Recognitize handwritten text to latex using API of myscript.com
which declares MyScript has been working on this matter for over two decades now and there is no question that our technology is ahead of Microsoft Ink Recognizer. at their article.
Create a account at https://developer.myscript.com/getting-started/web to get token.
It is free for 2000 calls/month.

Paste your token on PasteYourTokenHere, and push this into setting.json following the previous section Customize Buttons. The json generated from this script

{"type":"script","version":"0.1.2","function":"var token = {\n applicationKey: 'PasteYourTokenHere',\n hmacKey: 'PasteYourTokenHere',\n}\nvar addGap = true\n\nvar iink_cdn = 'https://cdn.jsdelivr.net/npm/iink-js@1.4.5/dist/iink.min.js'\n// var iink_cdn = https://unpkg.com/iink-js@1.4.5/dist/iink.min.js\n// var iink_cdn = https://myscript.github.io/iinkJS/dist/iink.min.js\n\nvar icon = 'square-root-alt'\nvar title = 'Recognize to latex'\n\ndocument.querySelector('div.svg-operate').insertAdjacentHTML('beforeEnd', (addGap ? \"<span class='svgiconspliter'></span>\" : \"\") + \"<div class='svg-btn fa fa-\" + icon + \"' title='\" + title + \"'><span></span></div>\");\nvar btnElement = document.querySelector('div.svg-operate > :last-child')\nbtnElement.onclick = ()=>{drawAPI.unstable.setTextContent('loading script')}\n\nvar svg = drawAPI.unstable.getSVGElement()\n\nvar getStrokeGroups = () => {\n let strokes = []\n for (const item of svg.children) {\n if (item.nodeName === 'path') {\n let points = item\n .getAttributeNS(null, \"d\")\n .split(/M |L /g)\n .slice(1)\n .map(item => {\n return {\n x: parseFloat(item.split(\",\")[0]),\n y: parseFloat(item.split(\",\")[1])\n };\n })\n strokes.push({ x: points.map(v => v.x), y: points.map(v => v.y) })\n }\n }\n console.log(strokes);\n return [{ \"penStyle\": null, \"strokes\": strokes }]\n}\n\nvar s1 = document.createElement('script')\ns1.setAttribute('nonce', drawAPI.unstable.nonce())\ns1.src = iink_cdn\ns1.onload = () => {\n\n const onclick = () => {\n drawAPI.unstable.setTextContent('calling the API')\n // Creating a recognizer\n const iinkRecognizer = iink.DefaultBehaviors.recognizerList.find(x => {\n const infos = x.getInfo();\n return infos.protocol === 'REST';\n });\n\n // Creating a empty model\n const model = iink.InkModel.createModel();\n // Filling the model with the stroke groups\n model.strokeGroups = getStrokeGroups();\n\n // Creating a recognizer context with the configuration attached\n const recognizerContext = iink.RecognizerContext.createEmptyRecognizerContext({\n configuration: iink.DefaultConfiguration\n });\n\n recognizerContext.editor.configuration.recognitionParams = {\n type: 'MATH',\n protocol: 'REST',\n server: {\n scheme: 'https',\n // host: 'webdemoapi.myscript.com',\n host: 'cloud.myscript.com',\n applicationKey: token.applicationKey,\n hmacKey: token.hmacKey\n },\n iink: {\n math: {\n mimeTypes: [\n 'application/x-latex',\n ],\n }\n }\n }\n\n // Assigning a theme to the document\n recognizerContext.editor.theme = iink.DefaultTheme;\n\n // Defining the behaviour on recognition result\n const recognitionCallback = (err, x) => {\n if (!err) {\n Object.entries(x.exports)\n .forEach(([mimeType, exportValue]) => {\n\n let latex = x.exports[mimeType]\n let content = '\\n$$'+latex.trim()+'$$ '+' '\n drawAPI.unstable.setTextContent('')\n drawAPI.unstable.editCurrentLine({\n control: 0,\n text: content\n })\n });\n }\n };\n\n // Triggering the recognition\n iinkRecognizer.export_(recognizerContext, model)\n .then((values) => {\n values.forEach((value) => {\n recognitionCallback(undefined, value);\n });\n })\n .catch(err => recognitionCallback(err, undefined));\n }\n\n btnElement.onclick = onclick\n}\ndocument.body.appendChild(s1)\n"}

Actually it is HTR not OCR.

The mathpix API is a popular choice for OCR to latex.
Its API it free for 1000/month, but you have to provide a card. They charge a one-time non-refundable setup fee of $1.

The following script also supports drag and paste image to latex by mathpix API.

And I am considering providing another independent extension to convert the clipbord-picture to latex.
Which need not to open a webview panel and provides a command and is able to bind keys.
(It will be https://github.com/zhaouv/vscode-paste-to-latex-mathpix.git if it becomes existence.)

{"type":"script","version":"0.1.2","function":"var token = { app_id: 'PasteYourTokenHere', app_key: 'PasteYourTokenHere' }\nvar addGap = true\nvar icon = 'square-root-alt'\nvar title = 'Recognize to latex'\n\ndocument.querySelector('div.svg-operate').insertAdjacentHTML('beforeEnd', (addGap ? \"<span class='svgiconspliter'></span>\" : \"\") + \"<div class='svg-btn fa fa-\" + icon + \"' title='\" + title + \"'><span></span></div>\");\nvar btnElement = document.querySelector('div.svg-operate > :last-child')\nbtnElement.onclick = () => {\n\n drawAPI.unstable.getPNG((dataURL) => {\n drawAPI.unstable.setTextContent('calling the API')\n xhrPost(dataURL, (err,ret)=>{\n console.log(err,ret)\n let latex = JSON.parse(ret)['latex']\n let content = '\\n$$'+latex.trim()+'$$ '+' '\n drawAPI.unstable.setTextContent('')\n drawAPI.unstable.editCurrentLine({\n control: 0,\n text: content\n })\n })\n })\n}\n\nfunction xhrPost(dataURL, callback) {\n var xhr = new XMLHttpRequest();\n xhr.onreadystatechange = function () {\n if (xhr.readyState == 4) {\n if ((xhr.status >= 200 && xhr.status < 300) || xhr.status == 304) {\n callback(null, xhr.responseText);\n } else {\n callback([xhr.status, xhr.responseText], null);\n }\n }\n }\n xhr.open('post', 'https://api.mathpix.com/v3/latex');\n xhr.setRequestHeader('app_id', token.app_id)\n xhr.setRequestHeader('app_key', token.app_key)\n xhr.setRequestHeader('Content-type', 'application/json')\n xhr.send(JSON.stringify({ 'url': dataURL }));\n}\n\n// drag and paste image\nfunction convertImage(dataURL) {\n drawAPI.unstable.setTextContent('calling the API to convert image')\n xhrPost(dataURL, (err,ret)=>{\n console.log(err,ret)\n let latex = JSON.parse(ret)['latex']\n let content = '\\n$$'+latex.trim()+'$$ '+' '\n drawAPI.unstable.setTextContent('')\n drawAPI.unstable.editCurrentLine({\n control: 0,\n text: content\n })\n })\n\n}\nfunction getImage(items, cb) {\n var file = null;\n if (items && items.length) {\n for (var i = 0; i < items.length; i++) {\n if (items[i].type.indexOf('image') !== -1) {\n file = items[i].getAsFile();\n break;\n }\n }\n }\n // console.log(file);\n if (file) {\n var reader = new FileReader()\n reader.onload = function (event) {\n cb(event.target.result);\n }\n reader.readAsDataURL(file);\n }\n}\nvar bindElement = document.body\nbindElement.addEventListener('paste', function (event) {\n var items = event.clipboardData?.items;\n getImage(items,convertImage);\n});\nbindElement.ondragover = function (ev) {\n ev.preventDefault();\n}\nbindElement.ondrop = function (ev) {\n ev.preventDefault();\n var items = ev.dataTransfer.items;\n getImage(items,convertImage);\n}\n"}

Paste your token on PasteYourTokenHere, and push this into setting.json following the previous section Customize Buttons. The json generated from this script