Transcription: dwindle model choice to Whisper Base + Whisper Large v3

Registry now lists exactly two models instead of five:
- 'Whisper Base' — names the actual bundled model (was 'Base (default)').
- 'Whisper Large v3' — kept; honest copy on size/speed/RAM.
tiny/small/medium removed from the registry (validate_model_name now
rejects them; existing rows keep their saved model — all current data
is 'base', still valid).
Detail screen shows the model that WILL run by real display name
('Whisper Base (default)') instead of 'Use default (base)'; the picker
lists the two models with the default marked inline.
Backend 78 passed/1 skipped; app 34/34; engine hot-reloaded clean.

app/src/main/kotlin/com/shonar/desktop/Screens.kt

@@ -601,24 +601,24 @@ fun DetailScreen(state: DesktopState) {
                 }
             }
         }
-        // Model override: "Use default" or a specific size.
+        // Model override: shows the model that WILL run, by its real name.
         Row(verticalAlignment = Alignment.CenterVertically) {
             Text("Model: ", style = MaterialTheme.typography.bodyMedium)
-            val current = detail.overrideModel
+            val current = detail.overrideModel ?: models?.defaultModel ?: "base"
-                ?: models?.defaultModel
+            val currentLabel = models?.byName(current)?.displayName ?: current
-                ?: "base"
             TextButton({ showModels = !showModels }) {
-                Text(if (detail.overrideModel == null) "Use default ($current)" else current)
+                Text(if (detail.overrideModel == null)
+                    "$currentLabel (default)" else currentLabel)
             }
         }
         if (showModels && models != null) {
             Column(verticalArrangement = Arrangement.spacedBy(4.dp)) {
-                ModelRow(state, detail, null, models.defaultModel,
-                    "Use default (${models.defaultModel})", "")
                 models.models.forEach { m ->
                     ModelRow(state, detail, m.name, models.defaultModel,
-                        m.displayName, m.description +
+                        m.displayName + (if (m.name == models.defaultModel)
-                            (if (!m.available) " — not downloaded" else ""))
+                            " (default)" else ""),
+                        m.description +
+                            (if (!m.available) " — downloads ~once from the app" else ""))
                 }
             }
             Spacer(Modifier.height(4.dp))

backend/shonar/services/ai/model_registry.py

@@ -37,42 +37,20 @@ class TranscriptionModelInfo:
 
 
 SUPPORTED_TRANSCRIPTION_MODELS: dict[str, TranscriptionModelInfo] = {
-    "tiny": TranscriptionModelInfo(
-        name="tiny",
-        display_name="Tiny",
-        description="Fastest and lightest. Good for quick drafts and slow machines.",
-        params="~39M",
-        approx_memory="~1 GB RAM",
-        relative_speed="~10x real-time (CPU)",
-    ),
     "base": TranscriptionModelInfo(
         name="base",
-        display_name="Base (default)",
+        display_name="Whisper Base",
-        description="Balanced default. Works reasonably well on ordinary computers.",
+        description="The bundled model — works offline out of the box. "
+        "Good for clear speech on ordinary computers.",
         params="~74M",
         approx_memory="~1 GB RAM",
         relative_speed="~7x real-time (CPU)",
     ),
-    "small": TranscriptionModelInfo(
-        name="small",
-        display_name="Small",
-        description="Better accuracy with higher resource usage.",
-        params="~244M",
-        approx_memory="~2 GB RAM",
-        relative_speed="~4x real-time (CPU)",
-    ),
-    "medium": TranscriptionModelInfo(
-        name="medium",
-        display_name="Medium",
-        description="Higher accuracy and slower performance.",
-        params="~769M",
-        approx_memory="~5 GB RAM",
-        relative_speed="~2x real-time (CPU)",
-    ),
     "large-v3": TranscriptionModelInfo(
         name="large-v3",
-        display_name="Large v3",
+        display_name="Whisper Large v3",
-        description="Highest accuracy and greatest resource requirements.",
+        description="Best accuracy (accents, names, messy audio). Downloads "
+        "~3 GB once; slow on CPU — best on machines with 16 GB+ RAM.",
         params="~1.5B",
         approx_memory="~10 GB RAM",
         relative_speed="~1x real-time (CPU)",

backend/tests/test_inline_queue.py

@@ -116,7 +116,7 @@ async def test_reprocess_reruns_in_place_with_model(client, monkeypatch):
         calls_before = stt.calls
 
         r = await client.post(
-            f"/api/v1/recordings/{rec['id']}/reprocess?job=transcribe&model=small",
+            f"/api/v1/recordings/{rec['id']}/reprocess?job=transcribe&model=large-v3",
             headers=h,
         )
         assert r.status_code == 200, r.text
@@ -127,7 +127,7 @@ async def test_reprocess_reruns_in_place_with_model(client, monkeypatch):
         # Same recording id, model override persisted on the row.
         r = await client.get(f"/api/v1/recordings/{rec['id']}", headers=h)
         assert r.json()["id"] == rec["id"]
-        assert r.json()["transcription_model"] == "small"
+        assert r.json()["transcription_model"] == "large-v3"
     finally:
         await inline_queue.stop()
 

backend/tests/test_models.py

@@ -59,16 +59,16 @@ async def upload_recording(client, token, client_id, session_model=None, finaliz
     return r
 
 
-async def test_models_endpoint_lists_five(client):
+async def test_models_endpoint_lists_two(client):
     token = await user_tokens(client)
     r = await client.get("/api/v1/models", headers={"Authorization": f"Bearer {token}"})
     assert r.status_code == 200, r.text
     body = r.json()
     assert body["default_model"] == "base"
-    assert [m["name"] for m in body["models"]] == ["tiny", "base", "small", "medium", "large-v3"]
+    assert [m["name"] for m in body["models"]] == ["base", "large-v3"]
     base = next(m for m in body["models"] if m["name"] == "base")
     assert base["is_default"] is True
-    assert "ordinary computers" in base["description"]
+    assert "bundled model" in base["description"]
     for m in body["models"]:
         assert isinstance(m["downloaded"], bool)
         assert isinstance(m["available"], bool)
@@ -79,32 +79,32 @@ async def test_default_get_put_validation(client):
     h = {"Authorization": f"Bearer {token}"}
     r = await client.get("/api/v1/models/default", headers=h)
     assert r.json() == {"default_model": "base"}
-    r = await client.put("/api/v1/models/default", json={"model": "small"}, headers=h)
+    r = await client.put("/api/v1/models/default", json={"model": "large-v3"}, headers=h)
     assert r.status_code == 200, r.text
-    assert r.json() == {"default_model": "small"}
+    assert r.json() == {"default_model": "large-v3"}
     r = await client.put("/api/v1/models/default", json={"model": "xxl-turbo"}, headers=h)
     assert r.status_code == 422
     assert "Supported models" in r.text
     # rejected change did not stick
     r = await client.get("/api/v1/models/default", headers=h)
-    assert r.json() == {"default_model": "small"}
+    assert r.json() == {"default_model": "large-v3"}
 
 
 async def test_finalize_override_and_default_history(client):
     token = await user_tokens(client)
     h = {"Authorization": f"Bearer {token}"}
-    r = await upload_recording(client, token, "m-override-1", finalize_model="small")
+    r = await upload_recording(client, token, "m-override-1", finalize_model="large-v3")
     assert r.status_code == 201, r.text
-    assert r.json()["transcription_model"] == "small"
+    assert r.json()["transcription_model"] == "large-v3"
 
     r = await upload_recording(client, token, "m-default-1")
     assert r.json()["transcription_model"] == "base"
 
     # Changing the default affects future rows only.
-    r = await client.put("/api/v1/models/default", json={"model": "tiny"}, headers=h)
+    r = await client.put("/api/v1/models/default", json={"model": "large-v3"}, headers=h)
     assert r.status_code == 200
     r = await upload_recording(client, token, "m-default-2")
-    assert r.json()["transcription_model"] == "tiny"
+    assert r.json()["transcription_model"] == "large-v3"
 
     async with db_session._session_factory() as s:
         got = dict((await s.execute(
@@ -112,19 +112,20 @@ async def test_finalize_override_and_default_history(client):
             .where(Recording.client_recording_id.in_(
                 ["m-override-1", "m-default-1", "m-default-2"])))
         ).all())
-    assert got == {"m-override-1": "small", "m-default-1": "base", "m-default-2": "tiny"}
+    assert got == {"m-override-1": "large-v3", "m-default-1": "base",
+                   "m-default-2": "large-v3"}
 
 
 async def test_session_override_and_finalize_wins(client):
     token = await user_tokens(client)
-    r = await upload_recording(client, token, "m-sess-1", session_model="small")
+    r = await upload_recording(client, token, "m-sess-1", session_model="large-v3")
     assert r.status_code == 201, r.text
-    assert r.json()["transcription_model"] == "small"
+    assert r.json()["transcription_model"] == "large-v3"
 
-    r = await upload_recording(client, token, "m-sess-2", session_model="small",
+    r = await upload_recording(client, token, "m-sess-2", session_model="large-v3",
-                               finalize_model="tiny")
+                               finalize_model="base")
     assert r.status_code == 201, r.text
-    assert r.json()["transcription_model"] == "tiny"
+    assert r.json()["transcription_model"] == "base"
 
 
 async def test_finalize_invalid_override_rejected(client):
@@ -175,15 +176,15 @@ async def _run_with_spy(monkeypatch, recording_id):
 
 async def test_worker_uses_saved_model(client, monkeypatch):
     token = await user_tokens(client)
-    rec = (await upload_recording(client, token, "m-work-1", finalize_model="small")).json()
+    rec = (await upload_recording(client, token, "m-work-1", finalize_model="large-v3")).json()
     await _run_with_spy(monkeypatch, rec["id"])
     # Last construction is the worker rebuild from the saved model.
-    assert SpyTranscriber.seen_models[-1] == "small"
+    assert SpyTranscriber.seen_models[-1] == "large-v3"
 
     h = {"Authorization": f"Bearer {token}"}
     r = await client.get(f"/api/v1/recordings/{rec['id']}/transcript", headers=h)
     assert r.status_code == 200
-    assert r.json()["model"] == "small"
+    assert r.json()["model"] == "large-v3"
 
     r = await client.get(f"/api/v1/recordings/{rec['id']}/jobs", headers=h)
     job = next(j for j in r.json() if j["job_type"] == "transcribe")
@@ -197,7 +198,7 @@ async def test_worker_keeps_history_after_default_change(client, monkeypatch):
     h = {"Authorization": f"Bearer {token}"}
     rec = (await upload_recording(client, token, "m-work-2")).json()
     assert rec["transcription_model"] == "base"
-    r = await client.put("/api/v1/models/default", json={"model": "tiny"}, headers=h)
+    r = await client.put("/api/v1/models/default", json={"model": "large-v3"}, headers=h)
     assert r.status_code == 200
     await _run_with_spy(monkeypatch, rec["id"])
     assert SpyTranscriber.seen_models[-1] == "base"
@@ -229,15 +230,15 @@ async def test_unavailable_model_fails_with_instructions(client, monkeypatch):
 
 async def test_reupload_explicit_override_updates_model(client):
     token = await user_tokens(client)
-    r = await upload_recording(client, token, "m-reup-1", finalize_model="small")
+    r = await upload_recording(client, token, "m-reup-1", finalize_model="large-v3")
-    assert r.json()["transcription_model"] == "small"
+    assert r.json()["transcription_model"] == "large-v3"
     # Same client id, new explicit choice: model history moves with it.
-    r = await upload_recording(client, token, "m-reup-1", finalize_model="tiny")
+    r = await upload_recording(client, token, "m-reup-1", finalize_model="base")
     assert r.status_code == 201, r.text
-    assert r.json()["transcription_model"] == "tiny"
+    assert r.json()["transcription_model"] == "base"
     # Same client id, no choice: history untouched.
     r = await upload_recording(client, token, "m-reup-1")
-    assert r.json()["transcription_model"] == "tiny"
+    assert r.json()["transcription_model"] == "base"
 
 
 async def test_models_need_auth(client):