Accept alias subtitle suffixes when detecting existing outputs

src/scripto/core/languages.py

@@ -16,6 +16,10 @@ class LanguageSpec:
     code: str          # config value, e.g. "zh"
     prompt_name: str   # name used inside translation prompts
     suffix: str        # filename suffix, e.g. ".zh"
+    # Foreign suffixes accepted as this language when *detecting* existing
+    # outputs (subtitles named by other tools / by hand). Writing always
+    # uses ``suffix``.
+    aliases: tuple[str, ...] = ()
 
 
 _REGISTRY: dict[str, LanguageSpec] = {}
@@ -43,7 +47,18 @@ def suffix_map() -> dict[str, str]:
     return {spec.code: spec.suffix for spec in _REGISTRY.values()}
 
 
-register_language(LanguageSpec(code="en", prompt_name="English", suffix=".en"))
-register_language(LanguageSpec(code="zh", prompt_name="Simplified Chinese", suffix=".zh"))
+def alias_suffixes(code: str) -> tuple[str, ...]:
+    """Detection-only suffixes for ``code`` (empty for unknown codes)."""
+    spec = _REGISTRY.get(code)
+    return spec.aliases if spec else ()
+
+
+register_language(LanguageSpec(
+    code="en", prompt_name="English", suffix=".en", aliases=(".eng",),
+))
+register_language(LanguageSpec(
+    code="zh", prompt_name="Simplified Chinese", suffix=".zh",
+    aliases=(".cn", ".chs", ".zh-cn", ".zh-hans", ".zh_cn"),
+))
 register_language(LanguageSpec(code="ja", prompt_name="Japanese", suffix=".ja"))
 register_language(LanguageSpec(code="ko", prompt_name="Korean", suffix=".ko"))

src/scripto/core/output.py

@@ -16,7 +16,7 @@
 from pathlib import Path
 
 from ..engines.base import TranscribeResult
-from .languages import suffix_map
+from .languages import alias_suffixes, known_languages, suffix_map
 
 DEFAULT_SUFFIXES = suffix_map()  # single source of truth: core/languages.py
 FORMATS = ("srt", "txt", "vtt", "json")
@@ -64,14 +64,22 @@ def existing_transcript(
     Export-dir runs always re-check the exact default name only.
     """
     suffixes = suffix_map or DEFAULT_SUFFIXES
+    directory = export_dir if export_dir is not None else source.parent
     if language:
         path = output_path(
             source, language=language, fmt=fmt,
             suffix_map=suffix_map, export_dir=export_dir,
         )
-        return path if path.exists() else None
-    directory = export_dir if export_dir is not None else source.parent
-    for suffix in suffixes.values():
+        if path.exists():
+            return path
+        for suffix in alias_suffixes(language):
+            candidate = directory / f"{source.stem}{suffix}.{fmt}"
+            if candidate.exists():
+                return candidate
+        return None
+    detect = list(suffixes.values())
+    detect += [a for spec in known_languages() for a in spec.aliases]
+    for suffix in detect:
         candidate = directory / f"{source.stem}{suffix}.{fmt}"
         if candidate.exists():
             return candidate

src/scripto/translate/stage.py

@@ -64,8 +64,13 @@ def translate(
         if out_path == srt_path:
             # transcript already carries the target language suffix — nothing to do
             return []
-        if out_path.exists() and not self._overwrite:
-            return [out_path]
+        if not self._overwrite:
+            # Accept alias-named files (lecture.cn.srt dropped in by hand or
+            # by another tool) as the existing translation — never redo it.
+            for suffix in (self._target.suffix, *self._target.aliases):
+                candidate = source.with_name(f"{source.stem}{suffix}.srt")
+                if candidate != srt_path and candidate.exists():
+                    return [candidate]
 
         content = srt_path.read_text(encoding="utf-8")
         translated = self.translate_content(content, stop_check=stop_check, progress=progress)

tests/test_output.py

@@ -77,3 +77,17 @@ def test_vtt_txt_json_writers(tmp_path):
     payload = json.loads((tmp_path / "o.json").read_text(encoding="utf-8"))
     assert payload["language"] == "en"
     assert len(payload["segments"]) == 2
+
+
+def test_existing_transcript_accepts_alias_suffixes(tmp_path):
+    from scripto.core.output import existing_transcript
+
+    source = tmp_path / "lecture.mp4"
+    source.write_bytes(b"x")
+    (tmp_path / "lecture.cn.srt").write_text("1\n", encoding="utf-8")
+
+    forced = existing_transcript(source, fmt="srt", language="zh")
+    assert forced is not None and forced.name == "lecture.cn.srt"
+    auto = existing_transcript(source, fmt="srt", language=None)
+    assert auto is not None and auto.name == "lecture.cn.srt"
+    assert existing_transcript(source, fmt="srt", language="ja") is None

tests/test_translate_stage.py

@@ -157,3 +157,21 @@ def test_release_unloads_model(tmp_path):
     stage, _src, _srt = make_stage(client, tmp_path)
     stage.release()
     assert client.unloaded
+
+
+def test_translate_skips_when_alias_named_translation_exists(tmp_path):
+    # A hand-dropped lecture.cn.srt counts as the zh translation: returned
+    # as-is, nothing sent to the model.
+    source = tmp_path / "lecture.mp4"
+    source.write_bytes(b"x")
+    transcript = tmp_path / "lecture.en.srt"
+    transcript.write_text(SAMPLE, encoding="utf-8")
+    existing = tmp_path / "lecture.cn.srt"
+    existing.write_text(SAMPLE, encoding="utf-8")
+
+    client = FakeClient()
+    stage = OllamaTranslateStage(client, model="m", target="zh")
+    out = stage.translate(transcript, source)
+
+    assert out == [existing]
+    assert client.calls == []