The spatial audio should be relative to the camera.
If you consider it, even in a 3D game, the audio is relative to the camera. Sounds that come from the right side of the screen are heard from the right speaker, and sounds that come from the left side of the screen are heard from the left speaker. It just happens that the avatar is looking in the same direction as the camera. (For example, an FPS with a cutscene that shows your avatar doesn't switch the direction of the audio when viewing your character's front side).
Remember that the reason you're implementing spatial audio is for the player. It would be rather unintuitive to hear a sound coming from your speakers and need to reference the direction the avatar is looking to determine which direction they heard it from. The audio is for the human playing the game, not the avatar in the game. With FPS style games, when in first person mode, audio is
The only reason you might implement spatial audio relative to the avatar instead of being relative to the player is if it was specific a game mechanic your game relied on.