Gemini: add media resolution select (#4775)

5accfbc5d8566eae5aaeaa8b9f833d518e9893d0

Cohee <18619528+Cohee1207@users.noreply.github.com>

Signed

public/index.html+1 -1 Viewed

                                             <strong data-source="makersuite,vertexai" data-i18n="video_inlining_hint_4">Videos must be less than 20 MB and under 1 minute long.</strong>
                                             <strong data-source="makersuite,vertexai" data-i18n="audio_inlining_hint_2">Audio must be less than 20 MB.</strong>
                                         </div>
                                         <div class="flex-container flexFlowColumn wide100p textAlignCenter marginTop10" data-source="openai,custom,xai,pollinations,cohere,cometapi,nanogpt,moonshot,aimlapi,openrouter,mistralai,electronhub,azure_openai,zai,siliconflow,chutes,makersuite,vertexai">
                                             <div class="flex-container oneline-dropdown">
                                                 <label for="openai_inline_image_quality" data-i18n="Inline Image Quality">
                                                     Inline Image Quality

public/scripts/openai.js+2 -1 Viewed

+        }
         // Note: No compression for videos (unlike images)
 -        this.content.push({ type: 'video_url', video_url: { 'url': video } });
 +        const quality = oai_settings.inline_image_quality || default_settings.inline_image_quality;
 +        this.content.push({ type: 'video_url', video_url: { 'url': video, 'detail': quality } });
         try {
             // Using Gemini calculation (263 tokens per second)

src/prompt-converters.js+24 -6 Viewed

     SINGLE: 'single',
 };
 +// 'auto' is intentionally unmapped
 +const GEMINI_MEDIA_RESOLUTION = {
 +    low: 'media_resolution_low',
 +    high: 'media_resolution_high',
 +};
++
 /**
  * @typedef {object} PromptNames
  * @property {string} charName Character name

         //create the prompt parts
         const parts = [];
         message.content.forEach((part) => {
             const addDataUrlPart = (/** @type {string} */ url, /** @type {string} */ defaultMimeType, /** @type {string?} */ detail = null) => {
                 if (url && url.startsWith('data:')) {
                     const [header, base64Data] = url.split(',');
                     const mimeType = header.match(/data:([^;]+)/)?.[1] || defaultMimeType;
 +                    const mediaResolution = GEMINI_MEDIA_RESOLUTION[detail] || null;
 -                    parts.push({
 +                    const part = {
                         inlineData: {
                             mimeType: mimeType,
                             data: base64Data,
                         },
                     });
++
 +                    // https://ai.google.dev/gemini-api/docs/gemini-3#media_resolution
 +                    if (/gemini-3/.test(model) && mediaResolution) {
 +                        part.mediaResolution = {
 +                            level: mediaResolution,
 +                        };
 +                    }
++
 +                    parts.push(part);
+                }
             };

                 });
             } else if (part.type === 'image_url') {
                 const imageUrl = part.image_url?.url;
 -                addDataUrlPart(imageUrl, 'image/png');
 +                const detail = part.image_url?.detail;
 +                addDataUrlPart(imageUrl, 'image/png', detail);
             } else if (part.type === 'video_url') {
                 const videoUrl = part.video_url?.url;
 -                addDataUrlPart(videoUrl, 'video/mp4');
 +                const detail = part.video_url?.detail;
 +                addDataUrlPart(videoUrl, 'video/mp4', detail);
             } else if (part.type === 'audio_url') {
                 const audioUrl = part.audio_url?.url;
                 addDataUrlPart(audioUrl, 'audio/mpeg');

@@ -572,7 +590,7 @@ export function convertGooglePrompt(messages, model, useSysPrompt, names) {
572	590	contents[contents.length - 1].parts.push(part);
573	591	}
574	592	}
575	593	if (part.inlineData \|\| part.functionCall \|\| part.functionResponse \|\| part.thoughtSignature \|\| part.mediaResolution) {
576	594	contents[contents.length - 1].parts.push(part);
577	595	}
578	596	});