Key Takeaways
- EssayHero now has a
/speakingsection for HKDSE Paper 4 and IELTS Speaking practice - Record yourself, see your transcript, then get AI feedback on vocabulary, grammar, ideas, and fluency
- We do not assess pronunciation — and we're upfront about why
- 66 prompts, cue cards, fluency metrics, and a read-aloud checker
What's New
EssayHero started as a writing feedback tool. But students preparing for HKDSE or IELTS don't just write — they speak.
Paper 4 and the IELTS Speaking test are worth real marks, and most students get very little structured practice outside the classroom.
Speaking Practice for Two Major Exams
So we built a speaking section. It covers:
- HKDSE Paper 4 Part B — Individual Response
- IELTS Speaking Parts 1, 2, and 3 — Full test format
You can find it at /speaking.
How It Works
The flow has two phases, and the separation is deliberate.
Phase 1: See Your Words
First, you record yourself responding to a prompt. When you stop, you get a transcript — your actual words, laid out on screen.
This step matters on its own. Most students have never seen what they sound like written down. It's often surprising.
You notice the fillers, the half-finished thoughts, the sentences that go nowhere. That awareness is valuable before any AI gets involved.
Phase 2: AI Analysis
Second, the AI analyses the transcript. It gives you feedback on:
- Vocabulary range — variety and sophistication of word choice
- Grammatical accuracy — correctness and complexity of structures
- Idea development — depth and coherence of your response
- Fluency — smoothness and natural flow of speech
Each area gets a separate assessment, so you can see specifically where to focus.
Why This Order Matters
We show you your words first, then we analyse them. You get to form your own impression before the AI offers its view.
What We Don't Assess (and Why)
Here's the honest part. We cannot reliably assess pronunciation, intonation, stress patterns, or body language from a transcript.
And we don't pretend to.
How Speech-to-Text Works
Speech-to-text systems convert your audio into words. Once that conversion happens, the acoustic information — how you said something — is gone.
What remains is what you said.
What Pronunciation Assessment Requires
Assessing pronunciation properly requires either:
- A human listener, or
- Specialised acoustic models built specifically for that purpose
A language model reading a transcript is not that.
Our Position on Pronunciation Scoring
Some tools on the market will give you a pronunciation score anyway. We think that's irresponsible.
An unreliable pronunciation score is worse than no pronunciation score, because students might trust it and stop working on an area that actually needs attention.
A Design Choice, Not a Missing Feature
So we score what we can assess reliably (vocabulary, grammar, ideas, fluency patterns) and we're transparent about the rest.
This is a design choice, not a missing feature.
Fluency You Can Measure
Even without pronunciation scoring, there's a lot of useful data in how you speak.
Quantitative Fluency Metrics
The speaking section gives you:
- Words per minute — overall speaking rate
- Pause frequency — how often you stop mid-flow
- Average pause length — duration of hesitations
- Filler word count — "um", "like", "you know", etc.
Real Measurements, Not AI Judgement
These numbers are derived directly from your recording. They don't involve AI judgement — they're measurements.
If you're speaking at 90 words per minute with frequent pauses and a lot of "um" and "like", that tells you something concrete to work on.
Track Your Progress
You can record yourself again next week and see whether the numbers improve.
66 Prompts, No Repeats
We've built a bank of 66 speaking prompts across all supported formats:
- HKDSE Individual Response prompts
- IELTS Part 1 topics
- IELTS Part 2 cue cards (with the visual layout you'll see in the real exam)
- IELTS Part 3 discussion questions
Systematic Practice Without Repetition
The prompts are served without repeats, so you can work through them systematically.
If you're doing a prompt a day in the weeks before your exam, that's over two months of practice.
Read-Aloud and Voice Dictation
Two smaller features worth mentioning.
Read-Aloud Fluency Checker
The read-aloud fluency checker lets you paste in a text, read it aloud, and compare your spoken version against the original.
It highlights where you diverged — skipped words, substitutions, additions.
This is useful for:
- Building reading fluency
- Catching patterns in how you process written English aloud
- Identifying words that trip you up
Voice Dictation on Essay Form
Voice dictation is now available on the main essay form. You can brainstorm by speaking instead of typing.
Some students find it easier to get ideas flowing out loud, then edit the transcript into a proper essay. It's there if you want it.
Same Philosophy, New Format
If you've used EssayHero for writing, the approach will feel familiar.
Our Approach to Feedback
We give you the most useful feedback we can, we tell you what we can't do, and we let you decide how to use the information.
- No inflated scores — honest assessment based on official criteria
- No vague encouragement — specific, actionable feedback
- Just honest assessment — of the things we can measure well
Try It Now
Speaking practice is live now at /speaking. Try it and let us know what you think.
Feedback or questions? Email hello@essayhero.app.
Related Articles
Mark an Entire Class in Minutes: EssayHero's Teacher Marking Suite
Upload one PDF, get every essay marked. Score overrides, custom rubrics, class analytics, and full export — built for teachers.
Read moreSpeaking Practice When Nobody's Listening
Practical tips for practising HKDSE Paper 4 and IELTS Speaking on your own, with honest advice on what works.
Read moreWhat Your Vocabulary Score Actually Means
A plain-language guide to how EssayHero measures vocabulary diversity, sophistication, and readability — and what the numbers mean for your writing.
Read more