Bahasa Indonesia, Melayu, Filipino
Memuat data interpretasi naratif secara real-time...
The Full Story
Selamat datang di Detroit tahun 2038, kota penuh android yang mulai punya perasaan dan niat buat demo berjilid-jilid. Ikuti kisah Connor, Kara, dan Markus dalam nentuin nasib ras mereka—dan nasib umat manusia yang kebetulan lagi banyak drama.
Gak mau kan main game interaktif tapi terjemahannya kaku kayak robot karatan? Project gila ini ngerombak 116 ribu kata pake pipeline neural 8-tahap biar bahasanya luwes abis. Kita masukin bumbu-bumbu slang lokal dan gaya bahasa yang pas buat tiap karakter—omelan Hank yang judes sampe pidato Markus yang bikin merinding semuanya disesuaiin biar pas di telinga orang kita. Gak ada terjemahan asal-asalan, ini murni kerja keras modder yang gak mau liat gamenya kerasa asing!
Experimental Build
Author's Notes
24-03-2026: Alpha-1. This one edit binary file. So you should back up your file first.
30-03-2026: Alpha-2. Improved contextual translations by sorting, yeah, sorting text by scene can make translations better.
=== Audit Teknis & Semantik Lokalisasi DETROIT BECOME HUMAN ===
1. SKALA LINGUISTIK & CAKUPAN
- Skala Proyek: Sekitar 116,663 kata diproses melalui alur neural 8-tahap.
- Cakupan Bahasa: Dukungan trilingual penuh untuk pasar Indonesia, Malaysia, dan Filipina.
- Status Kelengkapan: Indonesia: 94.7%, Malay: 95.8%, Filipino: 91.3%
- Analisis Variasi Leksikal: Source -> Density: 69.0% | Diversity: 6.1%, Indonesia -> Density: 72.8% | Diversity: 7.8%, Malay -> Density: 73.9% | Diversity: 6.4%, Filipino -> Density: 64.0% | Diversity: 7.7%
2. VALIDASI NEURAL & AKURASI
- Skor Keselarasan Semantik (Platt Score): Indonesia: 87%, Malay: 85%, Filipino: 84%
(Skor ini mengukur seberapa akurat terjemahan mempertahankan makna asli dari teks sumber.)
- Gaya Bahasa Karakter: Penyesuaian gaya (gaul, formal, santai) telah diterapkan pada 54 karakter unik.
- Pemulihan Struktur Otomatis (Tag Repair): 252 tag kode game telah dipulihkan secara presisi.
3. KAPABILITAS ENGINE
- Pipeline: Austronesian Localization System (Neural LoRA-Adaptive Architecture).
- Pengenalan Entitas: Ekstraksi penuh untuk terminologi spesifik game dan konstanta lore.
Attention: This version contains 1.7% watermarks. Support this project on Trakteer or Ko-fi to download NON-WATERMARKED version.
Linguistic Analysis Report
Discourse analysis using Gemma embeddings. Classifies rhetorical register across the corpus to ensure tonal consistency with source narrative assets.
Emotional tone mapped via dot-product similarity between extracted dialog embeddings and predefined sentiment anchors using zero-shot semantic alignment.
DISCLOSURE: Profiling data generated algorithmically via zero-shot inference and semantic vector alignment. Represents AI interpretation of the dataset corpus, not explicit ground-truth statistics from the underlying game engine or internal metrics. Use as a heuristic guide for context mapping.
Cross-Lingual Quality Matrix
Semantic alignment quantified via Multilingual E5 Large Instruct (RoBERTa based) bitext mining. NER entities preserved using GLiNER heuristic extraction protocols to maintain terminological invariance.
* Sim = Cosine Similarity (Vector Space) · Density = Content/Total Tokens · Diversity = TTR (Type-Token Ratio) · "src" = Source Baseline · Named Entities enforced via GLiNER mining.
Heuristic markup verification utilizing multi-pass validation and correction to ensure syntactical integrity of control codes and visual tags.
Name
Narrative Profile
Associated Entities
Semantic Archetypes
NLP Pipeline Intelligence