diff --git a/docs/fern/pages/recipes/cli-templates/vllm.mdx b/docs/fern/pages/recipes/cli-templates/vllm.mdx
index 003b322194c6..1fa09a95add8 100644
--- a/docs/fern/pages/recipes/cli-templates/vllm.mdx
+++ b/docs/fern/pages/recipes/cli-templates/vllm.mdx
@@ -48,16 +48,38 @@ The default model is `Qwen/Qwen3-0.6B`.
 </Accordion>
 <Accordion title="Qwen3-ASR Realtime Transcription">
 
-**Experimental.** Transcribe English audio over `/v1/realtime` using vLLM 0.30.0's native
-Qwen3-ASR support. The launch script requires an explicit `--hf-overrides` argument:
+**Experimental.** Transcribe audio over `/v1/realtime` using vLLM's native Qwen3-ASR
+support. This example was validated with English audio.
+
+The default model is `Qwen/Qwen3-ASR-0.6B`. The launch script selects its
+`Qwen3ASRRealtimeGeneration` architecture automatically.
+
+Follow the [vLLM installation instructions](../../developer-guide/knowledge-base/modular-components/backends/vllm/overview.md#installation)
+to use Dynamo's compatible backend version.
+
+**Command-line flags:** `--model` sets the model identifier or path. `--hf-overrides`
+selects the realtime architecture; it is optional for the default model and required
+for other models.
+
+**Pass-through arguments:** Additional unrecognized arguments are forwarded to the backend worker command.
+
+Start the default model:
+
+```bash
+./launch/agg_realtime_transcription.sh --max-model-len 4096 --max-num-seqs 2
+```
+
+For another realtime-capable model, replace `MODEL_ID` and `REALTIME_ARCHITECTURE`
+with its model identifier and realtime architecture:
 
 ```bash
-./launch/agg_realtime_transcription.sh --model Qwen/Qwen3-ASR-0.6B \
-    --hf-overrides '{"architectures":["Qwen3ASRRealtimeGeneration"]}' \
-    --max-model-len 4096 --max-num-seqs 2
+./launch/agg_realtime_transcription.sh --model MODEL_ID \
+    --hf-overrides '{"architectures":["REALTIME_ARCHITECTURE"]}'
 ```
 
-In another terminal, send a speech file longer than five seconds to check incremental delivery:
+In another terminal, send a speech file longer than five seconds to check incremental
+delivery from the default model. Replace `speech.wav` with your audio file and set
+`--model` to the served model if you changed it:
 
 ```bash
 python launch/realtime_audio_client.py \
diff --git a/docs/fern/pages/use-cases/multimodal-serving/voice.md b/docs/fern/pages/use-cases/multimodal-serving/voice.md
index dd699edff778..227aa4e1aa1d 100644
--- a/docs/fern/pages/use-cases/multimodal-serving/voice.md
+++ b/docs/fern/pages/use-cases/multimodal-serving/voice.md
@@ -27,7 +27,7 @@ for a cascaded pipeline, or a vLLM-Omni model for audio input and output.
   Follow the example's runtime-image requirements; it currently requires
   compatible source-built Dynamo images, not the published 1.5.0 images.
 - **Separate vLLM/vLLM-Omni speech workers:** Use the
-  [Voxtral realtime transcription launcher](https://github.com/ai-dynamo/dynamo/blob/main/examples/backends/vllm/launch/agg_realtime_transcription.sh)
+  [Qwen3-ASR realtime transcription launcher](https://github.com/ai-dynamo/dynamo/blob/main/examples/backends/vllm/launch/agg_realtime_transcription.sh)
   and [Qwen3-TTS launcher](https://github.com/ai-dynamo/dynamo/blob/main/examples/backends/vllm/launch/agg_omni_audio.sh)
   to exercise each endpoint independently before connecting them to an LLM.
 - **Audio-in/audio-out with vLLM-Omni:** Run the
diff --git a/examples/backends/vllm/launch/agg_realtime_transcription.sh b/examples/backends/vllm/launch/agg_realtime_transcription.sh
index 060da9ce52b6..5091b10dbc6a 100755
--- a/examples/backends/vllm/launch/agg_realtime_transcription.sh
+++ b/examples/backends/vllm/launch/agg_realtime_transcription.sh
@@ -8,8 +8,10 @@ SCRIPT_DIR="$(dirname "$(readlink -f "$0")")"
 source "$SCRIPT_DIR/../../../common/gpu_utils.sh"
 source "$SCRIPT_DIR/../../../common/launch_utils.sh"
 
-MODEL="mistralai/Voxtral-Mini-4B-Realtime-2602"
+DEFAULT_MODEL="Qwen/Qwen3-ASR-0.6B"
+MODEL="$DEFAULT_MODEL"
 HF_OVERRIDES=""
+HF_OVERRIDES_SET=false
 
 EXTRA_ARGS=()
 while [[ $# -gt 0 ]]; do
@@ -24,10 +26,12 @@ while [[ $# -gt 0 ]]; do
                 exit 2
             fi
             HF_OVERRIDES="$2"
+            HF_OVERRIDES_SET=true
             shift 2
             ;;
         --hf-overrides=*)
             HF_OVERRIDES="${1#*=}"
+            HF_OVERRIDES_SET=true
             shift
             ;;
         *)
@@ -37,9 +41,13 @@ while [[ $# -gt 0 ]]; do
     esac
 done
 
+if [[ "$HF_OVERRIDES_SET" == false && "$MODEL" == "$DEFAULT_MODEL" ]]; then
+    HF_OVERRIDES='{"architectures":["Qwen3ASRRealtimeGeneration"]}'
+fi
+
 if [[ -z "$HF_OVERRIDES" ]]; then
     echo "Error: select the model's realtime architecture explicitly with --hf-overrides." >&2
-    echo "Usage: $0 [--model MODEL] --hf-overrides JSON [vLLM args...]" >&2
+    echo "Usage: $0 [--model MODEL] [--hf-overrides JSON] [vLLM args...]" >&2
     exit 2
 fi