EN
LAB · SYNTHETIC
Human creativity and machine in the same conversation.
Ei, Humano speaks to you through voice and reads the intent; Campaign Management shows what the piece's evidence supports.
Who is Labtech
A laboratory for communication between people and machines.
Labtech is an independent applied technology laboratory. In Synthetic, it brings together two experiences built on the same foundation: language, data, and artificial intelligence running on its own infrastructure.
Ei, Humano studies the conversation: what the person says, with what intention, in what context — and how to respond with personality, without turning the conversation into an interrogation.
Campaign Management studies advertising communication: what the ad shows and says, in which network it works best, and with what evidence this can be affirmed.
In both, the rule is the same: every statement carries a stamp. What the source does not support appears as a gap, and what is still in the laboratory is presented as laboratory.
Research rigor
Evidence with stamp and source
The human decides
Responsibility for the outcome
Experiences
Two experiments.
The same intelligence.
The conversation and communication under the same method: read carefully, measure what can be measured, cite the source, and declare what is still missing.
Ei, Humano
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Voice conversation
In the tabletop app, you speak; the speech is transcribed on our GPU, the response comes back in voice, and the audio is deleted. Only the text remains, with proof that the sound was discarded.
View in detailIntention and context
Each turn is read to estimate intention, intensity, topic change, and repeated question. The reading turns into a response plan that avoids repeating questions and avoids interrogation.
View in detailLexicon and personality
A lexicon of 37 concepts — epistemic, conversational, creative, social — gives the conversation a distinct tone: colloquial, jargon-free, with restrained humor when the subject requires caution.
View in detailReading Layers
Affective, conversational, professional, and limits: the screen displays, turn by turn, what each layer read — and declares the affective as not yet implemented, instead of pretending.
View in detailLearning from speeches
Each response can receive a «like» or «dislike» and a critique, which the AI rewrites with greater clarity. The chain becomes a curated example — and nothing is trained alone: the example only advances after validation.
View in detailTaxonomy, semantics, and ontology
The next layer: concepts and relations between concepts, for the conversation to understand what is being talked about and not just the words. The data model already exists; the engine that uses it is still under construction.
View in detailCampaign Management
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Reading of the play
Video, image, or public link: the speech is transcribed, the scenes are read block by block by a vision model, and the result arrives in stages, live.
View in detailNetworks and channels
Instagram, TikTok, YouTube, Facebook, and WhatsApp evaluated side by side. The chosen network comes with the reason — and each network not chosen as well.
View in detailScenario simulation
Adjustable audience weights, sensitivity, and a recalculated winner in the moment by a deterministic engine: the same scenario always yields the same result.
View in detailEvidence with seal
Confirmed, hypothesis, gap, divergence: each public reading comes marked, with the test that would refute it and the excerpt from the play that supports it.
View in detailScientific base
References checked one by one support the criteria of the analysis — and those that did not withstand the check were removed, with a record of the reason.
View in detailSuggestions and script
Objective suggestions for the play and a rewritten script by AI, for those who decide to compare the original version with an alternative.
View in detailSuccess and failure criteria
The next step: comparing campaigns with each other and learning from cases of success and failure. Not yet built — and for that reason presented here as a laboratory.
View in detailIn detail
A section for each capability.
What each part does, how it was built, and what stage it is in — including what is still in the laboratory.
Voice conversation
What does
In the tabletop app, you speak; the speech is transcribed on our GPU, the response comes back in voice, and the audio is deleted. Only the text remains, with proof that the sound was discarded.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Intention and context
What does
Each turn is read to estimate intention, intensity, topic change, and repeated question. The reading turns into a response plan that avoids repeating questions and avoids interrogation.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Lexicon and personality
What does
A lexicon of 37 concepts — epistemic, conversational, creative, social — gives the conversation a distinct tone: colloquial, jargon-free, with restrained humor when the subject requires caution.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Reading Layers
What does
Affective, conversational, professional, and limits: the screen displays, turn by turn, what each layer read — and declares the affective as not yet implemented, instead of pretending.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Learning from speeches
What does
Each response can receive a «like» or «dislike» and a critique, which the AI rewrites with greater clarity. The chain becomes a curated example — and nothing is trained alone: the example only advances after validation.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Taxonomy, semantics, and ontology
What does
The next layer: concepts and relations between concepts, for the conversation to understand what is being talked about and not just the words. The data model already exists; the engine that uses it is still under construction.
How we build
A voice conversation between a human and a machine: the machine listens, estimates the speaker's intention, and responds with its own personality — and each response can be rated by the person who conversed.
Reading of the play
What does
Video, image, or public link: the speech is transcribed, the scenes are read block by block by a vision model, and the result arrives in stages, live.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Networks and channels
What does
Instagram, TikTok, YouTube, Facebook, and WhatsApp evaluated side by side. The chosen network comes with the reason — and each network not chosen as well.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Scenario simulation
What does
Adjustable audience weights, sensitivity, and a recalculated winner in the moment by a deterministic engine: the same scenario always yields the same result.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Evidence with seal
What does
Confirmed, hypothesis, gap, divergence: each public reading comes marked, with the test that would refute it and the excerpt from the play that supports it.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Scientific base
What does
References checked one by one support the criteria of the analysis — and those that did not withstand the check were removed, with a record of the reason.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Suggestions and script
What does
Objective suggestions for the play and a rewritten script by AI, for those who decide to compare the original version with an alternative.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
Success and failure criteria
What does
The next step: comparing campaigns with each other and learning from cases of success and failure. Not yet built — and for that reason presented here as a laboratory.
How we build
Intelligence applied to advertising: the ad is read in full, each statement bears the seal of the evidence that supports it, and the decision rests with those who know the brand.
From conversation to decision