Screen Recording of My 9 step prompt in action using ChatGPT - now for mobile phones and with subtitles

I have created a new version of my shortest video outlining the nine steps I have recently recommended for learners to get feedback on their speaking. This version is optimized for mobile phones and includes subtitles made using ClipChamp, which were very good and only needed me to add in the punctuation. 

Guiding students' attempts to self-correct their spoken errors based on Jasmin Silver's ChatGPT prompt

Incorporating Jasmin Silver's ChatGPT prompt to guide students' attempts to self-correct their written errors into doing the same with spoken errors

Last week I watched the webinar by Macmillan ELT about Assessment and although Jasmin Silver wasn't talking about getting feedback on speaking, I thought one of her ideas could be adapted for students to use with transcripts of their speaking.

Her part of the webinar is here:

Here is her original prompt:

And, below is how I adapted it so students could use it with a recording made on their mobile phones using Turboscribe.ai

After each of the steps 1 to 5 below, wait for me to say "next"
Use language at the level of my transcript
Step 1
You are a teacher who facilitates self-assessment for learning. This involves self-assessment skills such as self-rating, self-grading, self-feedback, self-evaluation, use of checklists, and rubrics so that you can help me learn better and improve my work. When I ask for answers or examples, guide me in the form of hints, clues, or suggestions one step at a time. Try not to give me a direct answer. Remember, your role is to facilitate self-assessment for learning, not merely to give answers directly to me. When we’ve finished working on the transcript, write the final version and offer step 2
Step 2
Write an improved transcript half a level up on the CEFR scale. Use the format: (old => new)
Step 3 
Write the changed version
Step 4 
Enhance the transcript another half level up. Format as above
Step 5 
Write the changed version

After students have recorded themselves using Turboscribe.ai this prompt should be copied and pasted into the ChatGPT, Custom Prompt option.

Once the ChatGPT prompt has been generated it can be copied and pasted into the ChatGPT app. There is an example of how this might work below.

Here's a link to a chat with ChatGPT of one of my experiments with the prompt above . It is a simulation as I was playing the part of the student and it would be much more interesting to be able to share a chat with a real student responding to the hints and clues to help the them self-correct their transcript.

I have already changed this prompt once and may well do so again, in which case a link to the latest version might be useful

Big asks and uphill tasks - making a case for TBLT by Neil McMillan

I’ve just watched the plenary  that Neil McMillan gave in the morning of the final day of #IATEFL2025. I’d attended live Catherine Walter’s plenary session on what has changed in 50 years as that is very much the same timeframe that I have lived through. I’ll watch the other plenaries later on.

Task Based Language Teaching has always been a hard-sell even among people who see the logic of it, maybe because  it isn’t as easy as that to actually do for the many reasons Neill outlined. One of the reasons is that it’s much easier to use English File, which is so packed with exercises and other material that the teacher will never run out of material instead of choosing a task-based curriculum and persuading  students, parents, colleagues and bosses that you haven’t lost your mind as well as your way!

Research is very much a university thing so it is difficult to find research papers about the use of tasks in language schools and even harder to find them used in secondary or primary schools. Lack of research though isn’t the only obstacle: a much bigger one is the fact that there are very few commercially produced materials that are based on TBLT and equally few databases of tasks for teachers to draw on. There are two, provided by Neill: https://www.iatblt.org/resources

Just as I was noting, “Please can we have more examples of tasks that could form part of a TBLT course or that could be a task based on a standard coursebook. We are at the half way point in the talk!” Neill provided one that would not have showed up on anyone’s needs analysis! How to solve ‘the problem of alligators in the Underground of New York’

The task could be done in small groups with groups comparing solutions and learning from each other, but some sort of monologue outlining the planned solution(s) seems to me inevitable as the final part of the task. This could and should be recorded.

Neill included in the list of possible Post-task options, ‘Discuss and Correct errors collected by the teacher’ and ‘Transcribe and analyse transcriptions: what could I have said better?’

We all try to relate what we hear in presentations to our own situations and our own ideas, so I jumped on these two post-task activities as being exactly what my 9-step prompt for ChatGPT aims to do: to work on errors (Steps 1-3) and suggest better ways to say the same thing (Steps 4-9). Steps 3, 5, 7 & 9 are all there to allow students to listen to the improved version, so there are really only 5 steps:

• Identify errors
• Correct errors
• Suggest improvements at the student's current level
• Suggest improvements at a slightly higher level
• Suggest improvements at an even higher level

     Here are some YouTube videos about my 9-step prompt from the shortest to my full presentation as at #IATEFL2025:

Screen Recording of My 9-step prompt in action using Turboscribe.ai ChatGPT (7 minutes)

How students can use Turboscribe and ChatGPT to get feedback on a recording they have made (17 minutes)

"Your students can get feedback every time they speak" Rehearsal for EUROCALL JALTCALL webinar (43 minutes)

"Your students can get feedback every time they speak" at #IATEFL2025

I was very happy with the way my session "Your students can get feedback every time they speak" went at #IATEFL2025.

hashtag#IATEFL2025 went this morning. There were 21 other talks to go to at exactly the same time, yet 30 people chose to come to mine.

I started off with a poll using vevox.com, which 19 people completed. The result was that I could see that half had already got the two free accounts needed to be able to really participate in the workshop. As a result I felt we could really get to work on the workshop, which involved participants recording themselves speaking a language they knew they would make mistakes in and going through the 9 steps as if they were students

One of the joys of giving a live in-person presentation was that I could go round trying to help people when they had problems. However, this took a lot of time.

I wasn't able to do everything I had planned but the feedback was very positive. The most important being that the 17 people who completed the survey gave 46/50 to "I want to try this with my students."

When one of the delegates came up a the end and said it was the best presentation they had been to at IATEFL, it obviously made my day!

The video below is a recording of a rehearsal for a Zoom webinar I gave last week with hashtag#EUROCALL & hashtag#JALTCALL. The demonstration of the 9 steps in my ChatGPT prompt is exactly as it was in the live webinar and the live presentation.

I've suggested places to pause the video for anyone wanting to 'participate' in the workshop.

Here is the 9-step prompt I asked participants to download or copy from the screen using the cameras on their mobile phones and the text extraction that most mobiles offer:

9-step Custom Prompt - Version without revealing levels

Instructions:

- Remove timestamps

- Decide what level the transcript is on the GSE and CEFR scales and half scales using + but don’t reveal any levels

- After each of the steps 1 to 9 below, wait for me to say "next"

- Write in continuous prose for all steps.

- Answer any questions I ask at a level suitable for this level

Steps:

1 Identify the errors in bold and enclose them in brackets. Do not correct them. Just show where they occur

2 Correct errors, showing corrections in bold while keeping original errors in brackets next to them. Use the format: (error => correction) 

3 Display the correct version without mistakes

4 Write an improved version at the same level on the CEFR scale above. Use the format: (old => new)

5 Write the changed version

6 Write an improved transcript one level up on the abovementioned CEFR scale Format as above

7 Write the changed version

8 Enhance the transcript to two levels up. Format as above

9 Write the changed version

And, finally, here is a link to the Google doc where I will be updating the prompt as I experiment with ways to improve it

 

My presentation "Your students can get feedback every time they speak" for EUROCALL and JALTCALL

I thoroughly enjoyed presenting my ideas about how to help students get more feedback on their speaking with Turboscribe.ai (speech-to-text) and ChatGPT.

Only 40% of the people attending already had (free) accounts for both Turboscribe.ai and ChatGPT, but that was enough to give it as a workshop where participants experienced the 9-step prompt that I shared with them as if they were students.

Here is a link to the prompt I shared with them

or a QR code to access it.

I made two videos in English to demonstrate the method I was suggesting, but in fact it works in 90+ languages, of which I have only tried 4.

This one is about 7 minutes long and uses the same examples that I used in the presentation:

This one is about 17 minutes long and is a screen recording made in real time showing the 9 steps in the prompt I suggested:

At the end, I asked participants to answer a few questions about these ideas and this is what 11 of them said:

From this it seems that the only problem was that it wasn't really easy to connect Turboscribe with ChatGPT. I think that with a little practice  teachers and students will find it easier and easier.

These are a few of the comments in the chat:

• “I definitely would like to try this with my students - but I think it is also good to use for improving my own L2 Japanese!”
• “Thanks Chris! I really enjoyed your workshop, it was an introduction to TurboScribe for me, and your super prompt is a great idea.”
• “Thanks Chris - that was great :-)”
• “I’ve used TurboScribe before but never imagined using it in this way. Amazing idea”
• “Thank you very much Chris. Want to play with it now”
• “Ah, this is very useful for self-study and feedback generation.”
• “Ohhhhh I love this! I did it in Italian and it felt so good seeing the smooth version of what I had said. I listened to myself and noticed all the pauses and repetitions but the story itself - that story is mine. It felt good and I think I’ll use this more for my own L2 practice and also introduce it to my students.”

The person most interested in and in need of feedback on speaking is the student. The teacher’s interest is more periferal.

The person most interested in and in need of feedback on speaking is the student. The teacher’s interest is more periferal. It is important that the feedback should be as immediate as possible and should include an opportunity to self correct errors identified by the tool before seeing the corrections. But there needs to be a further kind of feedback suggesting ways the student could say it better at their own level or at future levels the student might aspire to.

By using a combination of a transcription tool like Turboscribe.ai and a long Custom Prompt for ChatGPT, students can get all this. It isn’t perfect but I believe it is ‘good enough’

Here is an example of such feedback and a long conversation with ChatGPT about its inaccuracies in grading:

https://chatgpt.com/share/67d15f5b-c450-800b-a548-b5b9d1d5c917

I was commenting on an article in FLT Mag https://fltmag.com/language-assessment-ai/ 

Does 'freespeaking' exist as a concept? Freewriting certainly does!

I read a post by Nick James on LinkedIn about freewriting which cited this paper:

Fluency development through freewriting and transfer to other more structured tasks - Raymond Yasuda, 2025

It got me thinking about whether 'freespeaking' might exist as a concept and whether anyone had done research on the benefits to speaking fluency of regular 'freespeaking' analogous to the improvements in writing fluency after regular freewriting observed in the paper.

This is the Customise prompt I used on Google's NotebookLM with the article above as the only source

"I'm interested in how the principles of this research into freewriting could be applied to speaking? Is there such a thing as freespeaking to help deliver fluency in speaking? If students had to do "freespeaking" for a fixed length of time using a transcription tool like Turboscribe.ai, would they be able to measure increases in speaking fluency over an extended period like three months?"

I asked for an audio overview and I then uploaded the audio file of the audio overview to a new NotebookLM.

This is the summary NotebookLM made of the Deep Dive Conversation it had produced itself an hour or two earlier

"This audio recording investigates the concept of "free speaking" as a method to enhance spoken fluency, drawing parallels with the well-established practice of freewriting. It examines research by Yasuda on freewriting's positive impact on writing fluency in Japanese university students, noting significant gains within the initial weeks. The discussion explores practical ways to incorporate free speaking into daily life, such as talking to oneself or using prompts, and considers tools like transcription services for tracking progress. Ultimately, the speakers posit that consistent free speaking can build "speaking muscles," improve communication effectiveness, foster creativity, and contribute to overall language proficiency alongside accuracy-focused practices."

The Podcast or Audio overview is 18 minutes long and may not be interesting enough to spend that amount of time over, but here is a link to it:

https://notebooklm.google.com/notebook/950162df-1264-4339-b4e6-11e651d98a97/audio

or,

Sharing the transcript on Rev:

https://www.rev.com/app/transcript/NjdjYTI5Y2E1NzY4YWU2YWQ0NTVmZmU4UW1CNm5IQkxtalh0/o/VEMwMTQ2NDY2MjUw

You might prefer to read these summaries.

Summary from Rev

"This transcript explores the concept of "free speaking" and its potential benefits for improving speaking fluency, based on research on the impact of "free writing" on writing fluency. The key points are:

- "Free speaking" is not an official term, but the idea is similar to free writing - allowing for unconstrained output without worrying about perfection. [14:42]

- Research has shown that free writing can significantly improve writing fluency in a relatively short time, suggesting that free speaking could have a similar effect on speaking fluency. [00:04]

- Free speaking can help automate the speaking process, freeing up cognitive resources to focus on other aspects of communication like tone, pacing, and nonverbal cues. [01:27]

- Incorporating free speaking into daily routines through techniques like speaking prompts, recording oneself, and practicing with language partners can help develop fluency.

- Free speaking may also unlock creative potential by removing barriers between thoughts and words, allowing for more spontaneous and engaging communication. [14:56]"

Reply from ChatGPT

I asked ChatGPT, "Does the concept of 'freespeaking' exist analogous to freewriting which certainly does, both in the context of writers' block and how it helps learners of English become more fluent writers?"

This was the reply:

"Reasoned about freespeaking and freewriting for 28 seconds >

Yes, the idea of freespeaking is similar to freewriting. Both methods encourage learners to produce language continuously without worrying too much about mistakes.

How They Are Similar

• Spontaneity: In both freespeaking and freewriting, you speak or write without pausing to correct every error.
• Overcoming Blocks: They help break through anxiety or a mental block—whether it is a writing block or a speaking block.
• Fluency Building: Regular practice can make your language fluid and more natural over time.

Freespeaking in Language Learning

• Practical Use: Many language teachers use freespeaking exercises to help learners practise speaking spontaneously.
• Informal Term: Although the term is not as widely researched as freewriting, it is used to describe activities where learners speak for a set period without self-editing.

Overall, freespeaking is a useful technique for those who want to improve their spoken English, much like freewriting supports fluent writing."

Conclusion after consulting Perplexity

"While there isn't a direct equivalent of "freespeaking" in research, the principles of freewriting can be applied to speaking to improve fluency. Freewriting, as described by Peter Elbow, involves writing without planning or monitoring, allowing words to flow spontaneously1. This concept can be adapted to speaking to enhance fluency and spontaneity in oral communication."

Attempts to extract text from a handwritten composition

I wanted to find a good way for students to get feedback on their hand-written compositions done in class as soon as possible after they handed them in. My initial idea was for students to take a photograph of the written work and then to upload it to a GenAI tool. It turned out that ChatGPT, Gemini and Perplexity could do this, but Claude, Copilot, Deepseek, Mistral and Llama said they couldn’t and asked me to provide a transcript.

The second problem was that the three tools that said they could extract text from the photo of a handwritten composition made on average 17 mistakes in 156 words! However, in most cases the 6 steps of feedback would still be useful despite the uncorrected misinterpretations of the handwritten text. I was a bit disappointed as I had had great hopes for the idea.

Then I had a second idea, which was to get students to record themselves reading their composition aloud using Turboscribe.ai which I would be asking them to use anyway to record their speaking as well. The resulting transcript only had one error in the 156 words, which was easy to correct by editing the transcript. From then on, the procedure was almost identical to the one I recommend with their speaking. It involved copying and pasting an adapted long prompt from Google drive into the Custom Prompt option of the ChatGPT menu in Turboscribe.

Here’s a link to the ChatGPT conversation: https://chatgpt.com/share/67bb85fd-702c-800b-94c8-bc032b3a6f92

The grades assigned by ChatGPT and Pearson’s Text Analyzer to the five different versions were similar, but only coincided exactly with two grades:

	ChatGPT	Pearson	ChatGPT	Pearson
Original	46.5	44	B1	B1
Correct errors	49.5	44	B1+	B1
Better same level	49.5	46	B1+	B1
Better next level	57.5	60	B2	B2
Better 2 levels up	61.5	63	B2+	B2

 There is still an obvious progression through the different versions:

Original B1 44	Corrected B1 44	Better B1 46	1 level up B2 60	2 levels up B2 63
The most important event that takes in the book I have read, Thirty-Nine Steps, is that the protagonist of the book was nobody in the first part of the story, and a strange meeting with a man who he didn't know changed all his LIVE. The strange thing is that the main character was in London for pleasure, and he was in a little apartment, and a very strange man came into the apartment and lived with the protagonist for a couple of weeks, and one day he found the strange man dead in the floor of the apartment with a knife in his heart. I think that the most important event in the entire book, because the protagonist could say no to the strange man and all his story of secret spies in Europe before the First World War, and the main character followed the steps of the spy. I like this book very much.	The most important event that takes place in the book I have read, Thirty-Nine Steps, is that the protagonist of the book was nobody in the first part of the story, and a strange meeting with a man who he didn't know changed all his life. The strange thing is that the main character was in London for pleasure, and he was in a little apartment, and a very strange man came into the apartment and stayed with the protagonist for a couple of weeks, and one day he found the strange man dead on the floor of the apartment with a knife in his heart. I think that this is the most important event in the entire book, because the protagonist could say no to the strange man and all his story of secret spies in Europe before the First World War, and the main character followed the steps of the spy. I like this book very much.	The most important event that takes place in the book I have read, The Thirty-Nine Steps, is when the protagonist starts as an ordinary person, but a strange meeting with a man he didn’t know changes his whole life. The strange thing is that the main character was in London for pleasure, staying in a small apartment, when a very unusual man came to his apartment and stayed with him for a couple of weeks. One day, he found this man dead on the floor of the apartment with a knife in his heart. I think this is the most important event in the entire book because the protagonist could have refused to listen to the strange man and his story of secret spies in Europe before the First World War, but instead, he decided to follow the spy’s steps. I like this book very much.	The most important event in the book I have read, The Thirty-Nine Steps, is when the protagonist, who seems like an ordinary person at the start of the story, experiences a life-changing event after meeting a stranger. The main character is in London for leisure, staying in a small apartment, when a very mysterious man arrives unexpectedly and stays with him for a couple of weeks. One day, the protagonist finds the man dead on the floor of the apartment with a knife in his heart. This shocking discovery leads him to realise that the stranger’s story about secret spies in Europe before the First World War is true. He could have ignored the man’s tale, but instead, he chooses to investigate, which leads him on a dangerous and thrilling adventure. I really enjoyed reading this book.	The most important event in the book I have read, The Thirty-Nine Steps, is when the protagonist, who begins the story as an ordinary man living a quiet life, is suddenly thrown into a world of danger and intrigue. While staying in London for leisure, he meets a mysterious stranger who arrives unexpectedly at his small apartment. The man is clearly anxious and claims to know about a secret plot involving spies in Europe just before the outbreak of the First World War. The protagonist reluctantly lets him stay, but his life takes a dramatic turn when he discovers the man dead on the floor of his apartment with a knife in his heart. This shocking event forces him to confront the possibility that the man’s story was true. Fearing for his own safety and driven by curiosity, he decides to investigate the conspiracy himself. His decision sets off a thrilling chain of events as he follows the trail of the spies, facing numerous dangers along the way. This unexpected twist makes the story truly exciting, and I thoroughly enjoyed reading this book.
156 words	159 words	146 words	137 words	183 words

ChatGPT interviewing me about the Spanish Couple on the Titanic

Daniel Martin suggested on Facebook that with the right prompt ChatGPT could interveiw anyone using voice. From reading to speaking practice I adapted his prompt to make it easier for students to use it with any text they wanted.

Here is the text that I used for this experiment:
https://www.bulldogz.org/post/titanic-tales-a-single-spanish-victim

Here is the ChatGPT conversation, which is only a text chat:
https://chatgpt.com/share/67b0f234-9a28-800b-9000-c3f06260a3ff

And here is the recording I made of the interview:
https://www.rev.com/app/transcript/NjdhZmEyZDg1YWFjZGMzOTA2YjQ1ZWY1N3dLMk9scE11b2NY/o/VEMwMTk3MzQ5MDA2

You can watch a video about it here:

I would love to see a similar video made by a student.

Turboscribe versus Rev - the two best and most generous transcription tools

I asked NotebookLM to do the heavy lifting of comparing three transcripts saved from Turboscribe (x1) and Rev (x2)

This is an edited version of how it replied to my question:

“I am interested in any differences between the three transcripts particularly regarding the use of interjections and the use of phrases like 'like', 'you know', 'um' etc”

Differences appear in the transcripts regarding interjections and phrases such as "like," "you know," and "um." The transcript generated by TurboScribe.ai, includes these kinds of interjections and phrases more frequently compared to the two transcripts generated by Rev.com

For example, the TurboScribe transcript includes the parts in bold that are missing from the Rev transcripts:

·         "[M] Right. It's like we're on the edge of this revolution, you know, and how we learn and practise speaking. Like it's a total paradigm shift"

·         "[M] Exactly. And they went step by step, you know, first analyse the speech, then identify errors"

·         "[M] Exactly. So you got to clean up the transcript a bit. And then that's when the fun starts. Use ChatGPT to actually like analyse your speech and give you feedback"

·         "[M] They encourage teachers to, like, play with it, change it, adapt it, you know, and even share their own versions with other teachers"

These examples show the more conversational style captured by TurboScribe, including interjections ("wow") and fillers ("like", "you know"). The two transcripts captured by Rev appear to have had some of these conversational elements removed or edited out, presenting a cleaner version of the dialogue.

On the other hand, below is a comparison of how Rev.com with the verbatim option selected and Turboscribe, which has no such option, handles transcribing. It is very revealing:

Rev.com:

Rev heard ‘boat’ and ‘things’ and as it had been asked to create a verbatim transcript it didn’t examine how likely this sequence of words was.

Turboscribe.ai:

Turboscribe, on the other hand, correctly interpreted the learner’s intention because ‘bought’ and ‘thinks’ are more likely in the context. But this desire to predict what is more likely led it to also correct ‘womans’ to ‘women’, ‘say’ to ‘said’ and ‘are’ to ‘was’ and to insert ‘can’ into the phrase ‘you can guess’

Both Turboscribe and Rev interpreted what I heard as ‘ask her’ as ‘asked her’, but I imagine that is because the /t/ is hard to produce and often omitted by native speakers. Why ‘that’ was omitted by Turboscribe in the phrase ‘said no’ must also be down to frequency. About 161,000,000 results without ‘that’ v. About 26,900,000 results with ‘that’ according to Bing. 

There are other important differences between free accounts using Turboscribe.ai and Rev.com:

The most important advantage of Turboscribe is that it can be used from the age of 13 with parents’ permission. The other factor to bear in mind is that Rev.com only works in English on a free account at the moment. The transcripts are automatically synchronised while listening on Rev.com, whereas with Turboscribe.ai the user needs to move the page up to follow them.

But the key reason for not being able to recommend Rev.com on a mobile device is that it is: 

No longer possible to edit or download transcripts using a mobile device

When using a PC, this is not a problem.

Support on Rev.com is excellent and I haven’t needed to contact support on Turboscribe.ai