Fix output text not working.

docs/Decisions.md

@@ -325,3 +325,34 @@ on both engines. Three independent causes, all length-dependent:
 **Revisit:** when Transformers.js changes the seek-loop condition (currently
 `return_timestamps && !max_new_tokens`, `src/models/whisper/modeling_whisper.js`),
 retest whether the explicit `max_new_tokens` is still needed.
+
+## D-022 — Medium models never default to fp16 in the browser
+**Status:** Accepted (2026-07-15)
+Reported symptom: transcribing `en_aiproblem` in-browser with Whisper Medium
+printed just "the". The medium catalog entries defaulted to fp16 on WebNN
+(whole model) and WebGPU (encoder) — and D-017's "auto" order now tries WebNN
+first, so browsers exposing `navigator.ml` silently moved medium onto fp16.
+Two problems with those fp16 conversions
+(`onnx-community/whisper-medium(.en_timestamped|-ONNX)`):
+1. The fp16 ONNX files crash onnxruntime's graph optimizer at session
+   creation (`SimplifiedLayerNormFusion` fails on their
+   `InsertedPrecisionFreeCast_*` nodes) — reproduced on both onnxruntime-node
+   and onnxruntime-web 1.26.0-dev (wasm EP). EPs whose optimizer pipeline
+   skips that fusion can load the model and hit problem 2 instead.
+2. Whisper medium/large encoder activations are the classic fp16 overflow
+   case — degenerate output with no error raised.
+Meanwhile q4 medium was verified correct end-to-end on the real browser
+runtime (onnxruntime-web wasm EP driven under Bun via the
+`Symbol.for("onnxruntime")` override, `graphOptimizationLevel: "basic"` as in
+D-014, `max_new_tokens: 440` as in D-021; both the `.en` and multilingual
+conversions, auto language detect included). **Fix:** medium's `webgpu` and
+`webnn` defaults are q4 (`ACCEL_Q4` in `src/lib/models.ts`). A backend that
+cannot execute q4 (`MatMulNBits`) throws at pipeline creation and falls down
+the D-017 device chain — strictly better than silently emitting garbage.
+**Scope:** medium only — that's where load-failure evidence exists.
+tiny/base/small keep fp16 on WebNN and small keeps fp16+q4 on WebGPU;
+large-v3-turbo keeps fp16+q4 on WebGPU (the widely-used Transformers.js demo
+configuration). If a one-word/garbage transcript ever shows up on those, suspect
+their fp16 encoders and extend this decision.
+**Revisit:** if the onnx-community medium conversions are re-exported with
+working fp16 files, or once WebNN/WebGPU runs can be exercised in CI.

src/lib/models.ts

@@ -128,6 +128,23 @@ const WEBNN_FP16: DtypeDefault = {
   note: "fp16 is commonly used with WebNN.",
 };
 
+/**
+ * The medium fp16 conversions are unusable (D-022, verified 2026-07-15): the
+ * ONNX files crash onnxruntime's graph optimizer (SimplifiedLayerNormFusion
+ * chokes on their InsertedPrecisionFreeCast nodes — reproduced on both
+ * ort-node and ort-web 1.26.0-dev), and Whisper medium/large encoder
+ * activations are the classic fp16 overflow case besides. In the browser the
+ * symptom was a degenerate one-word transcript with no error. q4 is verified
+ * correct for medium on the WASM runtime; a backend that cannot run it
+ * throws at session creation and falls down the D-017 device chain instead
+ * of silently producing garbage.
+ */
+const ACCEL_Q4: DtypeDefault = {
+  dtype: "q4",
+  verified: false,
+  note: "q4: the medium fp16 conversions fail to load / overflow (D-022).",
+};
+
 // ---------------------------------------------------------------------------
 // The catalog (D-006: curated, not exhaustive)
 // ---------------------------------------------------------------------------
@@ -203,7 +220,7 @@ export const MODELS: readonly WhisperModel[] = [
     multilingual: false,
     encoderParamsM: 306,
     decoderParamsM: 463,
-    defaults: { wasm: WASM_Q4, webgpu: WEBGPU_FP16_Q4, webnn: WEBNN_FP16 },
+    defaults: { wasm: WASM_Q4, webgpu: ACCEL_Q4, webnn: ACCEL_Q4 },
   },
   {
     id: "onnx-community/whisper-medium-ONNX",
@@ -212,7 +229,7 @@ export const MODELS: readonly WhisperModel[] = [
     multilingual: true,
     encoderParamsM: 306,
     decoderParamsM: 463,
-    defaults: { wasm: WASM_Q4, webgpu: WEBGPU_FP16_Q4, webnn: WEBNN_FP16 },
+    defaults: { wasm: WASM_Q4, webgpu: ACCEL_Q4, webnn: ACCEL_Q4 },
   },
   {
     id: "onnx-community/whisper-large-v3-turbo",

tests/models.test.ts

@@ -107,6 +107,17 @@ describe("resolveDefaultDtype", () => {
     ).toEqual({ encoder_model: "fp16", decoder_model_merged: "q4" });
   });
 
+  test("medium never defaults to fp16 on any backend (D-022: broken conversions)", () => {
+    for (const id of [
+      "onnx-community/whisper-medium.en_timestamped",
+      "onnx-community/whisper-medium-ONNX",
+    ]) {
+      for (const device of ["wasm", "webgpu", "webnn"] as const) {
+        expect(resolveDefaultDtype(id, device)).toBe("q4");
+      }
+    }
+  });
+
   test("unknown model ids fall back to q8", () => {
     expect(resolveDefaultDtype("someone/some-model", "wasm")).toBe("q8");
   });