Listening with context.

Open Synapse

1. Open Synapse and sign in with your PARALLAX Account.
2. Connect a supported music service when requested.
3. Choose music directly or open Synapse Q for a guided session.

Install the Synapse DJ engine

Synapse Q needs the packaged local language model and voice engine to deliver its complete DJ experience. Install both together from the Q tab when prompted. Synapse verifies the package, registers the local runtime and reports when the engine is ready. Keep the model and voice package in their managed location so updates and integrity checks can work.

What happens when Q opens

Q begins with a short greeting while the session is assembled. It then introduces the first track before music starts. The greeting may connect naturally to the first song, but track, artist and session details remain grounded in verified playback metadata.

Local engine priority

When the Synapse DJ engine is installed and healthy, Q uses it before every cloud provider. Deterministic session code controls playback, timing, skipping and exact song identity; the local model selects tone and structure from that grounded context. Optional Groq or Gemini enhancement is reserved for occasional richer moments and cannot replace routine local narration.

Prepared transitions

Five seconds into the current song, Q begins preparing and caching narration for the upcoming track. This keeps transitions ready before playback changes. Curated narration remains available when model enhancement is slow, while verified metadata prevents invented artist or track facts.

Manual skips

A manual skip cancels speech belonging to the previous track, advances the authoritative session state and uses or regenerates narration for the song that actually became current. Repeated skips must not leave an earlier track introduction pinned on screen or permanently disable later narration.

Mid-session takeovers

Q may occasionally step in between ordinary transitions with a short, relevant observation about the listening arc, an artist or a verified music fact. These moments are scheduled by the session engine, never over active speech, and never by interrupting playback without a prepared handoff.

Narration quality boundaries

Q should speak only from the current track, upcoming queue, session history and verified metadata. It should not output prompt labels, JSON, role names, quotation scaffolding or invented listener statements. If malformed model text appears, stop the session, capture the visible output and report it through Support.

When Q cannot speak

Confirm that both parts of the Synapse DJ engine show as installed, then restart Synapse. Check the selected audio output and ensure device security software has not quarantined the runtime. A cloud quota warning should not disable a healthy local engine. Include the Synapse version, engine version, failing track, whether the failure followed a manual skip and the exact visible error when requesting help.

Refresh and recovery

Refreshing Q should preserve current playback until a complete new session and its opening narration are ready. If recovery fails, Synapse should leave the existing session intact and explain which engine layer could not start. Use the troubleshooting guide for the diagnostic order.