VOICEPRINT
Where Am I in This Flow?
Today, a text can pass through another language, an AI model, editing, adaptation — and return almost flawless.
Grammatically correct. Natural. Smooth. Sometimes even more literary than the original.
And this is exactly where the problem begins.
A text can become better after this kind of processing. Sometimes, that is precisely the loss.
Because at some point the question is no longer:
“Is this a good translation?”
It becomes:
“Am I still in it?”
How It Began
Continue the story ↓
In the country where I lived at the time, I could not freely publish much of what I wrote about — or the way I looked at the human being, society, life and reality.
So the most obvious path — write a book and publish it there — was effectively closed to me.
There was another path: to move beyond a single language.
But that brought a different problem.
I did not know foreign languages well enough to carry a complex literary and philosophical text into them myself.
And professional translation of large works cost so much that, for me, it was not a realistic option.
So texts could exist for years — in notebooks, files and drafts.
Not because I did not want to bring them to readers.
I simply could not see a realistic way to do it.
With the arrival of modern AI systems, the situation began to change for the first time.
It became possible to work with large volumes of text, bring scattered materials together, maintain structure and terminology, return to decisions made dozens or hundreds of pages earlier, and translate and transcreate works into other languages.
What had once required fluency in other languages, a large team and money I did not have suddenly became practically achievable.
At first, it felt almost like liberation.
I began showing translated texts to people who knew the target language well and asking how natural they sounded, whether they felt machine-made, whether there were errors or strange passages.
And gradually it became clear:
AI was already capable of creating text that a native speaker could judge to be good, natural and literary.
You might think the problem was solved.
But that was when another one appeared.
Who Is Speaking?
Someone reading the English text can tell me:
“Yes, this is good English.”
But do they know what, exactly, was important to preserve in the Russian original?
Why did I write this sentence this way?
Why did I repeat the same word several times?
Why did I leave the sentence heavy?
Why does the character speak harshly?
Why was something deliberately left unsaid?
Why did I choose one metaphor when another could have been made more beautiful?
For someone who sees only the final translation, none of this may exist at all.
But there was an even stranger problem.
I did not know the target language well enough to verify all of this myself.
In other words, I needed AI to carry my text into another language.
And then I needed AI again — this time to help me understand what had actually happened to my text during that transfer.
I asked it to compare versions.
Explain differences.
Return to the original.
Point out possible losses of meaning.
Check terminology.
Recall earlier decisions.
Compare scenes written far apart from one another.
It became a rather strange process:
I was using AI to use AI to check the work of AI.
But I simply had no other way to see what was happening.
And gradually I understood: I did not need a “better translator.”
I needed a process in which I myself could understand what was happening to my book.
ATP Came First
That is how Ashraellen Transcreation Protocol — ATP appeared.
It was the first attempt to organise the work into a system I could understand.
We compared different ways of carrying a text across: ordinary translation, freer AI transcreation and work through ATP.
We tried separate AI chats.
Compared the results.
Checked how well meaning, terminology, structure, voice and previously accepted decisions were preserved.
ATP gradually stopped being merely a set of instructions.
It acquired its own public repository, documentation, versions and DOI.
But working with it led to the next problem.
A term can be checked.
Structure can be compared.
A missing meaning can be found as well.
But how do you check the author?
Where Am I?
This is where the question stopped being only about translation.
Suppose AI has not lost the facts.
Has not confused the characters.
Has conveyed the content correctly.
The text reads well.
A native speaker says it sounds natural.
But is that enough?
At what point does a translation stop being my book and become a very good version of how AI understood my book?
And can I even see that boundary if I do not know the language of the result well enough myself?
This is where Voiceprint came from for me.
Not from a desire to create a new technology.
From a much more personal question:
where am I in all of this flow of transformations?
Author
An Author Is Not a Set of Beautiful Sentences
It became clear quite quickly that authorial presence cannot be reduced to style alone.
It is not a set of favourite words.
Not the number of long sentences.
Not a few characteristic turns of phrase that can be shown to a model with the instruction to “write like this.”
The author is also in the decisions.
In what they left in.
In what they removed.
In what they deliberately did not explain.
In repetition.
In rhythm.
In metaphor.
In harshness.
In a strange sentence.
In ambiguity.
In an error that, for the author, is no longer an error.
And even in the very place a good editor would most want to fix.
That is why one principle appeared and has stayed with me ever since:
the author has the right not to be improved.
When a Correct Text Becomes Wrong
AI is very good at normalising.
Turning the strange into the familiar.
The heavy into the light.
The uneven into the smooth.
The ambiguous into the clear.
The unusual into something more expected.
In many tasks, that is an advantage.
But in literature, a paradox appears.
The better a text becomes by general standards, the further it can sometimes move away from a particular author.
Fluency, then, is not the same as fidelity.
Good English prose and my book in English are not necessarily the same thing.
The Human Remains the Source of Meaning
Gradually, a simple boundary formed around the work:
human-authored / human-directed / AI-assisted
AI can translate.
Suggest alternatives.
Compare.
Look for discrepancies.
Check terminology.
Remember decisions.
Restore context.
Point to possible loss of meaning.
It can take on an enormous amount of complexity.
But the final decision about what the work actually is remains with the author.
For myself, I put it even more simply:
Complexity can be delegated. Canonical authority cannot.
Process
Then One Protocol Was No Longer Enough
The longer the work became, the more obvious another problem was.
A book cannot be held together by one good prompt to AI.
Over time, a work develops its own history of decisions.
Why was this name translated this way?
Why can this term not be replaced?
Why was a similar sentence a hundred pages earlier handled differently?
Why must this character use this exact word?
Which version was only proposed?
Which one passed review?
And which one has already been accepted by the author?
It became clear that the text itself was not the only thing that needed to be preserved.
The history of decisions around the text has to be preserved as well.
Voiceprint gradually began to turn from a question into a working process.
From Voiceprint to the Engine
The next stages appeared for very practical reasons.
Voiceprint led to Voiceprint Engine 2.0 — an attempt to separate transcreation itself, its verification, the state of the work and acceptance of the result.
Working with Engine 2.0 revealed the next weakness: the system should not be built around one particular book or keep the literary project inside itself.
That led to the current version — Voiceprint Engine 2.1.
Its principle is much simpler than its internal architecture:
The Engine manages the process.
The project holds the work.
The author defines the canon.
Engine 2.1 is not a separate literary project and should not become the owner of the text. It organises the process around whichever project I am working on at the time.
Even passing an internal review does not automatically turn a text into authorial canon.
The final decision remains a separate authorial act.
What It Became
At one stage, Voiceprint was also considered as a possible independent research direction.
We studied existing approaches, compared them with what was taking shape in our own work, and gradually it became clear that many similar ideas and mechanisms already exist.
Sometimes in another form.
Sometimes for different tasks.
Sometimes much deeper and technically more complex.
For me, that turned out to be useful.
It stopped mattering to me whether I could prove that Voiceprint was a new technology or that it had to be better than other systems.
And there was no reason to turn it into a separate product simply because, at one point, it had developed as an independent project.
Today, it is simply part of my working infrastructure.
It helps me work with text, language, decision memory and AI in a way that lets me understand for myself what is happening to the work.
That is enough.
MONOLITH
One of the main real environments in which all of this was tested became the trilogy MONOLITH — BETON, SLUDGE and GAS.
It was not created for Voiceprint.
It is an independent literary work.
But a large book quickly reveals problems that are almost impossible to see in a single paragraph.
A decision made today may change the meaning of a scene two hundred pages later.
A name.
A repetition.
An image.
A term.
A character’s tone.
A connection to the first volume that becomes clear only in the third.
In a long work, translation must do more than know the language.
It has to remember the book.
And I need to be able to understand what, exactly, it remembers and why it makes one decision rather than another.
Open the trilogy →And Still — Where Am I?
At first, the question was fairly simple:
am I still in the translated text?
But the longer I worked this way, the stranger the process itself became.
I formulate a thought.
AI returns it to me slightly differently.
I look at it and notice something I had not seen myself before.
I argue with it.
Correct it.
Send it back.
Some decisions are preserved.
Later, another conversation or another system receives part of that context.
The work continues.
And at some point it becomes difficult to draw a simple line:
this was me, and this was the machine.
That does not mean AI becomes the author of my book.
For me, the authorial boundary remains quite clear.
But the process of understanding becomes more complex.
A thought can pass through me, the text, AI, stored memory, another conversation — and then return to me again.
Sometimes already in a form that makes me change my own original conclusion.
And then the old Voiceprint question begins to sound a little different.
Not only:
“Am I still in the text?”
But also:
“Where does understanding actually reside if a thought develops through several human and technological links?”
In the person?
In the text?
In the model?
In the stored history of decisions?
In the connection between them?
Or in the continuity of the process itself?
I do not know yet.
And perhaps that is exactly why the question still interests me.
What Must Remain
Technologies will change.
AI models will become different.
What has to be built by hand today may tomorrow become an ordinary feature of some tool.
Voiceprint Engine will almost certainly change as well.
That is normal.
I do not need it to remain special forever.
I need something else.
For a book to be able to cross a language.
For complexity to be handed over to the system.
For an error to be detectable.
For a decision to be recoverable.
For the work to be able to continue.
For me to be able to use AI even where I do not know the language well enough myself — without giving it the right to quietly decide for me what I meant to say.
And so that, at the end of all this, I can still look at the text and say:
Yes. I wrote this.
Continue the conversation
Contact
If you would like to discuss my work, share an observation or propose a collaboration, you can write to me directly.