starker Nora-Start
Observations
Field notes from an ongoing experiment.
Elenavision runs in public.
The characters meet real platforms, real recommendation systems and real people.
Some hypotheses work. Others do not. Some results look strong without it being immediately clear why. And sometimes almost nothing happens.
This page documents what we observe along the way.
Not as a success story smoothed out afterwards.
But while the experiment is running.
Facts. Reactions. Learnings. And sometimes: no learning yet.
The experiments share methods and findings.
The characters themselves stay separate.
Learn globally. Apply locally. Declare the transfer.
Experiment 01
Early August to early September 2026 — Elena starts
Elena was the first public experiment of Elenavision.
The start was in early August, first on Instagram. Facebook was added later as a second test channel.
In the first weeks it was not only about reach.
First it had to become clear whether a synthetic character could be produced stably enough to remain the same person across many different scenes, systems and production methods.
Published
In the first evaluated Facebook period there appeared:
23publications
18short videos
4images
~13.9KViews
of which:
18 short videos
4 images
1 Textpost
Result
Zusammen kamen diese publications auf rund:
13.900 Views
Distribution was very uneven.
The strongest video reached about 3,100 views and 8 new followers.
Another came to about 1,600 views and 3 new followers.
Many other publications stayed at a few hundred views despite comparable technical quality.
No clear link between production effort and distribution was visible.
Our response
Do not change Elena radically.
The visual identity stays stable.
Instead, bounded variables are tested:
Content, visual idea, location, outfit, thumbnail, timing and platform.
A weak post is no reason to reinvent the character.
Stand
Technical and visual stability works.
A reproducible content formula for reach does not, so far.
Status: we keep observing.
Experiment 02
Late August 2026 — Nora starts differently
Nora was built as a second experiment under different conditions.
Character profile, positioning and large parts of the creative direction emerged much more strongly through AI-supported iteration.
The public reaction differed from Elena almost immediately.
Published
In the first evaluated Facebook period there appeared:
11publications
6short videos
5image posts
~233KViews
>1,200Facebook followers in about a week
of which:
6 short videos
5 image posts
Result
Diese 11 publications erreichten zusammen rund:
233.000 Views
Several videos were strongly distributed to people outside the existing follower base.
Individual publications came to about:
80.000 Views
75.000 Views
47.000 Views
and each produced several hundred new followers.
Within about a week Nora passed 1,200 followers on Facebook.
That was a rocket start.
Elena vs. Nora — the first seven days
Facebook views in the first seven days from the first Facebook post. Start and end values are measured; the curve between them is visualised.
Both characters start at zero.
Die Endwerte entsprechen den realen Resultsen der ersten sieben Tage.
The curve between them is for visualisation and does not claim exact daily measurement.
Our response
Do not copy it immediately.
Do not claim we have found a formula.
Let Nora keep running and watch whether the pattern holds.
Denn hinter dem Result steckt keine einzelne Variable.
Possible factors include:
- Character and visual effect
- Cabin-crew context
- Audience
- Content
- Platform
- Timing
- algorithmic classification
- Chance
These factors cannot currently be cleanly separated.
Stand
The result is exceptional.
But exceptional is not the same as reproducible.
Perhaps we hit several factors very well.
Perhaps part of it was a lottery win.
The truth is probably somewhere in between.
Status: rocket start. Not a recipe yet.
Experiment 02 — Update
10 September 2026 — Nora reaches 2.5k on Facebook
Observed
Nora started on Facebook with zero followers.
On 10 September she passed the 2,500 mark.
The documented figure on 10 September: 2.506 Follower.
About 17 days after the first stored Facebook post of 24 August.
Growth was not even.
A small beginning was followed by a clear surge.
Nora on Facebook: from 0 followers at the start to 2,506 on 10 September 2026. About 17 days since the first stored Facebook post. The end values are documented; the curve between them is schematic and does not show measured daily values.
More than 1,200 Facebook followers within about a week — an earlier, already documented threshold, without an exact day in this chart.
Reading
Our current reading: outside what we would expect from a normal organic cold start. The shape does fit a recommendation or viral hit: little movement at first, then a strong surge and a tail.
This describes an observed pattern. It isolates no cause.
How much character, content, platform, timing or chance contributed, we still cannot cleanly separate.
Our response
Celebrate briefly. Then keep observing.
Do not turn it into a daily growth rate.
And still claim no playbook.
Stand
A measurable milestone.
Not yet evidence of a repeatable method or a durably stable audience.
Status: 2.5k reached. Not a recipe yet.
Update — Ende September / Anfang Oktober 2026
Highlight — Nora passes 1 million views
1.232.823
views since the start on 24 August 2026
Nora Layover has reached 1,232,823 views in total since her start on 24 August 2026.
Experiment 02 thus passes the mark of 1 million views in less than six weeks.
For Elenavision this is an important milestone.
Not because a formula would thereby be proven.
But because a cold start has by now become a large real dataset.
The open question remains:
Which parts of that are reproducible — and which are not?
Nora — Raptor, Warp and late rediscovery
Several Nora reels developed much more strongly than earlier content.
Intern unterscheiden wir inzwischen:
- Diesel — langsam und stetig
- Raptor — anhaltende Ausbruchswelle
- Warp — clearly above that again
Dabei wurde eine Regel besonders wichtig:
Not the peak makes a Raptor, but the wave.
Ein einzelner Spike reicht nicht.
Entscheidend ist eine anhaltende Verteilungswelle.
Particularly interesting was an older Mykonos reel that was strongly distributed again only about two weeks after publication.
Learning
Alte Inhalte sind nicht automatisch „tot“.
We do not yet know why such a late rediscovery happens.
Elena — a clearer content mechanism
Bei Elena hat sich ein wiederkehrendes Muster herausgebildet:
small action → disproportionate consequence → minimal reaction
Intern:
Overreaction outside. Underreaction inside.
Oder einfach:
Wow.
Der Mechanismus ist wiedererkennbar.
Er erzeugt aber nicht automatisch dieselbe Distribution.
Learning
Ein klarer Content-Mechanismus kann Wiedererkennbarkeit schaffen.
Er ist noch keine Reichweitenformel.
Instagram — retention instead of likes?
On Elena, the few Instagram reels that were distributed more strongly shared a pattern:
The retention curves sat clearly above those of the other content.
Likes and other simple single metrics showed no similarly clear pattern.
Daraus entstand:
Herbert Gate Hypothesis v1
Early retention may matter more for further distribution than simple engagement numbers.
Noch keine Gewissheit.
Kleine Stichprobe.
Aber erstmals eine klar testbare Hypothese.
AI as a source of ideas
Ein weiteres Learning betrifft das AI-System selbst.
Our AI roles increasingly deliver not only variants or execution.
Sie erweitern manchmal selbst den Suchraum:
neue Achsen,
neue Mechanismen,
neue Kombinationen.
Learning
For us, interesting AI creativity begins where not only more variants appear, but new spaces for thinking and ideas.
System learning
With growing complexity it also became clear:
A universal bot is not automatically the best solution.
Specialised roles, clear handovers and independent cross-checks work better.
Learning
Nicht nur bessere Modelle verbessern ein AI-System. Bessere Architektur tut es auch.
What we still do not know
- warum genau bestimmte Posts ausbrechen,
- why older content is suddenly distributed again,
- which retention features are actually causal,
- und welche Muster langfristig reproduzierbar sind.
Facts. Reactions. Learnings. And sometimes: no learning yet.
Observing further.
Eine Figur. Zwei Platformen. Zwei Welten.
Nora showed at the same time how little results simply transfer from one platform to another.
bislang deutlich kleiner
While Facebook distributed strongly within a few days and quickly brought new followers, Instagram stayed small in the same period.
Auch bei Elena entwickelte sich keine vergleichbare Dynamik.
Our response
Facebook und Instagram nicht mehr gedanklich als dieselbe Versuchsumgebung behandeln.
Dieselbe Figur kann auf unterschiedlichen Platformen vollkommen unterschiedlich laufen.
Stand
Warum das passiert, ist noch nicht sauber isoliert.
Hence no algorithm mythology after the fact.
Status: beobachten, vergleichen, nicht raten.
Production note — world → placement → identity
With Nora a new production path appeared for more complex scenes.
Instead of forcing environment, character and identity in a single generation step, they are separated more strongly:
World → character placement → identity
Welt
Zuerst wird die Umgebung stabilisiert.
Platzierung
Then the character is placed correctly in the space.
Identity
Only after that is the final character identity fixed.
Our response
Use the workflow where complex scenes benefit from stable spatial geometry.
Nicht als Dogma.
Wenn ein einfacherer Weg funktioniert, bleibt der einfachere Weg erlaubt.
Stand
The workflow reduces typical problems such as wrong scale, unstable geometry and identity drift.
Status: echtes Produktions-Learning.
Community Note — Eine Figur braucht auch eine Stimme
Mit Noras schnellerem Wachstum kamen mehr Kommentare und damit eine neue Frage:
Wie bleibt eine synthetische Figur nicht nur optisch, sondern auch sozial konsistent?
For that, WIB appeared: a two-stage AI-supported community system.
01
Comment / Emoji / GIF
02
AI Preprocessing
03
Reply Author
04
Human Gate
05
Response
A first AI takes ingest, technical assignment, pre-classification and context.
Not only text is taken into account, but also non-verbal communication:
Emoji, stickers, GIFs and visual attachments.
If an individual reply makes sense, a second AI prepares it.
For Nora a style of her own is built:
schnippisch, charmant, selbstbewusst — aber nicht sexuell flirty.
Replies can be prepared in more than one language.
Before publication the human gate remains.
Our response
Nicht jede Interaktion beantworten.
Nicht jede positive Reaktion braucht Text.
Nicht jede Anmache verdient Aufmerksamkeit.
Der Stil soll konsistent werden, ohne Nora zu einem Chatbot zu machen.
Stand
Die technische Pipeline funktioniert.
Ob daraus langfristig wirklich eine wiedererkennbare soziale Stimme entsteht, muss erst die Zeit zeigen.
Status: laufendes Experiment.
Production note — museum-grade AI failures
This belongs to the running experiment as well: AI systems can become unusable in quite peculiar ways.
One model stayed stuck in a Spock persona.
Ein anderes lieferte nur noch images, obwohl Text gebraucht wurde.
And Beta AI protected the master assets so thoroughly that it practically took itself out of operation.
Intern: museumsreif.
In operation: failures that block the work.
These are operational anecdotes from the production system. Not a controlled evaluation, and not evidence of consciousness or sentience.
Our response
Mit Humor nehmen. Die Aufgabe trotzdem ernst nehmen.
Check whether a system is still doing the work required — even when its answers sound very consistent.
Stand
Die Episoden sind dokumentiert. Eine gemeinsame technische Ursache ist daraus noch nicht abgeleitet.
Status: Betriebsnotiz. Noch kein allgemeines Learning.
Facts. Reactions. Learnings. And sometimes: no learning yet.
We record that too, while it is happening.
What we know so far
Einige Dinge wirken inzwischen belastbar:
- Synthetic character identity can be kept stable across many different productions.
- A single good AI output is not enough for that. Repeatability is the real work.
- Platforms can treat the same or similar production very differently.
- Produktionsaufwand garantiert keine Reichweite.
- AI can not only generate output, but increasingly support character development, QA, analysis and community processes.
- Human Gate bleibt besonders dort relevant, wo Geschmack, Kontext, Verantwortung oder Mehrdeutigkeit ins Spiel kommen.
What we do not know yet
Die interessanteren Fragen sind weiterhin offen:
- Why does one character receive cold distribution and another hardly any?
- Welche Faktoren sind reproduzierbar?
- Which effects are only temporary?
- Wie stark ist Chance?
- Wie viele unterschiedliche Figuren lassen sich mit derselben Produktionsarchitektur sinnvoll betreiben?
- Entsteht aus Aufmerksamkeit ein stabiles Publikum?
- Und irgendwann: entsteht daraus wirtschaftlicher Wert?
Wir haben darauf noch keine fertigen Antworten.
Das ist der Punkt des Experiments.
Next entry
When something measurable happens.
Not because the calendar says we should blog again.
Observing further.