Processor list

This page contains detailed information on all processors offered by Document AI. You can see a list of all processors by solution type.

All Document AI processors adhere to the Data Processing and Security Terms.

Refer to the Managing processor versions documentation for more details. Also, specific processor limits apply in addition to overall product quotas and limits.

Digitize text

Enterprise Document OCR (Optical Character Recognition)

Description

Identify and extract text in different types of documents.

This processor allows you to identify and extract text, including handwritten text, from documents in more than 200 languages. The processor also uses machine learning to perform a quality assessment of a document based on the readability of its content.

Category Digitize

Functions OCR, Quality Analysis

Release stage General availability

Access status Public

Type in API OCR_PROCESSOR

Supported languages

Full list of languages

Language Name	BCP 47 Tag	Script
Afrikaans	`af`	`Latn`
Albanian	`sq`	`Latn`
Arabic	`ar`	`Arab`
Armenian	`hy`	`Armn`
Belarusian	`be`	`Cyrl`
Bangla	`bn`	`Beng`
Bengali	`bn`	`Beng`
Bulgarian	`bg`	`Cyrl`
Catalan	`ca`	`Latn`
Chinese	`zh`	`Hani`
Croatian	`hr`	`Latn`
Czech	`cs`	`Latn`
Danish	`da`	`Latn`
Dutch	`nl`	`Latn`
English	`en`	`Latn`
Estonian	`et`	`Latn`
Filipino	`fil`	`Latn`
Finnish	`fi`	`Latn`
French	`fr`	`Latn`
German	`de`	`Latn`
Greek	`el`	`Grek`
Gujarati	`gu`	`Gujr`
Hebrew	`iw`	`Hebr`
Hindi	`hi`	`Deva`
Hungarian	`hu`	`Latn`
Icelandic	`is`	`Latn`
Indonesian	`id`	`Latn`
Italian	`it`	`Latn`
Japanese	`ja`	`Jpan`
Kannada	`kn`	`Knda`
Khmer	`km`	`Khmr`
Korean	`ko`	`Kore`
Lao	`lo`	`Laoo`
Latvian	`lv`	`Latn`
Lithuanian	`lt`	`Latn`
Macedonian	`mk`	`Cyrl`
Malay	`ms`	`Latn`
Malayalam	`ml`	`Mlym`
Marathi	`mr`	`Deva`
Nepali	`ne`	`Deva`
Norwegian	`no`	`Latn`
Persian	`fa`	`Arab`
Polish	`pl`	`Latn`
Portuguese (Portugal & Brazil)	`pt`	`Latn`
Punjabi	`pa`	`Guru`
Romanian	`ro`	`Latn`
Russian	`ru`	`Cyrl`
Serbian	`sr`	`Cyrl`
Slovak	`sk`	`Latn`
Slovenian	`sl`	`Latn`
Spanish	`es`	`Latn`
Swedish	`sv`	`Latn`
Tagalog	`tl`	`Latn`
Tamil	`ta`	`Taml`
Telugu	`te`	`Telu`
Thai	`th`	`Thai`
Turkish	`tr`	`Latn`
Ukrainian	`uk`	`Cyrl`
Vietnamese	`vi`	`Latn`
Yiddish	`yi`	`Hebr`

Processor versions

Version ID	Release Channel	Release Maturity	Description
`pretrained-ocr-v1.2-2022-11-10`	Stable	GA	Frozen model version of v1.0: Model files, configurations, and binaries of a version snapshot frozen in a container image for up to 18 months.
`pretrained-ocr-v2.0-2023-06-02`	Stable	GA	Production-ready model specialized for document use cases. Includes access to all OCR add-ons.
`pretrained-ocr-v2.1-2024-08-07`	Stable	GA	The main areas of improvement for v2.1 are: better printed text recognition, more precise checkbox detection and more accurate reading order.
`pretrained-ocr-v2.1.1-2025-01-31`	Release candidate	Public Preview	v2.1.1 is similar to V2.1, and is available in all regions except: `US`, `EU`, and `asia-southeast1`.

For more information, see Managing processor versions.

Quotas and limits

Maximum pages (online/synchronous requests):	15
Maximum pages (batch/offline/asynchronous requests):	500
Maximum pages (imageless mode online/synchronous requests):	30

Note: To extend the maximum page limit for online and synchronous requests up to 30, be sure to enable imageless_mode in the ProcessRequest.

Uptraining

Sample Input File Open in new window.

Sample Output Open in new window.

Supported regions

asia-south1
asia-southeast1
australia-southeast1
eu
europe-west2
europe-west3
northamerica-northeast1
us

More information Enterprise Document OCR

Extract entities from documents

Refer to Sample datasets for sample labeled and unlabeled datasets to use for training.

Custom Extractor

Description

Extract fields from documents using generative AI or custom models; fine-tune models to accurately extract data from your documents.

Category Extract

Functions OCR, Entity Extraction

Release stage General availability

Access status Public

Type in API CUSTOM_EXTRACTION_PROCESSOR

Notes

If using generative AI for extraction, then:
- Only the English language is officially supported.
- Region availability is in the US, EU, northamerica-northeast1 and asia-southeast1.

Supported languages

Full list of languages

Language Name	BCP 47 Tag	Script
Afrikaans	`af`	`Latn`
Arabic	`ar`	`Arab`
Azerbaijani	`az`	`Latn`
Azerbaijani (Cyrillic)	`az-Cyrl`	`Cyrl`
Belarusian	`be`	`Cyrl`
Bulgarian	`bg`	`Cyrl`
Bosnian	`bs`	`Latn`
Catalan	`ca`	`Latn`
Cebuano	`ceb`	`Latn`
Czech	`cs`	`Latn`
Welsh	`cy`	`Latn`
Danish	`da`	`Latn`
German	`de`	`Latn`
Greek	`el`	`Grek`
English	`en`	`Latn`
Esperanto	`eo`	`Latn`
Spanish	`es`	`Latn`
Estonian	`et`	`Latn`
Basque	`eu`	`Latn`
Persian	`fa`	`Arab`
Finnish	`fi`	`Latn`
Filipino	`fil`	`Latn`
French	`fr`	`Latn`
Irish	`ga`	`Latn`
Galician	`gl`	`Latn`
Hindi	`hi`	`Deva`
Croatian	`hr`	`Latn`
Haitian Creole	`ht`	`Latn`
Hungarian	`hu`	`Latn`
Indonesian	`id`	`Latn`
Icelandic	`is`	`Latn`
Italian	`it`	`Latn`
Hebrew	`iw`	`Hebr`
Japanese	`ja`	`Jpan`
Javanese	`jv`	`Latn`
Kazakh	`kk`	`Cyrl`
Korean	`ko`	`Kore`
Kyrgyz	`ky`	`Cyrl`
Latin	`la`	`Latn`
Lithuanian	`lt`	`Latn`
Latvian	`lv`	`Latn`
Macedonian	`mk`	`Cyrl`
Mongolian	`mn`	`Cyrl`
Marathi	`mr`	`Deva`
Malay	`ms`	`Latn`
Maltese	`mt`	`Latn`
Nepali	`ne`	`Deva`
Dutch	`nl`	`Latn`
Norwegian	`no`	`Latn`
Polish	`pl`	`Latn`
Pashto	`ps`	`Arab`
Portuguese (Portugal & Brazil)	`pt`	`Latn`
Romanian	`ro`	`Latn`
Russian	`ru`	`Cyrl`
Russian (Petrine Orthography)	`ru-PETR1708`	`Cyrl`
Sanskrit	`sa`	`Deva`
Slovak	`sk`	`Latn`
Slovenian	`sl`	`Latn`
Albanian	`sq`	`Latn`
Serbian	`sr`	`Cyrl`
Swedish	`sv`	`Latn`
Swahili	`sw`	`Latn`
Tagalog	`tl`	`Latn`
Turkish	`tr`	`Latn`
Ukrainian	`uk`	`Cyrl`
Urdu	`ur`	`Arab`
Uzbek	`uz`	`Latn`
Uzbek (Cyrillic)	`uz-Cyrl`	`Cyrl`
Vietnamese	`vi`	`Latn`
Yiddish	`yi`	`Hebr`
Chinese simplified	`zh-Hans`	`Hani`
Chinese traditional	`zh-Hant`	`Hani`
Zulu	`zu`	`Latn`

Processor versions

Version ID	Release Channel	Release Maturity	Description
`pretrained-foundation-model-v1.5-2025-05-05`	Stable	GA	Production-ready candidate powered by Gemini 2.5 Flash LLM. Recommended for those who want to experiment with newer models.
`pretrained-foundation-model-v1.5-pro-2025-06-20`	Stable	GA	Production-ready model powered by the Gemini 2.5 Pro LLM. Supports a quota of up to 30 pages per minute for online process requests. This model has improved quality compared to v1.5, and may have a higher latency.
`pretrained-foundation-model-v1.5.1-2025-08-07`	Release candidate	Public Preview	Public preview model powered by the Gemini 2.5 Flash LLM. This model has the same features as v1.5, and has improved adaptive few-shot learning.
`pretrained-foundation-model-v1.6-pro-2025-12-01`	Release candidate	Public Preview	Preview model powered by the Gemini 3 Pro LLM. Note: This processor version uses the Vertex AI Gemini global endpoint and is not compliant with Data Residency (DMZ) standards. For example, requests in US and EU endpoints might route to anywhere globally.
`pretrained-foundation-model-v1.6-2026-01-13`	Release candidate	Public Preview	Preview model powered by the Gemini 3 Flash LLM. Note: This processor version uses the Vertex AI Gemini global endpoint and is not compliant with Data Residency (DMZ) standards. For example, requests in US and EU endpoints might route to anywhere globally.
`pretrained-foundation-model-v3.5-2026-05-26`	Release candidate	Public Preview	Preview model powered by the Gemini 3.5 Flash LLM.

For more information, see Managing processor versions.

Quotas and limits

Maximum pages (online/synchronous requests):	15
Maximum pages (batch/offline/asynchronous requests):	200
Maximum pages (imageless mode online/synchronous requests):	30

Note: To extend the maximum page limit for online and synchronous requests up to 30, be sure to enable imageless_mode in the ProcessRequest.

Normalized data types

You can find more information in the Enrichment & normalization, and Create dataset pages.

Full list of normalized data types

dateTime as STRING
currency as STRING
money as google.type.Money
number as FLOAT or INTEGER

Uptraining

Sample Input File Open in new window.

Sample Output Open in new window.

Supported regions

asia-south1
asia-southeast1
australia-southeast1
eu
europe-west2
europe-west3
northamerica-northeast1
us

More information Custom Extractor

Form Parser

Description

Extract general key-value pairs (entity and checkbox), tables, and generic entities from documents in addition to OCR text.

This processor applies advanced machine learning technologies to extract key-value pairs, checkboxes, and tables from documents more than 200 languages. This processor also leverages deep learning models to extract 11 generic entities that are common in various document types.

Category Extract

Functions OCR, Form Parsing, Entity Extraction

Release stage General availability

Access status Public

Type in API FORM_PARSER_PROCESSOR

Supported languages

Full list of languages

Language Name	BCP 47 Tag	Script
Afrikaans	`af`	`Latn`
Albanian	`sq`	`Latn`
Arabic	`ar`	`Arab`
Belarusian	`be`	`Cyrl`
Catalan	`ca`	`Latn`
Chinese	`zh`	`Hani`
Croatian	`hr`	`Latn`
Czech	`cs`	`Latn`
Danish	`da`	`Latn`
Dutch	`nl`	`Latn`
English	`en`	`Latn`
Estonian	`et`	`Latn`
Filipino	`fil`	`Latn`
Finnish	`fi`	`Latn`
French	`fr`	`Latn`
German	`de`	`Latn`
Hebrew	`iw`	`Hebr`
Hindi	`hi`	`Deva`
Hungarian	`hu`	`Latn`
Icelandic	`is`	`Latn`
Indonesian	`id`	`Latn`
Italian	`it`	`Latn`
Japanese	`ja`	`Jpan`
Korean	`ko`	`Kore`
Latvian	`lv`	`Latn`
Lithuanian	`lt`	`Latn`
Macedonian	`mk`	`Cyrl`
Malay	`ms`	`Latn`
Marathi	`mr`	`Deva`
Nepali	`ne`	`Deva`
Norwegian	`no`	`Latn`
Persian	`fa`	`Arab`
Polish	`pl`	`Latn`
Portuguese (Portugal & Brazil)	`pt`	`Latn`
Romanian	`ro`	`Latn`
Russian	`ru`	`Cyrl`
Serbian	`sr`	`Cyrl`
Slovak	`sk`	`Latn`
Slovenian	`sl`	`Latn`
Spanish	`es`	`Latn`
Swedish	`sv`	`Latn`
Tagalog	`tl`	`Latn`
Turkish	`tr`	`Latn`
Ukrainian	`uk`	`Cyrl`
Vietnamese	`vi`	`Latn`
Yiddish	`yi`	`Hebr`

Processor versions

Version ID	Release Channel	Release Maturity	Additional fields detected	Description
`pretrained-form-parser-v1.0-2020-09-23`	Stable	GA	None	Legacy version. For best quality and full feature set, use the Form Parser v2.0.
`pretrained-form-parser-v2.0-2022-11-10`	Stable	GA	Show fields `email` `phone` `url` `date_time` `address` `person` `organization` `quantity` `price` `id` `page_number`	Recommended version. Supports generic entities and includes upgraded table, KVP, and checkbox model, as well as more than 200 languages.
`pretrained-form-parser-v2.1-2023-06-26`	Release Candidate	Public Preview	None	Public Preview version. Same model as v2.0 with native text extraction from digital PDF files enabled.

For more information, see Managing processor versions.

Quotas and limits

Maximum pages (online/synchronous requests):	15
Maximum pages (batch/offline/asynchronous requests):	100
Maximum pages (imageless mode online/synchronous requests):	30

Note: To extend the maximum page limit for online and synchronous requests up to 30, be sure to enable imageless_mode in the ProcessRequest.

Uptraining

Sample Input File Open in new window.

Sample Output Open in new window.

Supported regions

asia-south1
asia-southeast1
australia-southeast1
eu
europe-west2
europe-west3
northamerica-northeast1
us

More information Form Parser

Layout Parser

Description

Extracts document content elements (text, tables, and lists) and creates context-aware chunks.

Layout Parser extracts document content elements like text, tables, and lists, and creates context-aware chunks that facilitate information retrieval in generative AI and discovery applications.

Category Extract

Functions Layout Parsing, Document Chunking

Release stage General availability

Access status Public

Type in API LAYOUT_PARSER_PROCESSOR

Notes

This parser supports PDF, HTML, DOCX, PPTX, and XLSX/XLSM files.

Supported languages

Full list of languages

Language Name	BCP 47 Tag	Script
Afrikaans	`af`	`Latn`
Albanian	`sq`	`Latn`
Arabic	`ar`	`Arab`
Armenian	`hy`	`Armn`
Belarusian	`be`	`Cyrl`
Bangla	`bn`	`Beng`
Bengali	`bn`	`Beng`
Bulgarian	`bg`	`Cyrl`
Catalan	`ca`	`Latn`
Chinese	`zh`	`Hani`
Croatian	`hr`	`Latn`
Czech	`cs`	`Latn`
Danish	`da`	`Latn`
Dutch	`nl`	`Latn`
English	`en`	`Latn`
Estonian	`et`	`Latn`
Filipino	`fil`	`Latn`
Finnish	`fi`	`Latn`
French	`fr`	`Latn`
German	`de`	`Latn`
Greek	`el`	`Grek`
Gujarati	`gu`	`Gujr`
Hebrew	`iw`	`Hebr`
Hindi	`hi`	`Deva`
Hungarian	`hu`	`Latn`
Icelandic	`is`	`Latn`
Indonesian	`id`	`Latn`
Italian	`it`	`Latn`
Japanese	`ja`	`Jpan`
Kannada	`kn`	`Knda`
Khmer	`km`	`Khmr`
Korean	`ko`	`Kore`
Lao	`lo`	`Laoo`
Latvian	`lv`	`Latn`
Lithuanian	`lt`	`Latn`
Macedonian	`mk`	`Cyrl`
Malay	`ms`	`Latn`
Malayalam	`ml`	`Mlym`
Marathi	`mr`	`Deva`
Nepali	`ne`	`Deva`
Norwegian	`no`	`Latn`
Persian	`fa`	`Arab`
Polish	`pl`	`Latn`
Portuguese (Portugal & Brazil)	`pt`	`Latn`
Punjabi	`pa`	`Guru`
Romanian	`ro`	`Latn`
Russian	`ru`	`Cyrl`
Serbian	`sr`	`Cyrl`
Slovak	`sk`	`Latn`
Slovenian	`sl`	`Latn`
Spanish	`es`	`Latn`
Swedish	`sv`	`Latn`
Tagalog	`tl`	`Latn`
Tamil	`ta`	`Taml`
Telugu	`te`	`Telu`
Thai	`th`	`Thai`
Turkish	`tr`	`Latn`
Ukrainian	`uk`	`Cyrl`
Vietnamese	`vi`	`Latn`
Yiddish	`yi`	`Hebr`

Processor versions

Version ID	Release Channel	Release Maturity	Description
`pretrained-layout-parser-v1.0-2024-06-03`	Stable	GA	General availability version for document layout analysis. This is the default pre-trained processor version.
`pretrained-layout-parser-v1.5-2025-08-25`	Release Candidate	Public Preview	Preview version powered by Gemini 2.5 Flash LLM for better layout analysis on PDF files. Recommended for those who want to experiment with new versions. Note: If it's used for non-PDF files, it will have the same behavior as the stable `pretrained-layout-parser-v1.0-2024-06-03`.
`pretrained-layout-parser-v1.5-pro-2025-08-25`	Release Candidate	Public Preview	Preview version powered by Gemini 2.5 Pro LLM for better layout analysis on PDF files. v1.5-pro has higher latency than v1.5. Note: If it's used for non-PDF files, it will have the same behavior as the stable `pretrained-layout-parser-v1.0-2024-06-03`.
`pretrained-layout-parser-v1.6-pro-2025-12-01`	Release Candidate	Public Preview	Preview version powered by Gemini 3.0 Pro LLM. Note:This processor version uses the Vertex AI Gemini global endpoint and is not compliant with Data Residency (DMZ) standards. For example, requests in US and EU endpoints might route to anywhere globally.
`pretrained-layout-parser-v1.6-2026-01-13`	Release Candidate	Public Preview	Preview version powered by Gemini 3.0 Flash LLM. Note:This processor version uses the Vertex AI Gemini global endpoint and is not compliant with Data Residency (DMZ) standards. For example, requests in US and EU endpoints might route to anywhere globally.

For more information, see Managing processor versions.

Quotas and limits

Maximum pages (online/synchronous requests):	15
Maximum pages (batch/offline/asynchronous requests):	500
Maximum pages (imageless mode online/synchronous requests):	30

Note: To extend the maximum page limit for online and synchronous requests up to 30, be sure to enable imageless_mode in the ProcessRequest.

Uptraining

Sample Input File Open in new window.

Sample Output Open in new window.

Supported regions

eu
us

More information Layout Parser

Explore pretrained processors

Bank Statement Parser

Description

Extract from bank statements including name, account, transactions, etc.

Category Pretrained

Functions OCR, Entity Extraction

Release stage General availability

Access status Public

Type in API BANK_STATEMENT_PROCESSOR

Notes

If a page of a multi-page input file is the correct document type and one of the supported versions, the processor performs entity extraction on the first supported document. If the processor doesn't find any applicable documents in the input file, the processor returns an error message.

Supported languages

Language Name	BCP 47 Tag	Script	Handwriting supported
English	`en`	`Latn`

Processor versions

Version ID	Release Channel	Release Maturity	Description
`pretrained-bankstatement-v1.0-2021-08-08`	Stable	GA
`pretrained-bankstatement-v1.1-2021-08-13`	Stable	GA
`pretrained-bankstatement-v2.0-2021-12-10`	Stable	GA
`pretrained-bankstatement-v3.0-2022-05-16`	Stable	GA	This version assumes that the input file contains a single bank statement. Unlike the default version, this version does not check the input file for bank statements and will not return an error if no bank statements are found.
`pretrained-bankstatement-v4.0-2023-07-31`	Release Candidate	Public Preview
`pretrained-bankstatement-v5.0-2023-12-06`	Stable	GA

For more information, see Managing processor versions.

Quotas and limits

Maximum pages (online/synchronous requests):	15
Maximum pages (batch/offline/asynchronous requests):	30
Maximum pages (imageless mode online/synchronous requests):	30

Note: To extend the maximum page limit for online and synchronous requests up to 30, be sure to enable imageless_mode in the ProcessRequest.

Fields detected in the earliest version

You can also find this information in the Field detected page.

Full list of fields

account_number
account_type
bank_address
bank_name
client_address
client_name
ending_balance
starting_balance
statement_date
statement_end_date
statement_start_date
table_item

table_item/transaction_deposit
table_item/transaction_deposit_date
table_item/transaction_deposit_description
table_item/transaction_withdrawal
table_item/transaction_withdrawal_date
table_item/transaction_withdrawal_description

Enriched fields

You can find more information in the Enrichment & normalization page.

Full list of enriched fields

bank_address
bank_name

Normalized fields

You can find more information in the Enrichment & normalization page.

Full list of normalized fields

ending_balance
starting_balance
statement_date
statement_end_date
statement_start_date
table_item/transaction_deposit
table_item/transaction_deposit_date
table_item/transaction_withdrawal
table_item/transaction_withdrawal_date

Uptraining

Labeling Instructions Open in new window.

Sample Input File Open in new window.

Sample Output Open in new window.

Supported regions

eu
us

W2 Parser

Description

Extract from Form W2, including employee, employer, wages, etc.

Category Pretrained

Functions OCR, Entity Extraction

Release stage General availability

Access status Public

Type in API FORM_W2_PROCESSOR

Notes

If a page of a multi-page input file is the correct document type and one of the supported versions, the processor performs entity extraction on the first supported document. If the processor doesn't find any applicable documents in the input file, the processor returns an error message.

Supported languages

Language Name	BCP 47 Tag	Script	Handwriting supported
English	`en`	`Latn`

Supported form/versions

Processor list Stay organized with collections Save and categorize content based on your preferences.

Digitize text

Enterprise Document OCR (Optical Character Recognition)

Full list of languages

Extract entities from documents

Custom Extractor

Full list of languages

Full list of normalized data types

Form Parser

Full list of languages

Show fields

Layout Parser

Full list of languages

Explore pretrained processors

Bank Statement Parser

Full list of fields

Full list of enriched fields

Full list of normalized fields

W2 Parser

Processor list