Cómo usar
- Copia el comando y revisa el paquete antes de instalarlo en tu entorno de agente.
- Consulta las herramientas, credenciales y entornos necesarios en el documento original.
- Describe las entradas y el resultado deseado; contrasta la salida con las fuentes.
npx skills add https://github.com/yushulx/dynamsoft-sdk-skill --skill capture-vision-webAntes de empezar
- La instalación no incluye suscripción al modelo, cuota API ni permisos sobre tus archivos.
- Las estrellas e instalaciones reflejan actividad, no una prueba independiente de calidad o seguridad.
Contenido del repositorio de origen con títulos, ejemplos, código, tablas, enlaces e imágenes preservados.
Capture Vision on the web
One package, one router, several capabilities. The capability is chosen by the template string you pass to capture() / startCapturing(), not by a different API:
| Capability | Template |
|---|---|
| Barcodes / QR | ReadBarcodes_Default (ReadSingleBarcode, ReadBarcodes_SpeedFirst, ReadBarcodes_ReadRateFirst) |
| Document boundaries | DetectDocumentBoundaries_Default |
| Boundaries + normalization | DetectAndNormalizeDocument_Default |
| Normalize with your own corners | NormalizeDocument_Default |
| MRZ (passport / ID) | ReadPassportAndId (ReadPassport, ReadId) |
| Text lines / labels | RecognizeTextLines_Default |
Which package
| Need | Package | Version |
|---|---|---|
| Barcode only | dynamsoft-barcode-reader-bundle | 11.6.5000 |
| Barcode + document + MRZ + labels | dynamsoft-capture-vision-bundle | 3.6.5000 |
| A preset document scan screen (no custom UX) | dynamsoft-document-scanner | 1.5.0 |
| A preset MRZ scan screen (no custom UX) | dynamsoft-mrz-scanner | 4.1.0 |
The first two expose the same classes (LicenseManager, CaptureVisionRouter, CameraEnhancer, CameraView, CapturedResultReceiver); only the import changes. The last two are higher-level products with a preset UI. Use them when that screen is exactly what the product needs; any custom UX or workflow must use the Capture Vision native interfaces (CaptureVisionRouter + templates), not these wrappers.
npm install dynamsoft-capture-vision-bundle@3.6.5000Do not install dynamsoft-barcode-reader, which stopped at v10.
License
get_trial_licensewith the product (dcvfor the full bundle,dbrfor barcode-only,mds
for the Document Scanner, mrz for the MRZ Scanner) and package: "web".
status: "login_required"means a browser window opened — ask the user to finish the Google
sign-in, then call again. The session is reused.
- Store
licenseKeyin a build-time environment variable (VITE_DYNAMSOFT_LICENSE_KEY,
NEXT_PUBLIC_…) and show only licenseKeyMasked in chat.
get_license_guidewithplatform: "web"returns the init call for the package you chose.
Never put a literal key in generated source, and never load it from a URL or a checked-in file.
import { LicenseManager } from "dynamsoft-capture-vision-bundle";
await LicenseManager.initLicense(import.meta.env.VITE_DYNAMSOFT_LICENSE_KEY);Camera scanner (barcode)
Self-host the bundle so wasm and models come from the app's own origin:
mkdir -p public
cp node_modules/dynamsoft-capture-vision-bundle/dist/dcv.bundle.min.js public/<script src="./dcv.bundle.min.js"></script>
<div id="camera" style="width:100%;max-width:640px;height:480px"></div>
<div id="results">Starting…</div>
<script>
(async () => {
const { LicenseManager, CaptureVisionRouter, CameraEnhancer, CameraView } = Dynamsoft.CVR;
await LicenseManager.initLicense(window.DYNAMSOFT_LICENSE_KEY);
const view = await CameraView.createInstance();
document.getElementById("camera").appendChild(view.getUIElement());
const enhancer = await CameraEnhancer.createInstance(view);
const router = await CaptureVisionRouter.createInstance();
router.setInput(enhancer);
router.addResultReceiver({
onDecodedBarcodesReceived: (result) => {
const items = result.items ?? [];
if (items.length === 0) return;
document.getElementById("results").textContent =
items.map((item) => `[${item.formatString}] ${item.text}`).join(" | ");
}
});
await enhancer.open();
await router.startCapturing("ReadBarcodes_Default");
})().catch((error) => {
document.getElementById("results").textContent = `Error: ${error.message ?? error}`;
});
</script>Images and PDFs
Same router, no camera:
import { CaptureVisionRouter, EnumCapturedResultItemType } from "dynamsoft-capture-vision-bundle";
const router = await CaptureVisionRouter.createInstance();
const result = await router.capture(blob, "ReadBarcodes_Default");
for (const item of result.items ?? []) {
if (item.type !== EnumCapturedResultItemType.CRIT_BARCODE) continue;
console.log(item.formatString, item.text, item.location);
}Verified against dynamsoft-barcode-reader-bundle 11.6.5000 and dynamsoft-capture-vision-bundle 3.6.5000:
capture()accepts a `Blob`, a URL string, a base64 string, anHTMLImageElement,
HTMLCanvasElement or HTMLVideoElement. It does not accept a Uint8Array — passing one throws Type of 'imageOrFile' should be ....
- There is no
result.barcodeResultItemson these builds. Iterateresult.itemsand filter on
item.type, or read result.decodedBarcodesResult.items from a receiver callback.
- For a PDF, render the pages (pdf.js) and capture each page as a blob — the web build does not page
through a PDF by itself.
Document capture
Document and MRZ templates ship only with the Capture Vision bundle. Asking the barcode-only dynamsoft-barcode-reader-bundle for DetectDocumentBoundaries_Default fails with [capture] [-10036] The template name is invalid. — a common wrong turn when the project installed the smaller package.
import { CaptureVisionRouter, EnumCapturedResultItemType } from "dynamsoft-capture-vision-bundle";
const result = await router.capture(blob, "DetectAndNormalizeDocument_Default");
for (const item of result.items ?? []) {
switch (item.type) {
case EnumCapturedResultItemType.CRIT_DETECTED_QUAD: // corners of the page
console.log(item.location?.points);
break;
case EnumCapturedResultItemType.CRIT_DESKEWED_IMAGE: // corrected image
console.log(item.image ?? item);
break;
}
}Result item types in this build: CRIT_ORIGINAL_IMAGE = 1, CRIT_BARCODE = 2, CRIT_TEXT_LINE = 4, CRIT_DETECTED_QUAD = 8, CRIT_DESKEWED_IMAGE = 16, CRIT_PARSED_RESULT = 32, CRIT_ENHANCED_IMAGE = 64. Check item.type rather than relying on an aggregate getter.
No item means "no document found in this frame" — a normal result, not an error. For a live flow, run DetectDocumentBoundaries_Default while framing and call capture() with DetectAndNormalizeDocument_Default only when the user confirms; normalizing every frame is expensive.
Resolution matters: a 640×480 frame cannot produce a readable A4 page. Request 1920×1080 or higher and capture from a still frame.
MRZ
import { CaptureVisionRouter, EnumCapturedResultItemType } from "dynamsoft-capture-vision-bundle";
const result = await router.capture(blob, "ReadPassportAndId");
const parsed = (result.items ?? []).find(
(item) => item.type === EnumCapturedResultItemType.CRIT_PARSED_RESULT
);
if (parsed) {
console.log(parsed.codeType); // MRTD_TD1_ID | MRTD_TD2_ID | MRTD_TD3_PASSPORT
console.log(parsed.getFieldValue("documentNumber"));
console.log(parsed.getFieldValue("dateOfExpiry"));
}MRZ templates, like document templates, are only in the Capture Vision bundle — the barcode-only bundle rejects them as an invalid template name.
Field names are camelCase: documentType, firstName, lastName, sex, nationality, documentNumber, dateOfBirth, dateOfExpiry, issuingState. ReadPassport on an ID card returns nothing — default to ReadPassportAndId.
The MRZ Scanner web SDK (dynamsoft-mrz-scanner) carries its own scan UI and pipeline — use it when that screen fits. When the MRZ UX or workflow must be custom, use the templates above instead; the packaged scanner is not designed to be reshaped.
Labels and structured content
RecognizeTextLines_Default returns text-line items for zonal text (price tags, VIN plates, ID cards). Pair it with the Code Parser to turn raw strings into typed fields — for example VIN, GS1 AI, AAMVA. Ask for the specific parser you need rather than writing a regex by hand when a parser exists.
Bundler apps
A bundler hides the SDK's own script URL, so the engine fetches *.wasm from the app origin, gets index.html, and the browser reports Unexpected token '<'. Set the root before initLicense:
import { CoreModule } from "dynamsoft-capture-vision-bundle";
CoreModule.engineResourcePaths.rootDirectory = "/dynamsoft/";The property is rootDirectory, not root. Copy assets under the path the SDK appends:
mkdir -p public/dynamsoft/dynamsoft-capture-vision-bundle@3.6.5000
cp -r node_modules/dynamsoft-capture-vision-bundle/dist \
public/dynamsoft/dynamsoft-capture-vision-bundle@3.6.5000/references/frameworks.md has the React hooks pattern (with StrictMode handling) and the same lifecycle translated to Vue and Angular.
Pitfalls
- Camera needs a secure context — HTTPS or
http://localhost, otherwisegetUserMediais
undefined and nothing starts.
- Dispose everything —
router.dispose(),enhancer.dispose(),view.dispose(), and empty the
container. React StrictMode mounts twice in development and leaks a camera stream otherwise.
- `startCapturing` needs an open enhancer — open the camera first.
- Do not default to the RTU wrappers (
BarcodeScanner,DocumentScanner,MRZScanner). They
have a fixed UI and little control; use CaptureVisionRouter for anything real.
- One router per scanner instance; reusing a disposed router throws.
- Templates are strings — enums are convenience wrappers whose
.valueis the string.
Verify
initLicenseresolves before any capture call.- No 404 for
*.wasmor*.worker.jsin the network panel. - The camera indicator turns off after unmount.
- A still image decodes in the same page, which separates camera problems from decode problems.
Samples
- JavaScript barcode samples: https://github.com/Dynamsoft/barcode-reader-javascript-samples
- Document Normalizer samples: https://github.com/Dynamsoft/document-normalizer-javascript-samples
- Label Recognizer samples: https://github.com/Dynamsoft/label-recognizer-javascript-samples
- Code Parser samples: https://github.com/Dynamsoft/code-parser-javascript-samples
- Document Scanner: https://github.com/Dynamsoft/document-scanner-javascript
- MRZ Scanner JS: https://github.com/Dynamsoft/mrz-scanner-javascript
- Barcode + MRZ + document in one page:
https://github.com/yushulx/javascript-barcode-qr-code-scanner/tree/main/examples/barcodemrzdocument