diff --git a/packages/api/src/routers/article_router.ts b/packages/api/src/routers/article_router.ts
index 027cf91cd..59fa773da 100644
--- a/packages/api/src/routers/article_router.ts
+++ b/packages/api/src/routers/article_router.ts
@@ -22,28 +22,9 @@ import { getPageById, updatePage } from '../elastic/pages'
 import { generateDownloadSignedUrl } from '../utils/uploads'
 import { enqueueTextToSpeech } from '../utils/createTask'
 import { createPubSubClient } from '../datalayer/pubsub'
-import { htmlToSsmlItems, SSMLItem } from '@omnivore/text-to-speech-handler'
-import { WordPunctTokenizer } from 'natural'
-import { htmlToText } from 'html-to-text'
-
-interface Utterance {
-  wordOffset: number
-  wordCount: number
-  voice?: string
-  text: string
-  idx: number
-}
-
-interface SSMLOutput {
-  wordCount: number
-  averageWPM: number
-  language: string
-  defaultVoice: string
-  utterances: Utterance[]
-}
+import { htmlToSpeechFile } from '@omnivore/text-to-speech-handler'
 
 const logger = buildLogger('app.dispatch')
-const WORDS_PER_MINUTE = 200
 
 export function articleRouter() {
   const router = express.Router()
@@ -93,16 +74,16 @@ export function articleRouter() {
   })
 
   router.get(
-    '/:id/:outputFormat/:priority/:voice?',
+    '/:id/:outputFormat/:priority?/:voice?/:secondaryVoice?',
     cors<express.Request>(corsConfig),
     async (req, res) => {
       const articleId = req.params.id
       const outputFormat = req.params.outputFormat
-      const voice = req.params.voice || 'en-US-JennyNeural'
-      const priority = req.params.priority
+      const voice = req.params.voice
+      const priority = req.params.priority || 'high'
       if (
         !articleId ||
-        !['mp3', 'speech-marks', 'ssml'].includes(outputFormat) ||
+        !['mp3', 'speech-marks', 'speech-file'].includes(outputFormat) ||
         !['low', 'high'].includes(priority)
       ) {
         return res.status(400).send('Invalid data')
@@ -120,26 +101,17 @@ export function articleRouter() {
         },
       })
 
-      if (outputFormat === 'ssml') {
+      if (outputFormat === 'speech-file') {
         const page = await getPageById(articleId)
         if (!page) {
           return res.status(404).send('Page not found')
         }
-        const ssmlItems = htmlToSsmlItems(page.content, {
+        const speechFile = htmlToSpeechFile(page.content, {
           primaryVoice: voice,
-          secondaryVoice: 'en-US-GuyNeural',
-          rate: '1',
-          language: page.language || 'en-US',
+          secondaryVoice: req.params.secondaryVoice,
+          language: page.language,
         })
-        const [utterances, wordCount] = ssmlItemsToUtterances(ssmlItems)
-        const ssmlOutput: SSMLOutput = {
-          wordCount,
-          averageWPM: WORDS_PER_MINUTE,
-          language: page.language || 'en-US',
-          defaultVoice: voice,
-          utterances,
-        }
-        return res.send(ssmlOutput)
+        return res.send(speechFile)
       }
 
       const existingSpeech = await getRepository(Speech).findOne({
@@ -219,24 +191,3 @@ const redirectUrl = async (speech: Speech, outputFormat: string) => {
       return generateDownloadSignedUrl(speech.audioFileName)
   }
 }
-
-const ssmlItemsToUtterances = (items: SSMLItem[]): [Utterance[], number] => {
-  const tokenizer = new WordPunctTokenizer()
-  let wordOffset = 0
-  return [
-    items.map((item) => {
-      const text = htmlToText(item.textItems.join(''), { wordwrap: false })
-      const wordCount = tokenizer.tokenize(text).length
-      const utterance: Utterance = {
-        wordOffset,
-        wordCount,
-        text,
-        voice: item.voice,
-        idx: item.idx,
-      }
-      wordOffset += wordCount
-      return utterance
-    }),
-    wordOffset,
-  ]
-}
diff --git a/packages/api/src/textToSpeech.d.ts b/packages/api/src/textToSpeech.d.ts
index ab955ae1b..051b9be8f 100644
--- a/packages/api/src/textToSpeech.d.ts
+++ b/packages/api/src/textToSpeech.d.ts
@@ -1,21 +1,29 @@
 declare module '@omnivore/text-to-speech-handler' {
-  export function htmlToSsmlItems(
+  export function htmlToSpeechFile(
     html: string,
     options: SSMLOptions
-  ): SSMLItem[]
+  ): SpeechFile
 
   export interface SSMLOptions {
-    primaryVoice: string
-    secondaryVoice: string
-    rate: string
-    language: string
+    primaryVoice?: string
+    secondaryVoice?: string
+    rate?: number
+    language?: string
   }
 
-  export interface SSMLItem {
-    open: string
-    close: string
-    textItems: string[]
+  interface Utterance {
     idx: number
+    wordOffset: number
+    wordCount: number
     voice?: string
+    text: string
+  }
+
+  export interface SpeechFile {
+    wordCount: number
+    averageWPM: number
+    language: string
+    defaultVoice: string
+    utterances: Utterance[]
   }
 }
diff --git a/packages/text-to-speech/src/index.ts b/packages/text-to-speech/src/index.ts
index a9b42a308..5ca4d54f7 100644
--- a/packages/text-to-speech/src/index.ts
+++ b/packages/text-to-speech/src/index.ts
@@ -9,11 +9,7 @@ import * as jwt from 'jsonwebtoken'
 import * as dotenv from 'dotenv' // see https://github.com/motdotla/dotenv#how-do-i-use-dotenv-with-import
 import { synthesizeTextToSpeech, TextToSpeechInput } from './textToSpeech'
 import { File, Storage } from '@google-cloud/storage'
-import { htmlToSsmlItems } from './htmlToSsml'
-
-interface SSMLInput {
-  text: string
-}
+import { htmlToSpeechFile } from './htmlToSsml'
 
 interface UtteranceInput {
   voice?: string
@@ -176,9 +172,9 @@ export const textToSpeechStreamingHandler = Sentry.GCPFunction.wrapHttpFunction(
         return res.status(500).send({ errorCode: 'SYNTHESIZER_ERROR' })
       }
       res.send({
+        idx: utteranceInput.idx,
         audioData: audioData.toString('hex'),
         speechMarks,
-        idx: utteranceInput.idx,
       })
     } catch (e) {
       console.error('Text to speech streaming error', e)
@@ -188,7 +184,7 @@ export const textToSpeechStreamingHandler = Sentry.GCPFunction.wrapHttpFunction(
 )
 
 module.exports = {
-  htmlToSsmlItems,
+  htmlToSpeechFile,
   textToSpeechStreamingHandler,
   textToSpeechHandler,
 }