diff --git a/.changeset/desktop-stream-typecheck.md b/.changeset/desktop-stream-typecheck.md
new file mode 100644
index 0000000000..071987baa2
--- /dev/null
+++ b/.changeset/desktop-stream-typecheck.md
@@ -0,0 +1,6 @@
+---
+"@runfusion/fusion": patch
+---
+summary: Desktop builds now typecheck streamed speech-model downloads across differing ReadableStream library definitions.
+category: fix
+dev: Casts the fetch response body through unknown before treating it as an async byte iterable, preserving runtime behavior while satisfying the desktop TypeScript library surface.
diff --git a/packages/dashboard/src/stt/model-manager.ts b/packages/dashboard/src/stt/model-manager.ts
index a38b5180ea..19b308ef68 100644
--- a/packages/dashboard/src/stt/model-manager.ts
+++ b/packages/dashboard/src/stt/model-manager.ts
@@ -120,7 +120,7 @@ export function createVoiceModelManager(options: VoiceModelManagerOptions = {}):
         if (!current(g, signal)) return state;
         const total = Number(response.headers.get("content-length")) || undefined;
         const hash = createHash("sha256"); let bytes = 0;
-        for await (const chunk of response.body as AsyncIterable<Uint8Array>) {
+        for await (const chunk of response.body as unknown as AsyncIterable<Uint8Array>) {
           if (!current(g, signal)) return state;
           const buffer = Buffer.from(chunk); hash.update(buffer); await file.write(buffer);
           if (!current(g, signal)) return state;