@@ -51,10 +51,12 @@ public class OmniRealtimeConfig {
5151 */
5252 @ Builder .Default OmniRealtimeAudioFormatConfig outputAudio = null ;
5353 /**
54- * Video input representation compactness. When set, it is serialized as {@code
55- * video.input.representation_compactness}.
54+ * Video input representation aggregation mode, serialized as {@code
55+ * video.input.representation_compact}. {@code none} preserves fine-grained video input
56+ * representations, while {@code normal} aggregates them to reduce computational overhead when
57+ * visual details are less important.
5658 */
57- @ Builder .Default Integer inputVideoRepresentationCompactness = null ;
59+ @ Builder .Default String inputVideoRepresentationCompact = null ;
5860 /** enable transcription for input audio */
5961 @ Builder .Default boolean enableInputAudioTranscription = true ;
6062 /** model used for input audio transcription */
@@ -111,10 +113,9 @@ public JsonObject getConfig() {
111113 config .put (OmniRealtimeConstants .INPUT_AUDIO_FORMAT , inputAudioFormat );
112114 config .put (OmniRealtimeConstants .OUTPUT_AUDIO_FORMAT , outputAudioFormat );
113115 }
114- if (inputVideoRepresentationCompactness != null ) {
116+ if (inputVideoRepresentationCompact != null ) {
115117 Map <String , Object > videoInput = new HashMap <>();
116- videoInput .put (
117- OmniRealtimeConstants .REPRESENTATION_COMPACTNESS , inputVideoRepresentationCompactness );
118+ videoInput .put (OmniRealtimeConstants .REPRESENTATION_COMPACT , inputVideoRepresentationCompact );
118119 Map <String , Object > video = new HashMap <>();
119120 video .put (OmniRealtimeConstants .VIDEO_INPUT , videoInput );
120121 config .put (OmniRealtimeConstants .VIDEO , video );
0 commit comments