> This page is for NVIDIA Studio Voice on Holoscan for Media, version 1.2.0 (Latest) (default).
> For other versions, use one of these documentation indexes:
> - 1.2.0 (Latest) (default): https://docs.nvidia.com/nim/maxine/studio-voice-h4m/1.2.0/llms.txt

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nim/maxine/llms.txt.

# Pipeline Configuration

This page covers NIM service chart Helm values and environment variables specific to the Studio Voice pipeline, including pipeline flags and SMPTE ST 2110 session wiring.

<Note>
Only the 48 kHz Low-Latency variant is shipped with the Studio Voice NIM. Other Studio Voice variants (Quality 48 kHz, Quality 16 kHz) are part of the gRPC NIM and are not currently exposed in this release.
</Note>

---

## Studio Voice Pipeline Flags

| Configuration | Helm Key | Environment Variable | Comment |
|---------------|----------|----------------------|---------|
| NMOS pipeline front-end | `studiovoice.nmosEnabled` | `STUDIOVOICE_NMOS_ENABLED` | Default: `true`; controls whether the pipeline uses the NMOS front-end (`1`) or SMPTE ST 2110 static SDP front-end (`0`). Keep in sync with `nmos.enabled`. |
| NIM inference | `studiovoice.nimEnabled` | `STUDIOVOICE_NIM_ENABLED` | Default: `true`; set `false` for passthrough or debug. This flag is *runtime-only* and does not relax the pod's resource requests; the pod still reserves `nvidia.com/gpu: 1` and an SR-IOV VF from the configured media pool even when inference is disabled. |
| CPU affinity range | `env.CPU_AFFINITY` | `CPU_AFFINITY` | Default: `"0-1"` (2 cores). One inclusive range `"start-end"` or a single CPU index `"n"`; comma-separated lists and multiple ranges are not supported. Chart-only; the operator does not expose a CR field for this. |

---

## Audio Format

A single audio block drives both the SMPTE ST 2110-30 receiver or sender format and the Studio Voice inference input rate, so one value applies to both the network audio session and the model.

| Configuration | Helm Key | Environment Variable | Comment |
|---------------|----------|----------------------|---------|
| PCM format | `audio.pcmFormat` | `STUDIOVOICE_AUDIO_PCM_FORMAT` | Default: `S24BE` (24-bit signed big-endian PCM; matches ST 2110-30 `audio/L24`). |
| Sampling rate | `audio.samplingRate` | `STUDIOVOICE_AUDIO_SAMPLING_RATE` | Fixed at `48000` Hz. The shipped model is 48 kHz only; other values produce a broken pipeline and are rejected at admission by the operator CRD (`enum: [48000]`). |
| Channel count | `audio.numChannels` | `STUDIOVOICE_AUDIO_NUM_CHANNELS` | Default: `1` (mono); Studio Voice expects a single mono speech stream. |

---

## SMPTE ST 2110 (Static Mode) Input Session

| Configuration | Helm Key | Environment Variable | Comment |
|---------------|----------|----------------------|---------|
| Audio session name | `input.audio.sessionName` | `STUDIOVOICE_INPUT_AUDIO_SESSION_NAME` | Default: `audio_in`. |
| Interface | `input.audio.localInterfaceName` | `STUDIOVOICE_INPUT_AUDIO_LOCAL_INTERFACE_NAME` | Default: `net1`. |
| Multicast IP address | `input.audio.hostIp` | `STUDIOVOICE_INPUT_AUDIO_HOST_IP` | Site-specific; default: `234.5.8.9`. |
| Port | `input.audio.hostPort` | `STUDIOVOICE_INPUT_AUDIO_HOST_PORT` | Site-specific; default: `5003`. |
| Subnet bits | `input.audio.hostNumSubnetBits` | `STUDIOVOICE_INPUT_AUDIO_HOST_NUM_SUBNET_BITS` | Default: `24`. |

---

## SMPTE ST 2110 (Static Mode) Output Session

| Configuration | Helm Key | Environment Variable | Comment |
|---------------|----------|----------------------|---------|
| Audio session name | `output.audio.sessionName` | `STUDIOVOICE_OUTPUT_AUDIO_SESSION_NAME` | Default: `audio_out`. |
| Interface | `output.audio.localInterfaceName` | `STUDIOVOICE_OUTPUT_AUDIO_LOCAL_INTERFACE_NAME` | Default: `net1`. |
| Multicast IP address | `output.audio.hostIp` | `STUDIOVOICE_OUTPUT_AUDIO_HOST_IP` | Site-specific; default: `234.5.8.9`. |
| Port | `output.audio.hostPort` | `STUDIOVOICE_OUTPUT_AUDIO_HOST_PORT` | Site-specific; default: `5004`. |
| Subnet bits | `output.audio.hostNumSubnetBits` | `STUDIOVOICE_OUTPUT_AUDIO_HOST_NUM_SUBNET_BITS` | Default: `24`. |

---

## NIM Pod Resources

| Resource | Requests | Limits |
|----------|----------|--------|
| `cpu` | `2` | `4` |
| `hugepages-2Mi` | `256Mi` | `256Mi` |
| `memory` | `2Gi` | `4Gi` |
| `nvidia.com/gpu` | `1` | `1` |

The default `cpu` request (`2`) and limit (`4`) are tuned for real-time ST 2110 audio processing and must stay consistent with `env.CPU_AFFINITY`. The chart validates at render time that `resources.requests.cpu` and `resources.limits.cpu` are each equal to or greater than the number of cores in the affinity range. If `cpu` falls below the affinity range, `helm install` / `helm upgrade` fails with a descriptive error. To use a wider pinned range (for example, `env.CPU_AFFINITY: "0-3"`), raise `resources.requests.cpu` and `resources.limits.cpu` to at least `4`. For the affinity field, refer to [Studio Voice Pipeline Flags](#studio-voice-pipeline-flags).

---

## Logging

| Configuration | Helm Key | Environment Variable | Comment |
|---------------|----------|----------------------|---------|
| Log level | `logging.level` | `STUDIOVOICE_LOGGING_LEVEL` | `0` FATAL, `1` ERROR, `2` WARN, `3` INFO (default), `4` DEBUG, `5` VERBOSE. |
| GStreamer debug | `gstDebug` | `GST_DEBUG` | Default: `2,nvdsnmosbin:8,nvnmos:2,nvdssdp*:8`. |

---

## See Also

- [Configuration Reference](/nim/maxine/studio-voice-h4m/configuration)
- [Common Helm Configuration](/nim/maxine/studio-voice-h4m/configuration/common) — Cluster, NGC, PVCs, and NMOS.
- [Operator Configuration](/nim/maxine/studio-voice-h4m/configuration/operator) — Operator CR pipeline tuning equivalent.
- [Known Limitations](/nim/maxine/studio-voice-h4m/limitations)