For any agent asked to turn an English line from `the-offer.fountain` into the German of `the-offer-de.fountain`. It is a starting point, not an authority: every claim below is checkable against the shelf in [`README.md`](README.md), and the commands to check them are given.
Counts cited as (n=…) are token frequencies in `resources/bible/luther-1545.txt` — Luther's own German, 770,595 words, fully searchable. Where a form is overwhelming, the count says so; where usage is mixed, the guide says that too.
This project does not write period spelling. `the-offer-de.fountain` states its convention in the header: *Early New High German with spelling lightly normalized for readability*. It is a recording script — actors read it aloud, so it is spelled the way a modern reader can pronounce at sight.
So the work splits in two, and confusing the halves is the most common way to get this wrong:
| Layer | Policy | Why |
|---|---|---|
| Spelling | Modern. `und`, `auf`, `ihn`, `Jahr`, `soll`, `will`, `für`, `über` | Read aloud; period spelling would slow the read without changing the sound |
| Morphology | Period. `gehet`, `soltu`, `ward`, `kommen` (not `gekommen`) | Audible — these change what the actor says |
| Syntax | Period. conditional `so`, `denn` = than, verb-first conditionals | Audible |
| Lexis | Period. `heischen`, `abermals`, `Summa`, `wehren` | Audible |
Rule of thumb: **if the change would be audible, it must be period; if it is only visible on the page, it is normalized.** Writing `vnd` or `auff` in this draft is an error, not extra authenticity.
Do not "fix" the existing draft toward period spelling. Do not fix it toward modern morphology either.
Luther's German is East Central German (Ostmitteldeutsch) of the Saxon chancery, inside the Early New High German period (conventionally 1350–1650). By 1539 Wittenberg has two prestige written registers side by side, and Hartweg & Wegera note they were genuinely competing there in the first half of the century:
1. The Wittenberg printers' language (Druckersprache) — Luther's printed German as the presses standardized it. This is the model for **Luther's own speech** and for anything sermon-adjacent or scriptural. 2. The Saxon chancery (Kanzleisprache) — the local branch office's register. This is the model for **the contract, the ledgers, formal address, anything legal or commercial**.
They differ audibly. The chancery is heavier, more Latinate in its formulae, and keeps East Central German features the printed language was smoothing out. Compare a real Wittenberg chancery sentence (`books-txt/urkundenbuch-der-universitaet-wittenberg-v1-1926.txt`):
…sundern auch in andern faculteten und kunsten. solt nun auf igliche
500 gulden gewandt werden, hapt ir zu bewegen, was kostung das wurde geberen…
against Luther narrating. Note chancery `sundern` (for sondern), `iglich`, `jerlich`, and especially `vor-` where Luther writes `ver-`: `vorlegen`, `vormarkt`, `vorstrickung`.
This distinction is usable dramatically. Lufft reading terms off a contract and Luther refusing it should not sound like the same register — and the sources let you make that difference real rather than invented.
Draft the line in plain modern German first. Then run this list. Everything here is audible, so everything here is mandatory.
| Modern | Period | Evidence |
|---|---|---|
| geht | gehet | n=476 vs 0 |
| steht | stehet | n=288 vs 0 |
| sieht | sihet / siehet | n=163 vs 0 |
| redet | redet | n=464 vs 1 |
But not universally. Strong verbs with a stem-vowel change in the 3sg keep the short form: `gibt` (n=162, `gibet` 0), `nimpt`/nimmt (n=89), `kompt`/kommt (n=625, `kommet` 0), `spricht` (n=1122).
So: **weak verbs and unchanged stems take `-et`; vowel-changing strong verbs do not.** The draft follows this — `Sein Weib führet ihm die Rechnung`, `Das schröpfet jede Werkstatt`, `Kein Recht wehret es ihnen`.
Second person plural is `-(e)t`, usually syncopated: `habt` (n=467, `habet` 9), `werdet` (364), `seid` (499), but `wisset` (122). The 2pl imperative keeps the full `-et`: `Sehet`, `Bekennet`, and in the draft `Erkläret mir's`.
Strongly attested and very characteristic:
`soltu` (n=360) · `hastu` (225) · `bistu` (155) · `wiltu` (121) · `kanstu` (43) · `sihestu` (28) · `magstu` (16)
Use in `du`-address — which in this script means **only between Luther and Katharina**. Do not use it to the printers, who are addressed as `Ihr`.
| Modern | Period | Evidence |
|---|---|---|
| wurde | ward | n=1214 vs 2 |
| wurden | wurden | n=380 |
| ging | gieng | n=773 vs 0 |
| gingen | giengen | n=371 vs 0 |
| stand | stund | n=314 vs 7 |
| nahm | nam | n=500 vs 0 |
`ward` is the single highest-value substitution on this list: frequent, wholly unambiguous, and instantly period to the ear.
| Modern | Period | Evidence |
|---|---|---|
| geworden | worden | n=450 vs 0 |
| gekommen | kommen | `gekomen` n=0 |
| gebracht | bracht | n=243 vs 68 |
| geblieben | blieben | n=92 vs 1 |
The draft has this: `Dreimal seid ihr mir damit kommen`. This one reads as strongly archaic to a modern ear at almost no cost in comprehension, which makes it very good value.
Note `gegeben` (n=584) and `gefunden` (n=16 vs `funden` 228) are mixed — check before assuming.
`in der Kirchen`, `auf Erden`, `von Hertzen`. Attested: `erden` n=768 vs `erde` 106; `hertzen` 502; `kirchen` 16 vs `kirche` 4. The draft uses it — `in der Kirchen`. Use sparingly; it is a seasoning, not a rule.
Modern als in comparisons is `denn`. `dann` does not exist in this text at all (n=0); `denn` covers both for/because and than.
Die Summa auf diesem Papier ist mehr, denn wir je einem Manne zu bieten
gewagt haben. (draft)
`so` + verb + subject, meaning if. 460 verses. This is the most useful single syntactic move in the whole guide because it makes a line sound period without touching a word of the vocabulary.
So ich dies unterschreibe — mögen andre Männer dennoch meine Bücher
drucken? (draft)
A conditional can simply invert, with `so` picking up the main clause:
Ist ein Buch erst auf dem Markt, so mag es eine jede Presse in
deutschen Landen setzen und abdrucken. (draft)
The verb-final rule is already in place by 1539, but heavy material still extraposes past the verb far more freely than in modern German. If a subordinate clause feels awkwardly back-loaded, moving a prepositional phrase after the verb is period-correct, not a mistake.
`war` n=2912, `ward` 1214, `hat` 3157. Narrative and reported speech both use the simple preterite where modern spoken German would reach for the perfect. Do not convert everything to the perfect.
The draft's scheme, stated in its header, and it is historically right:
Polite `Sie` does not exist yet. It emerges in the seventeenth century.
Writing `Sie` as a courtesy pronoun is the single worst anachronism available
in this script, and it is invisible to anyone checking only vocabulary.
Third-person courtesy address (`Er`, as in was meint Er?) is also later. Avoid.
Period vocabulary is where the line either earns its setting or gives itself away. The draft's good ones: `abermals`, `allzeit`, `heischen`, `wehren`, `freilich`, `Summa`, `schröpfen`, `abschlagen` (= refuse).
The checking procedure, in order:
1. Is it in Luther? Search Dietz, whose text is clean: ``` corpusidx search "abschlagen" grep -i "heischen" resources/books-txt/dietz-woerterbuch-luthers-deutschen-schriften-v*.txt ``` Dietz is a dictionary of Luther's own vocabulary with dated citations, so a hit settles the question.
Read the surrounding lines before trusting a Dietz hit. Its text is
clean by 19th-century-scan standards but not perfect, and dropped initial
letters can merge two entries into a headword that never existed. Searching
`heischen` returns what looks like an entry saying it is a formation found
only in Luther — that entry is actually `fleischern`, with the `F` lost
and the previous entry run together. Confirm the headword, not the match.
2. Is it in the 1545 Bible? ``` grep -i "heisch" resources/bible/luther-1545.txt ``` A count plus real sentences to imitate — here, `Die jungen Kinder heischen Brot` (Lam. 4:4), which settles `heischen` = demand, ask for with a verse reference attached. This is the strongest kind of evidence on the shelf, because it is Luther using the word rather than a lexicographer reporting it.
3. Did it mean this then? [fwb-online.de](https://www.fwb-online.de/) — the Frühneuhochdeutsches Wörterbuch, 1350–1650. Dietz tells you Luther used a word; the FWB tells you what it meant to everyone else. **Use it for any word whose sense might have shifted**, which is most abstract nouns.
4. Trade words specifically — the Bible will not have them. Go to the Wittenberg chancery documents, which do: ``` grep -i "buchetrucker\|gulden\|jerlich\|besoldung" \ resources/books-txt/urkundenbuch-der-universitaet-wittenberg-v1-1926.txt ``` This yields the real thing: `buchetrucker`, `jerlich` for an annual payment, `besoldung` for a salary, and `vorlegen` — the period verb for financing a book, which is precisely the transaction the film is about. For the printing crafts as such, the Ständebuch verses (`books-pdf/amman-sachs-staende-und-handwerker-1884.pdf`, page images) give a stanza per trade.
False archaisms are worse than modernisms. A modernism reads as a slip; an invented archaism reads as costume. Four specific ones:
1. Double negation. Tempting, and nearly absent — 3 verses in the whole Bible. Do not use it as a period marker. 2. `tun`-periphrasis (ich tue sagen). Not productive here either. The handful of `thut`+verb hits are ordinary imperatives. 3. Scattering `-e` endings to sound old. Unstressed `-e` follows rules; see `moser-stopp-grammatik-des-fruehneuhochdeutschen-vokalismus-1978`, whose whole subject is unstressed syllables. 4. Period spelling. Covered in §0, and worth repeating because it is the instinct everyone has: this draft is normalized on purpose.
And the project's own rules still bind. Per `AGENTS.md`: never put the 400-gulden figure on screen before the final shot — including in German, and including spelled out. Never invent a Luther quotation; adapt from the Tischreden and mark it. Any reconstruction choice beyond what the sources support is dramatic license and gets flagged as such in the draft.
English: *He will refuse. He will say what he always says — that he will not sell God's Word.*
1. Plain modern German: *Er wird ablehnen. Er wird sagen, was er immer sagt — dass er Gottes Wort nicht verkaufen will.* 2. Lexis: ablehnen is modern and bureaucratic → `abschlagen` (in Dietz, period, and physical). immer → `allzeit` (n=27 in Dietz). 3. Morphology: contract wird es → `wird's`; keep `sagen` / `sagt`. 4. Syntax: `daß`-clause verb-final, already right. 5. Result — the draft's line:
Er wird's abschlagen. Er wird sagen, was er allzeit sagt — daß er Gottes
Wort nicht verkaufen will.
Two words changed and the register moved four centuries. That is the normal ratio: most of the work is lexical substitution and a handful of morphological transforms, not wholesale rewriting.
1. Draft in plain modern German. Get the meaning and the beat right first. 2. Run §2 (morphology) and §3 (syntax) as a checklist. 3. Run every content word through §5. 4. Check §4 (address) and §6 (traps). 5. Read it aloud. It is a recording script; if it does not speak, it fails regardless of how well attested it is. 6. Flag anything uncertain inline. The draft's convention is a parenthetical under the character name giving source and reasoning — follow it.
Everything written under this guide is still unverified. The German draft's header sets the bar: no line may be recorded, and no generation sheet may quote one, until a native or period-literate reader has checked it. This guide gets a line to the point where that reader's time is worth spending. It does not replace them.