth at the Federal Revenue Service	`String`
federalRevenueData.matches	Object with `name`, `federalRevenueNumber`, `mothersName` and `birthdate`, each carrying `matched`	`Object`
classification.type	Type of the document	`String`
classification.subtype	Model of the document	`String`
classification.country	Issuing country	`String`
classification.sides	List with the `side`, `page`, `fieldname` and `confidence` of each side recognized	`Object[]`
postOfficeData	Same format as v4	`Object`
face	Same format as v4, but always present — `returnsFaceInfo` has no effect here	`Object`

Status Code: 200

json

{
  "id": "13cfec7c-b238-4820-a4d1-5173e4c1418e",
  "version": "v3",
  "data": [
    {
      "extraction": {
        "name": "JOAO DA SILVA SANTOS",
        "federalRevenueNumber": "12345678909",
        "mothersName": "MARIA DA SILVA SANTOS",
        "birthdate": "01/02/1990",
        "registerNumber": "01234567890",
        "driversLicenseCategory": "AB",
        "expireAt": "10/03/2029"
      },
      "federalRevenueData": {
        "name": "JOAO DA SILVA SANTOS",
        "federalRevenueNumber": "12345678909",
        "mothersName": "MARIA DA SILVA SANTOS",
        "birthdate": "01/02/1990",
        "matches": {
          "name": { "matched": true },
          "federalRevenueNumber": { "matched": true },
          "mothersName": { "matched": true },
          "birthdate": { "matched": true }
        }
      },
      "classification": {
        "type": "DriversLicense",
        "subtype": "Printed",
        "country": "BRA",
        "sides": [
          { "side": "OnlyFront", "page": 0, "fieldname": "frente", "confidence": 0.99 },
          { "side": "OnlyBack", "page": 0, "fieldname": "verso", "confidence": 0.97 }
        ]
      },
      "face": {
        "age": 34,
        "gender": { "value": "Male" },
        "boundingBox": { "top": 0.22, "left": 0.28, "width": 0.15, "height": 0.16 },
        "croppedBase64": "iVBORw0KGgoAAAANSUhEUgAAAlgAAAJYCAMAAA...K5CYII="
      }
    }
  ],
  "metadata": {
    "filesInfo": [
      {
        "fieldname": "frente",
        "name": "cnh-frente.jpg",
        "size": 567098,
        "pages": 1,
        "mimetype": "image/jpeg",
        "encoding": "7bit",
        "sha256": "41c2ece339abfcf9ba1e7d5a60666162a24ef34ba95adeb41ed1d9721c91c6b0"
      }
    ],
    "timeSpent": 8120
  }
}

What changes in v2

Besides having no enhanced, v2 has two differences that break consumption:

It does not merge front and back. Each side recognized becomes a separate data entry, each with page and fieldname at the root of the item. A driver's license sent in two files returns two entries, not one.
classification uses a different vocabulary. type comes translated into CNH, RG, CPF, PROOF-OF-RESIDENCE, SELFIE, IMPRESSOS, CARTAOCREDITO or OTHERS, and the side lives in classification.face, with the values front, back or front-back. There is no subtype and no country.

The Federal Revenue Service cross-check lives in federalRevenueData, with the matches in federalRevenueData.matched — note the singular name, different from v3 — and each field carrying matched, confidence (0 to 1) and strategy (char-to-char or by-tokens).

Status Code: 200

json

{
  "id": "13cfec7c-b238-4820-a4d1-5173e4c1418e",
  "version": "v2",
  "data": [
    {
      "page": 0,
      "fieldname": "frente",
      "extraction": {
        "name": "JOAO DA SILVA SANTOS",
        "federalRevenueNumber": "12345678909",
        "mothersName": "MARIA DA SILVA SANTOS",
        "birthdate": "01/02/1990",
        "registerNumber": "01234567890",
        "driversLicenseCategory": "AB"
      },
      "federalRevenueData": {
        "name": "JOAO DA SILVA SANTOS",
        "federalRevenueNumber": "12345678909",
        "mothersName": "MARIA DA SILVA SANTOS",
        "birthdate": "01/02/1990",
        "matched": {
          "name": { "matched": true, "confidence": 1, "strategy": "char-to-char" },
          "federalRevenueNumber": { "matched": true, "confidence": 1, "strategy": "char-to-char" },
          "mothersName": { "matched": true, "confidence": 0.97, "strategy": "by-tokens" },
          "birthdate": { "matched": true, "confidence": 1, "strategy": "char-to-char" }
        }
      },
      "classification": {
        "confidence": 0.99,
        "type": "CNH",
        "face": "front"
      },
      "face": {
        "age": 34,
        "gender": { "value": "Male", "confidence": 99.72 },
        "boundingBox": { "top": 0.22, "left": 0.28, "width": 0.15, "height": 0.16 },
        "croppedBase64": "iVBORw0KGgoAAAANSUhEUgAAAlgAAAJYCAMAAA...K5CYII="
      }
    }
  ],
  "metadata": {
    "filesInfo": [
      {
        "fieldname": "frente",
        "name": "cnh-frente.jpg",
        "size": 567098,
        "pages": 1,
        "mimetype": "image/jpeg",
        "encoding": "7bit",
        "sha256": "41c2ece339abfcf9ba1e7d5a60666162a24ef34ba95adeb41ed1d9721c91c6b0"
      }
    ],
    "timeSpent": 7450
  }
}

‼️ Switching versions breaks whoever reads the response

The path to one and the same piece of data changes across all three versions. The mother's name confirmed by the Federal Revenue Service is in data[].federalRevenueData.mothersName on v2 and v3, and in data[].taxData.mothersName on v4. The result of that check is in data[].federalRevenueData.matched.mothersName.matched on v2, in data[].federalRevenueData.matches.mothersName.matched on v3, and in data[].matches.mothersName on v4.

Code written for one version does not work on the other without changes.

Classifier

Full OCR ​

Request ​

Headers ​

Parameters ​

Accepted files ​

Example request ​

Response ​

data[].classification ​

data[].extraction and data[].enhanced ​

data[].taxData and data[].matches ​

data[].postOfficeData ​

data[].face ​

data[].documentBase64 ​

Document forensics ​

metadata ​

Full example ​

No document recognized ​

Response headers ​

Errors ​

Versions ​

What changes in v3 ​

What changes in v2 ​

Full OCR

Request

Headers

Parameters

Accepted files

Example request

Response

data[].classification

data[].extraction and data[].enhanced

data[].taxData and data[].matches

data[].postOfficeData

data[].face

data[].documentBase64

Document forensics

metadata

Full example

No document recognized

Response headers

Errors

Versions

What changes in v3

What changes in v2