YOUR NEXT STEP

Scenic Avatar

Instructions for the upcoming NeuralCompanion interface.

Scenic is a lightweight avatar route that selects a still image from a tagged pack. It does not generate lip-sync frames or require MuseTalk/Ditto models. Speech remains the responsibility of the selected text-to-speech provider.

Select a pack

  1. Enable Scenic Avatar in Addons if needed, then restart when requested.
  2. Open Runtime → Runtime Session → Avatar → Scenic.
  3. Choose Refresh packs, select a discovered pack, and inspect Pack availability, the preview, and its image/tag details.
  4. Choose Use this avatar, initialize NC, and test a short reply.

Selecting a pack and activating the avatar route are separate steps. An unavailable saved pack is shown as unavailable; selecting its name does not restore missing files.

Pack format

Keep each pack under ScenicPacks/<pack_id>/, with scenic_pack.json directly inside its folder. For example:

ScenicPacks/MyScenic/
  scenic_pack.json
  images/neutral.png
  images/happy.png
{
  "engine": "Scenic",
  "pack_name": "My Scenic",
  "images": [
    {"tag": "neutral", "image_path": "images/neutral.png"},
    {"tag": "happy", "image_path": "images/happy.png"}
  ]
}

Use pack-relative paths and supported PNG, JPG/JPEG, WEBP, or BMP images. Keep the complete folder together when moving or backing up the pack. Refresh the catalog after adding it. The current native settings surface selects and previews packs; it is not a general pack editor.

Reply emotion tags select corresponding images. When a requested tag has no matching image, the pack falls back to neutral, then its first image. Include a neutral image and give the persona the avatar mood tag instructions. A still image during speech is expected for Scenic.

Troubleshooting

  • No packs: check the exact folder and manifest layout, then Refresh packs.
  • Pack listed but no image: check its images entries and referenced files.
  • Image does not change: check the reply’s mood tag and whether the pack has that tag; do not diagnose this as a TTS or MuseTalk problem.
  • No speech: test the selected TTS provider separately with Avatar None.

See Avatars for activation and preview, and Avatar packs for the other pack formats. The format and fallback behavior are implemented in pack_runtime.py.