DazedAnon
56453e922b
- Add image readiness checks and navigation from workflow Step 6 - Keep long Image Manager folder lists bounded and scrollable - Update workflow guidance, button labels, and regression coverage - Disable code 357 and 108 translation by default
95 lines
4.5 KiB
Markdown
95 lines
4.5 KiB
Markdown
# Example walkthrough
|
|
|
|
This page is a concrete story of translating one game, step by step.
|
|
Use it to see the *order* of work. Details for each Workflow step live in the other Guide sections and in each step's **?** button.
|
|
|
|
Assume: Claude Sonnet 4.6 for paid quality + Batch, or Mistral if you are on the free tier.
|
|
Also assume Cursor (or VS Code) has the **game folder** open for glossary / UI script work.
|
|
|
|
---
|
|
|
|
## Example A - First RPG Maker MV/MZ game
|
|
|
|
You just downloaded a small MV game. You want English dialogue, menus that make sense, and a playable build - without boiling the ocean on day one.
|
|
|
|
### 1. Get the tool ready
|
|
|
|
1. Launch DazedTL (opens on this Guide).
|
|
2. Open **Configuration**, set your API key and model, save.
|
|
3. Open the game folder in Cursor so the agent can read JSON later.
|
|
|
|
### 2. Point Workflow at the game
|
|
|
|
1. Open **Workflow** → engine **RPG Maker**.
|
|
2. Step 0 - browse to the folder with `Game.exe` / `www/` (MV) or `data/` (MZ).
|
|
3. When the file list appears, start small: check core DB files (`Actors`, `Items`, `System`, …) plus **one** early map (`Map001.json`). Import those into `files/`.
|
|
|
|
You are not importing the whole game yet. A tiny slice proves the pipeline before you burn tokens on every map.
|
|
|
|
### 3. Set up names and voice (Step 2)
|
|
|
|
1. Run **Parse Speakers** if the game uses nameplates the tool can harvest.
|
|
2. **Copy Project Setup**, paste into Cursor with the game open, let the agent return glossary / speakers / quirks / game skill.
|
|
3. Paste each block into the Workflow editors and save.
|
|
|
|
Now the translation API has character genders, terms, and a short Translation Frame instead of guessing blind.
|
|
|
|
### 4. Translate a slice (Step 3)
|
|
|
|
1. Prefer **Normal** mode while you are learning.
|
|
2. Run **Phase 0** (database names/descriptions).
|
|
3. Run **Phase 1** on the checked map only.
|
|
4. Skim `translated/` and `log/` - do names look consistent? Does dialogue sound okay?
|
|
|
|
If something is wrong, fix vocab / quirks *now* before you translate fifty maps.
|
|
|
|
### 5. Put it in the game and play
|
|
|
|
1. Step 5 - **Export Active Files** (only what you imported).
|
|
2. Launch the game and walk that first map.
|
|
3. Note leftovers: overflow, wrong names, Japanese in a menu.
|
|
|
|
Fix leftovers next (re-translate a file, tweak wrap width in Config, or OCR → search in the IDE). Do not jump to Phase 2 yet.
|
|
|
|
### 6. Scale up
|
|
|
|
When the sample map looks good:
|
|
|
|
1. Step 0 - import the rest of the maps / CommonEvents (or the next chunk).
|
|
2. Step 3 Phase 1 again - use **Batch** with Claude if the set is large; stay on **Normal** with Mistral.
|
|
3. Export again and play further.
|
|
|
|
### 7. Risky codes and UI (only when needed)
|
|
|
|
1. Step 4 - copy the Plugin / risky-codes prompt, audit the game in your IDE, enable only codes that hold player-visible text, then run Phase 2 carefully.
|
|
2. Step 5 - for MV/MZ, copy vocab + the plugins.js prompt and edit `plugins.js` in the IDE (player-facing strings only). For **Ace**, edit `ace_json/scripts/*.rb` the same way, then Export so **RV2JSON** packs `Scripts.rvdata2`.
|
|
3. Step 6 (MV/MZ) - check image setup, then decrypt / translate / patch any bitmap UI text.
|
|
4. Step 7 (MV/MZ) - install TL Inspector / Forge if you want in-game inspection helpers.
|
|
|
|
### What “done enough” looks like
|
|
|
|
You can play through without walls of Japanese, menus are readable, and you know how to OCR → search → re-export when something slips through. Perfect first pass is rare; the loop is translate → play → fix → expand.
|
|
|
|
---
|
|
|
|
## Example B - Same idea on WolfDawn (short)
|
|
|
|
1. Workflow → **Wolf RPG (WolfDawn)** → Step 0 extract + import.
|
|
2. Step 2 - build vocab in Cursor from the extracted files (skip `names.json` value names).
|
|
3. Step 3 - translate **safe** names only.
|
|
4. Step 4 - foundation DB first; add narrative DB if the discovery report says the game is DB-heavy.
|
|
5. Step 5 - maps / events (Batch for huge CommonEvent files on Claude).
|
|
6. Step 6 Precheck → Step 7 **Inject all** → Step 8 Package → play.
|
|
7. Step 9 - fix wrap on overflowing lines → Inject → Package again.
|
|
|
|
Same philosophy as RPG Maker: prove quality on a small slice of content, then scale; pack/inject before you judge overflow in-game.
|
|
|
|
---
|
|
|
|
## Habits that save time
|
|
|
|
- **Git** the game folder before big Export / Inject passes so you can roll back.
|
|
- Start with **one map** (or one DB sheet), not the whole project.
|
|
- Fix glossary and speaker flags early - they multiply across every later file.
|
|
- Use Batch for bulk overnight work; Normal for quick iteration.
|
|
- Leftover Japanese: screenshot → OCR → search in the IDE → re-translate or hand-edit → export/inject again.
|