If you have to rebuild anyway, rebuild where the meter stops.
ABBYY is moving FlexiCapture customers to Vantage. The industry reading of that path is blunt: it is a rebuild, not a transfer. There is no converter that turns a FlexiCapture document definition into a Vantage skill, because the two are architected differently — and such deployments are commonly quoted at eight to twelve weeks.
In short
The window closes as each account completes its migration.
An incumbent normally enjoys a switching cost: your templates, your integrations, your trained operators. When the vendor itself asks you to rebuild on a new product, that advantage is gone for as long as the decision is open.
What differs is what you get for the weeks: a platform with no per-page meter, no station licences, and a rebuild you do once rather than again at the next product transition. Once you have rebuilt elsewhere, the window does not reopen.
How a migration actually runs
- Scanner audit, before the demoWe ask what is on your floor first. eSCL-capable multifunction devices and modern departmental scanners are driven server-side with nothing installed. An estate that is TWAIN or ISIS only is a blocker unless it can scan to a folder, and we say so before you spend time on an evaluation.
- Inventory, import and blueprintWe list document types, field sets, validations and export targets. A FlexiCapture 12 project goes through the importer; anything else starts from a blueprint that creates the project, workflow and a full field set in one action — so week one starts on your documents, not on configuration screens.
- Template authoring, model-assistedDrop in a sample page and a draft template comes back — zones, field names, data types. An operator corrects it; the template trains from those corrections. Per-supplier layouts become alternatives on the same template rather than new projects.
- Parallel run, with field-level agreementSame input into both systems, then a report of where the two disagree, field by field. Every migration guide ends this way; we would rather start there. It is also the only fair way to settle the accuracy question on your own paper. There is no automatic parallel-run mode that feeds both and reports; the comparison is done on a sample set with the accuracy harness against results you already trust.
- Cutover, and the licence conversation you no longer haveExport mappings switch over, the old meter stops. Scaling recognition later is a hardware decision — add a node, declare its duties — not a procurement one.
What transfers from FlexiCapture 12, and what does not
Upload a zip of the whole project folder. A four-step wizard — upload, review plan, options, report — lists every item it found with one of three actions: import, approximate or skip, each with a reason. Nothing is created until you apply, nothing is dropped silently, and the workflow always arrives as a draft.
| In your FlexiCapture project | Becomes | Note |
|---|---|---|
| Document definition | Document template, one per section | FlexiCapture id, revision and build kept in the template settings |
| Fields, groups, repeating groups, tables | Fields in the same hierarchy, with data types and recognition languages | Captions become labels; required, read-only and hidden flags are only partly decoded |
| Custom data types | A pattern on the field | From the type's regular expression or alphabet |
| Database check and checksum rules | Dataset lookup and arithmetic rules | Column mapping and tolerance carried over; rule tags become severity |
| Other rules and scripts (C#, VB.NET, JScript, VBScript) | Approximated: ported to Java or JavaScript, inactive | Recognised idioms become built-in rules; the rest can be ported by a language model, one call per script, for you to review and activate |
| Dataset caches | Datasets with columns, keys and rows | Rows marked invalid are left out |
| Embedded sample page | Reference file on the template | |
| Standard workflow stages | A draft workflow: import → OCR → recognition → review → export | Low-confidence and failed-validation documents route to the Verification Hub |
| Page classifier and Classifier projects | Approximated: classifier with one category per class | Trained models are not reusable; the classifier is retrained here from its labelled pages |
| Field-extraction training batches | A template alternative per vendor, with sample files | Confirmed values and regions become the accuracy ground truth |
| PDFs in the upload | Documents, one batch per folder | Received, not processed, until you route them through a workflow |
| "Export to data files" profile | Filesystem export goal | You name the output directory; an upload endpoint and import profile are created too |
| FlexiLayout source | Skipped — encrypted | Extraction is rebuilt from the project's keyword lists as label search, patterns and table detection |
| Block geometry, separators, letter case | Skipped | Template defaults apply |
| Custom-action DLLs, .NET assemblies, verification bundle | Skipped — reported only | The scripts behind custom actions are ported |
| Processing Server settings, hot folders, schedules | Not in the project | The wizard asks for the one hot folder that matters |
| Custom stages and routing conditions | Not decoded | A custom stage becomes a script step stub; routing uses the standard guards |
Finishing by hand
- 01Run the template against the imported sample page and fix the fields that read nothing — usually a label that needs another keyword, or a field that wants a zone.
- 02Open each classifier: check categories, samples and accuracy, and retrain if the report shows a warning.
- 03Work through the report's approximated rows and the inactive scripts: review each port against the original, then activate or delete it.
- 04Set the export goal's output directory, check the workflow in the designer and activate it.
- 05Send a handful of real documents through and review them in the Verification Hub before switching on the hot folder.
Send us 200 of your most relevant pages.
Not the clean ones. The faxed ones, the stapled ones, the supplier who changes layout every quarter. We run them through the Tesseract, hybrid and model paths and report field-level accuracy back to you — including the fields we lose on.
A bake-off is scoped up front and runs one to two days — two at most: reading your documents, authoring templates, running the engines and writing up the field-level results with the cost per thousand pages of each mode.
You own the report either way, and if we lose, it says so.
Ask for a bake-off →Common questions
01Is open-source OCR really good enough to replace a commercial engine?
On clean 300 DPI print, Tesseract 5 is competitive. On degraded scans, dense forms and non-Latin scripts, the best commercial engines still read measurably better — we will not pretend otherwise. The answer for those documents is the hybrid path: Tesseract first, and low-confidence words or whole pages re-read by a vision model, local or remote, at your choice. The only way to settle it is a bake-off on your own documents.
02What happens to our existing templates?
From FlexiCapture 12, the importer brings fields, rules, datasets, scripts, classifiers and alternatives across as a starting configuration. From any other system, field sets, validation rules and export mappings are rebuilt — the same work any move between capture platforms requires. Blueprints and model-assisted template authoring shorten it. Blueprints ship pre-authored field sets, not pre-trained models.
03Can we keep our scanners?
If they speak eSCL (AirScan / Mopria) — which covers most networked multifunction devices and modern departmental scanners — yes, and the server drives them directly, so operators install nothing. Older production scanners that are TWAIN or ISIS only cannot be driven directly; they can still feed a hot folder. We ask what is on your floor before the demo rather than after it.
04What does a bake-off involve?
It starts with a fifteen-minute scanner and volume call, and a walk through the demo. Then someone reads your documents, authors templates, runs the Tesseract, hybrid and model paths over them, and writes up field-level accuracy with the cost per thousand pages of each mode. It runs one to two days — two is the ceiling, never a week. You keep the report whatever it concludes, including when it concludes that a competitor reads your paper better.
05Can both systems run in parallel during the move?
Yes, as two separate systems fed the same input by you. There is no built-in mode that mirrors documents to the old system and reports differences automatically.
Related
Statements about other vendors' products and plans reflect public reporting checked in August 2026 and may have changed. Product names are trademarks of their owners; KognitCapture is not affiliated with them.