From b9f33749d2fc9678dc403ebe4263efbc67b951c3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=81=A5=E4=BB=99?= Date: Thu, 17 Sep 2026 14:08:15 +0800 Subject: [PATCH] feat(model/omni): name change input video representation_compact parameter --- .../dashscope/audio/omni/OmniRealtimeConfig.java | 13 +++++++------ .../dashscope/audio/omni/OmniRealtimeConstants.java | 4 ++-- 2 files changed, 9 insertions(+), 8 deletions(-) diff --git a/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConfig.java b/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConfig.java index 504f036..f5fc36f 100644 --- a/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConfig.java +++ b/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConfig.java @@ -51,10 +51,12 @@ public class OmniRealtimeConfig { */ @Builder.Default OmniRealtimeAudioFormatConfig outputAudio = null; /** - * Video input representation compactness. When set, it is serialized as {@code - * video.input.representation_compactness}. + * Video input representation aggregation mode, serialized as {@code + * video.input.representation_compact}. {@code none} preserves fine-grained video input + * representations, while {@code normal} aggregates them to reduce computational overhead when + * visual details are less important. */ - @Builder.Default Integer inputVideoRepresentationCompactness = null; + @Builder.Default String inputVideoRepresentationCompact = null; /** enable transcription for input audio */ @Builder.Default boolean enableInputAudioTranscription = true; /** model used for input audio transcription */ @@ -111,10 +113,9 @@ public JsonObject getConfig() { config.put(OmniRealtimeConstants.INPUT_AUDIO_FORMAT, inputAudioFormat); config.put(OmniRealtimeConstants.OUTPUT_AUDIO_FORMAT, outputAudioFormat); } - if (inputVideoRepresentationCompactness != null) { + if (inputVideoRepresentationCompact != null) { Map videoInput = new HashMap<>(); - videoInput.put( - OmniRealtimeConstants.REPRESENTATION_COMPACTNESS, inputVideoRepresentationCompactness); + videoInput.put(OmniRealtimeConstants.REPRESENTATION_COMPACT, inputVideoRepresentationCompact); Map video = new HashMap<>(); video.put(OmniRealtimeConstants.VIDEO_INPUT, videoInput); config.put(OmniRealtimeConstants.VIDEO, video); diff --git a/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConstants.java b/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConstants.java index e8d2fc2..686dfa3 100644 --- a/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConstants.java +++ b/src/main/java/com/alibaba/dashscope/audio/omni/OmniRealtimeConstants.java @@ -17,10 +17,10 @@ public class OmniRealtimeConstants { public static final String AUDIO_FORMAT = "format"; public static final String AUDIO_FORMAT_TYPE = "type"; // Video input configuration, see session.update: - // { "video": { "input": { "representation_compactness": 1 } } } + // { "video": { "input": { "representation_compact": "none" } } } public static final String VIDEO = "video"; public static final String VIDEO_INPUT = "input"; - public static final String REPRESENTATION_COMPACTNESS = "representation_compactness"; + public static final String REPRESENTATION_COMPACT = "representation_compact"; public static final String INPUT_AUDIO_TRANSCRIPTION = "input_audio_transcription"; public static final String INPUT_AUDIO_TRANSCRIPTION_MODEL = "model"; public static final String INPUT_AUDIO_TRANSCRIPTION_CORPUS = "corpus";