Add skip or change word
-
@river Hello,
I would like to suggest a feature that could improve the realism and expressiveness of TTS-based narration in this application.
Currently, most text-to-speech systems read dialogue exactly as written, relying mainly on punctuation for tone. This often results in speech that sounds flat, especially when reading novels or dialogue-heavy content where the writer intends specific emotions such as anger, sadness, affection, or sarcasm.
To improve this, I suggest introducing a simple annotation-based dialogue instruction system that allows writers or users to specify the intended emotion or delivery style directly in the text. The TTS engine could parse these annotations and adjust voice parameters such as pitch, speed, pauses, and emphasis accordingly.
Proposed Method
The idea is to use lightweight tags placed before dialogue lines to indicate emotional intent.Basic Format:
[emotion:intensity|style] Dialogue textWhere:
- emotion = emotional tone of the line
- intensity = level from 1–5
- style (optional) = delivery modifier such as slow, whisper, pause, etc.
Examples:
- Angry dialogue
[anger:3] Get out of my sight!
Meaning: moderate anger with stronger emphasis.
- Soft or calm speech
[soft:1] It's alright. Everything will be fine.
Meaning: gentle tone with relaxed delivery.
- Romantic or affectionate dialogue
[love:2] I missed you.
Meaning: warm and soft voice tone.
- Fearful dialogue with trembling
[fear:4|tremble] Did you hear that?
Meaning: high fear intensity with a shaking voice effect.
- Sad dialogue with slower pacing
[sad:3|slow] I thought you would stay.
Meaning: medium sadness with slower speech.
Optional Pause Markers
Additional inline markers could control pacing inside a sentence.{p} = short pause
{pp} = medium pause
{ppp} = long pauseExample:
[shock:3] What… {p} what happened here?How It Could Work Internally
-
The TTS reader scans text for annotation tags.
-
Tags are parsed and removed from the spoken text.
-
Each tag is mapped to voice parameters such as:
- pitch adjustment
- speaking rate
- pause duration
- emphasis
-
The processed dialogue is then passed to the TTS engine with these parameters applied.
Benefits
- More natural audiobook narration
- Better dialogue acting in novels
- Greater emotional realism
- Minimal complexity for users
This method would allow AI voices to deliver dialogue more like a human narrator or actor rather than simply reading text line by line.
Thank you for considering this suggestion.
-
Thank you for all the feedback. This is super valuable.
The only feature missing is an auto-scroll option.
Yeah, it's in my plans for the next version.
To improve this, I suggest introducing a simple annotation-based dialogue instruction system that ...
I’m not sure. This system seems to require calibration on the website itself, it can't be done via the PNL Reader alone, right? Do you know of any websites already using it?
It looks very interesting and well-organized. If this system is being used on any website, I’d definitely like to integrate it. It would be a significant improvement.
-
@river Here is it which is close to what I suggest.
https://typecast.ai/text-to-speech/ -
@wkhn66
Hi, good news to share with you. I have released PNL Reader v2.8.0, which has the fix for:- Smarter pauses for decorative dividers like ~~~ or *** that authors use between sections.
- Auto scroll mode for when you use "read whole page".
- Better content extraction for your favorite novel site novelfull.com, have you noticed?
For the annotation-based dialogue instruction system, I need to dig into this deeper. But at least I took inspiration from the AI tool you recommended, and upgraded the prompt for the AI characters in PNL Reader to have smarter emotion detections, give it a try, see if it has improved a bit
There is much more update with v2.8.0, check it out here:
https://pnl.dev/topic/1078/pnl-reader-v2-8-0-customize-the-reader-for-your-favorite-sitesThank you again for your valuable feedback!
-
@river thanks for the update! I’m really glad to see the smarter pauses and auto-scroll features. The improved support for websites is great too, really appreciate you taking the feedback into account.
-
@river You should also make mobile app for this reader, Mobile apps has much bigger scope compare to computers, I am currently using this extension on Mobile using Quetta browser which supports extensions as well, But it will be great if you can develop it for mobile devices as well.
Thanks in advance. -
@wkhn66
Thank you, glad to hear that the new features work for you. If you spot any issues, don't hesitate to tell me.
Yeah, mobile app is in my roadmap, though probably I will start with Dictionariez, which I already have got a good picture in my mind. For PNL Reader, I am not really sure what to do with an app, as it's simply just a prettifier running on other websites, right? So probably it still needs to be a browser?Anyway, thanks for the suggestions. Would you mind to leave PNL Reader a review on the webstore? It can really help. Or star it on github if you have an account...
-
@river You're welcome! The new features are working great so far. I’ve also left a positive review for PNL Reader on the Chrome Web Store to support the project. Thanks for your hard work on the extension, really appreciate it.
-
just jumping in to say I love the new v2.8.0 updates too! the smarter pauses for dividers (***) make a huge difference for immersion. keep up the great work!
-
Awesome. I would never know it has that much of impact without your feedback. So thank you.
Which website and novel are you reading by the way?Also, I am wondering, do you like the themes of PNL Reader's selection? Which one do you like the best? For me it's the
Nordic Darktheme I added in the last version.Which TTS Character do you like to use (Luna, Owen, Mona, Bob, Kaylee or Vincent)?
Thanks again. These are valuable feedback to me, as for I can keep improving the product, start making some good impact to all learners in the world
Support me!