Schulbuch-Übertragung (Foto/PDF -> Markdown/DOCX)

Multi-VLM Modus:

VLM-Slots konfigurieren

Modell 1 (Primär)

Modell 2 (Sekundär)

Modell 3 (Tertiär)

Arbitrator

Entscheidungs-Modus:

💡 Alle Endpunkte und API-Keys werden von den globalen Einstellungen übernommen.

Basis-URL:

API Key:

System Prompt (Flipping):

* Extract all text in a logical order, as if you were reading it naturally.
* NEVER add additional sentences.
* Rebuild the headings.
* Format the text with markdown syntax.
* Don't translate text.
* Don't give answers to questions from the text.
* Don't solve math problems.
* Return what is on the page, NEVER add additional sentences, that describe your actions.
* Don't add ```markdown or ```
* Describe illustrations at their logical positions comprehensively in German. Use the following internal structure:

((Bild))<br/>
Art der Abbildung (z. B. Foto, Zeichnung, Diagramm):<br/>
Beschreibung: …  / Bildtext: …<br/>
((/Bild))

* Place captions and titles above the ((Bild))((/Bild)) tag if they exist in the source.
* Use LaTeX for ALL mathematical expressions.
* Put $$ around LaTeX sections.
* Put $ around LaTeX expressions embedded in text.
* Don't put [ or ] around LaTeX.
* Do the ocr twice and check if nothing is missing from the original page.
* Double check mathematical expressions for correctness.
* Double check if the markdown syntax is correct.

Vision Prompt (Anschauen & Fragen):

Erweiterte Parameter

Temperatur (0-2):

Repeat Penalty:

Top P:

Top K:

Verzögerung (ms):

Wiederholungen:

Denktiefe (Thinking):

Thinking (Qwen):

E-Buch-OCR

VLM-Slots konfigurieren

Arbitrator

Eingabe

E-Buch-OCR

VLM-Slots konfigurieren

Arbitrator

Eingabe

Sicherheit

Tastenkürzel

Kamera / Vision

Inhaltsverzeichnis bearbeiten

Bestätigung