Added new models and modes: GPT Image 2.5, Wan, Output Formats (#1954)

2026-10-20 17:02:53 -04:00 · 2026-09-18 16:23:44 -04:00
parent 584fdb7044
commit 3c5038c159
58 changed files with 883 additions and 78 deletions
@@ -12,12 +12,19 @@ export enum CommonImageModel {
  GptImage2 = "gpt_image_2",
  GptImage2p5Flare = "gpt_image_2p5_flare",
  GptImage2p5Sunburst = "gpt_image_2p5_sunburst",
+  GrokImagineImage = "grok_imagine_image",
+  GrokImagineImageQuality = "grok_imagine_image_q",
+  Midjourney7 = "midjourney_7",
+  Midjourney7Niji = "midjourney_7_niji",
+  Midjourney8 = "midjourney_8",
  NanaBanana = "nano_banana",
  NanaBanana2 = "nano_banana_2",
  NanaBananaPro = "nano_banana_pro",
  Seedream4 = "seedream_4",
  Seedream4p5 = "seedream_4p5",
  Seedream5Lite = "seedream_5_lite",
+  Seedream5p0Pro = "seedream_5p0_pro",
+  Seedream5p0ProUltra = "seedream_5p0_pro_u",
  QwenEdit2511Angles = "qwen_edit_2511_angles",
  Flux2LoraAngles = "flux_2_lora_angles",
 }
@@ -3,6 +3,9 @@
 // In the future, we should use code gen (protobufs or similar) to keep the two sides in sync.

 export enum CommonQuality {
+  Auto = "auto",
+  Max = "max",
+  XHigh = "xhigh",
  High = "high",
  Medium = "medium",
  Low = "low",
@@ -4,6 +4,10 @@

 export enum CommonVideoModel {
  GrokVideo = "grok_video",
+  GrokImagineVideo = "grok_imagine_video",
+  GrokImagineVideo1p5 = "grok_imagine_video_1p5",
+  Flux3 = "flux_3",
+  Flux3Draft = "flux_3_draft",
  Kling16Pro = "kling_1p6_pro",
  Kling21Pro = "kling_2p1_pro",
  Kling21Master = "kling_2p1_master",
@@ -11,10 +15,28 @@ export enum CommonVideoModel {
  Kling2p6Pro = "kling_2p6_pro",
  Kling3p0Standard = "kling_3p0_standard",
  Kling3p0Pro = "kling_3p0_pro",
+  HappyHorse1p0 = "happy_horse_1p0",
+  Wan3p0 = "wan_3p0",
+  Wan3p0Prime = "wan_3p0_prime",
+  MinimaxH3 = "minimax_h3",
+  MinimaxH3Turbo = "minimax_h3_turbo",
+  MinimaxH3Ultra = "minimax_h3_ultra",
  Seedance10Lite = "seedance_1p0_lite",
  Seedance1p5Pro = "seedance_1p5_pro",
  Seedance2p0 = "seedance_2p0",
  Seedance2p0Fast = "seedance_2p0_fast",
+  Seedance2p0BytePlus = "seedance_2p0_bp",
+  Seedance2p0BytePlusFast = "seedance_2p0_bp_fast",
+  Seedance2p0Ultra = "seedance_2p0_u",
+  Seedance2p0UltraFast = "seedance_2p0_u_fast",
+  Seedance2p0BytePlusUltra = "seedance_2p0_bpu",
+  Seedance2p0BytePlusUltraFast = "seedance_2p0_bpu_fast",
+  Seedance2p0Mini = "seedance_2p0_mini",
+  Seedance2p0BytePlusMini = "seedance_2p0_bp_mini",
+  Seedance2p0BytePlusUltraMini = "seedance_2p0_bpu_mini",
+  Seedance2p5Preview = "seedance_2p5_preview",
+  Seedance2p5 = "seedance_2p5",
+  Seedance2p5Ultra = "seedance_2p5_u",
  Sora2 = "sora_2",
  Sora2Pro = "sora_2_pro",
  Veo2 = "veo_2",
@@ -22,4 +44,9 @@ export enum CommonVideoModel {
  Veo3Fast = "veo_3_fast",
  Veo3p1 = "veo_3p1",
  Veo3p1Fast = "veo_3p1_fast",
+  Veo3p1Lite = "veo_3p1_lite",
+  ViduQ3 = "vidu_q3",
+  ViduQ3Turbo = "vidu_q3_turbo",
+  PreviewModel = "preview_model",
+  PreviewModelFast = "preview_model_fast",
 }
@@ -3,12 +3,18 @@ import { CommonQuality } from "@storyteller/api-enums";
 /**
 * Convert a quality value to CommonQuality for the image cost estimate API.
 *
- * Accepts a CommonQuality string value from @storyteller/model-list ("high", "medium", "low").
+ * Preserves all API quality tiers, including values added after this build.
 */
 export function stringToCommonQuality(
  quality: string | undefined,
 ): string | null {
  switch (quality) {
+    case "auto":
+      return CommonQuality.Auto;
+    case "max":
+      return CommonQuality.Max;
+    case "xhigh":
+      return CommonQuality.XHigh;
    case "high":
      return CommonQuality.High;
    case "medium":
@@ -78,7 +78,7 @@ export function useImageCostEstimate(
        // Match GenerateImage exactly; omitted values use the API defaults.
        aspectRatioStr = imageModel.supportsNewAspectRatio() ? imageAspectRatio : undefined;
        resolutionStr = imageModel.supportsNewResolution() ? imageResolution : undefined;
-        qualityStr = imageModel.supportsQuality() ? imageQuality ?? imageModel.defaultQuality : undefined;
+        qualityStr = imageModel.resolveQuality(imageQuality);
        referenceImageCount = imageModel.canUseImagePrompt
          ? imageReferenceImages.filter((image) => image.mediaToken.length > 0).length : 0;
        generationCount = imageGenerationCount;
@@ -17,6 +17,7 @@ const mocks = vi.hoisted(() => ({
    referenceVideos: [{ mediaToken: "mf_video", duration: 7.25 }],
    referenceAudios: [{ mediaToken: "mf_audio", duration: 3.5 }],
    bitrate: "high",
+    outputFormat: "mov",
    generationCount: 1,
    generateWithSound: true,
  },
@@ -38,6 +39,7 @@ vi.mock("./cost-breakdown-modal-store", () => ({
 }));

 beforeEach(() => {
+  mocks.state.generationCount = 1;
  mocks.estimate.mockReset();
  mocks.setCredits.mockReset();
  mocks.estimate.mockResolvedValue({
@@ -61,6 +63,7 @@ it("quotes a future model with the chosen bitrate and reference durations in mil
      model: "future_video",
      resolution: "four_k",
      bitrate: "high",
+      output_format: "mov",
      duration_seconds: 30,
      reference_video_media_tokens: ["mf_video"],
      reference_audio_media_tokens: ["mf_audio"],
@@ -72,6 +75,20 @@ it("quotes a future model with the chosen bitrate and reference durations in mil
  );
 });

+it("prices all selected generations for any model with API batch options", async () => {
+  mocks.state.generationCount = 3;
+  renderHook(() => useVideoCostEstimate(ModelPage.ImageToVideo, model, "artcraft"));
+  await waitFor(() => expect(mocks.setCredits).toHaveBeenLastCalledWith(ModelPage.ImageToVideo, 126));
+});
+
+it("clears format options and clamps stale batch counts for unsupported models", async () => {
+  mocks.state.generationCount = 4;
+  const single = { ...model, generationCountOptions: [1], outputFormatOptions: undefined } as VideoModel;
+  renderHook(() => useVideoCostEstimate(ModelPage.ImageToVideo, single, "artcraft"));
+  await waitFor(() => expect(mocks.setCredits).toHaveBeenLastCalledWith(ModelPage.ImageToVideo, 42));
+  expect(mocks.estimate).toHaveBeenCalledWith(expect.objectContaining({ output_format: undefined }));
+});
+
 it("does not overwrite a newer model's quote with a slow earlier response", async () => {
  let finishOld: (result: unknown) => void = () => {};
  mocks.estimate.mockImplementationOnce(
@@ -111,6 +128,10 @@ const model = {
  resolutionOptions: ["4K"],
  bitrateOptions: ["normal", "high"],
  defaultBitrate: "normal",
+  outputFormatOptions: ["mp4", "mov"],
+  defaultOutputFormat: "mp4",
+  generationCountOptions: [1, 2, 3, 4],
+  defaultGenerationCount: 1,
  generateWithSound: true,
  sizeOptions: [{ tauriValue: "wide_sixteen_by_nine", textLabel: "16:9" }],
 } as VideoModel;
@@ -1,6 +1,6 @@
 import { useEffect, useState } from "react";
 import { ModelPage } from "@storyteller/ui-model-selector";
-import { Model, VideoModel, resolveVideoDuration, videoResolutionValue } from "@storyteller/model-list";
+import { Model, VideoModel, resolveVideoDuration, resolveVideoGenerationCount, resolveVideoOutputFormat, videoResolutionValue } from "@storyteller/model-list";
 import { GenerationProvider } from "@storyteller/api-enums";
 import { usePromptVideoStore } from "@storyteller/ui-promptbox";
 import {
@@ -32,6 +32,7 @@ export function useVideoCostEstimate(
  const referenceVideos = usePromptVideoStore((s) => s.referenceVideos);
  const referenceAudios = usePromptVideoStore((s) => s.referenceAudios);
  const bitrate = usePromptVideoStore((s) => s.bitrate);
+  const outputFormat = usePromptVideoStore((s) => s.outputFormat);
  const generationCount = usePromptVideoStore((s) => s.generationCount);
  const generateWithSound = usePromptVideoStore((s) => s.generateWithSound);

@@ -72,6 +73,7 @@ export function useVideoCostEstimate(
      duration_seconds: resolveVideoDuration(videoModel, duration, isReferenceMode) ?? undefined,
      generate_audio: videoModel.generateWithSound ? generateWithSound : undefined,
      bitrate: videoModel.bitrateOptions?.length ? bitrate ?? videoModel.defaultBitrate ?? videoModel.bitrateOptions[0] : undefined,
+      output_format: resolveVideoOutputFormat(videoModel, outputFormat),
      start_frame_image_media_token: !isReferenceMode ? referenceImages[0]?.mediaToken : undefined,
      end_frame_image_media_token: !isReferenceMode ? endFrameImage?.mediaToken : undefined,
      reference_image_media_tokens: isReferenceMode ? referenceImages.map((image) => image.mediaToken) : undefined,
@@ -88,7 +90,7 @@ export function useVideoCostEstimate(
        if (cancelled) return;
        if (isEstimateVideoCostSuccess(result)) {
          const unitCredits = result.payload.cost_in_credits;
-          const count = videoModel.id === "seedance_2p0" ? generationCount : 1;
+          const count = resolveVideoGenerationCount(videoModel, generationCount);
          const credits = unitCredits == null ? null : unitCredits * count;
          setEstimatedCreditsForPage(ModelPage.ImageToVideo, credits);
        } else {
@@ -115,6 +117,7 @@ export function useVideoCostEstimate(
    referenceVideos,
    referenceAudios,
    bitrate,
+    outputFormat,
    generationCount,
    endFrameImage,
    generateWithSound,
@@ -407,7 +407,7 @@ export const PromptBoxImage = ({
      }

      if (selectedModel?.supportsQuality()) {
-        request.quality = commonQuality ?? selectedModel.defaultQuality;
+        request.quality = selectedModel.resolveQuality(commonQuality);
      }

      if (
@@ -11,6 +11,9 @@ import { DynamicIcon } from "@storyteller/icons";
 import { arrayMove } from "@dnd-kit/sortable";
 import {
  resolveVideoDuration,
+  resolveVideoGenerationCount,
+  resolveVideoOutputFormat,
+  videoGenerationCounts,
  videoDurationRange,
  videoResolutionValue,
  effectivePromptMaxLength,
@@ -143,6 +146,8 @@ export const PromptBoxVideo = ({
  const duration = usePromptVideoStore((s) => s.duration);
  const bitrate = usePromptVideoStore((s) => s.bitrate);
  const setBitrate = usePromptVideoStore((s) => s.setBitrate);
+  const outputFormat = usePromptVideoStore((s) => s.outputFormat);
+  const setOutputFormat = usePromptVideoStore((s) => s.setOutputFormat);
  const setDuration = usePromptVideoStore((s) => s.setDuration);
  const inputMode = usePromptVideoStore((s) => s.inputMode);
  const setInputMode = usePromptVideoStore((s) => s.setInputMode);
@@ -361,6 +366,14 @@ export const PromptBoxVideo = ({
    }
  }, [selectedModel, setBitrate]);

+  useEffect(() => {
+    const current = usePromptVideoStore.getState().outputFormat;
+    const options = selectedModel?.outputFormatOptions;
+    if (!current || !options?.includes(current)) {
+      setOutputFormat(selectedModel?.defaultOutputFormat ?? options?.[0] ?? null);
+    }
+  }, [selectedModel, setOutputFormat]);
+
  useEffect(() => {
    const current = usePromptVideoStore.getState().aspectRatio;
    if (selectedModel?.sizeOptions.some((option) => option.textLabel === current)) return;
@@ -396,15 +409,16 @@ export const PromptBoxVideo = ({
    }
  }, [selectedModel]);

-  // Reset generation count when switching away from seedance 2.0.
-  // Read from store directly to avoid stale closure (same as duration above).
+  // Keep the selected count within the API's options when switching models.
  useEffect(() => {
-    const currentGenerationCount =
-      usePromptVideoStore.getState().generationCount;
-    if (selectedModel?.id !== "seedance_2p0" && currentGenerationCount > 1) {
-      setGenerationCount(1);
-    }
-  }, [selectedModel]);
+    const current = usePromptVideoStore.getState().generationCount;
+    const next = selectedModel ? resolveVideoGenerationCount(selectedModel, current) : 1;
+    if (next !== current) setGenerationCount(next);
+  }, [selectedModel, setGenerationCount]);
+
+  const generationCounts = selectedModel ? videoGenerationCounts(selectedModel) : [1];
+  const effectiveGenerationCount = selectedModel
+    ? resolveVideoGenerationCount(selectedModel, generationCount) : 1;

  const durationRange = selectedModel
    ? videoDurationRange(selectedModel, inputMode === "reference")
@@ -1049,8 +1063,7 @@ export const PromptBoxVideo = ({

    gtagEvent("enqueue_video");

-    const isSeedance2 = selectedModel.id === "seedance_2p0";
-    const count = isSeedance2 ? generationCount : 1;
+    const count = effectiveGenerationCount;

    const isRefMode =
      inputMode === "reference" && !!selectedModel.supportsReferenceMode;
@@ -1145,6 +1158,9 @@ export const PromptBoxVideo = ({
      if (selectedModel.bitrateOptions?.length) {
        request.bitrate = bitrate ?? selectedModel.defaultBitrate ?? selectedModel.bitrateOptions[0];
      }
+      if (selectedModel.outputFormatOptions?.length) {
+        request.output_format = resolveVideoOutputFormat(selectedModel, outputFormat);
+      }

      // Pass the chosen resolution when the model exposes a resolution picker.
      // Guarded on `resolutionOptions` so a stale store value (left over from a
@@ -1454,6 +1470,21 @@ export const PromptBoxVideo = ({
                </Tooltip>
              )}

+              {!!selectedModel?.outputFormatOptions?.length && (
+                <Tooltip content="Output format" position="top" className="z-50">
+                  <PopoverMenu
+                    panelTitle="Output format"
+                    mode="toggle"
+                    items={selectedModel.outputFormatOptions.map((value) => ({
+                      label: value.toUpperCase(),
+                      action: value,
+                      selected: value === outputFormat,
+                    }))}
+                    onSelect={(item) => setOutputFormat(item.action ?? item.label)}
+                  />
+                </Tooltip>
+              )}
+
              {durationRange && durationRange.max > durationRange.min && (
                <Tooltip content="Duration" position="top" className="z-50">
                  <PopoverMenu
@@ -1522,10 +1553,10 @@ export const PromptBoxVideo = ({
                disabled={!hasClearableContent}
                confirmClear={hasAttachedRefs}
              />
-              {selectedModel?.id === "seedance_2p0" && (
+              {generationCounts.length > 1 && (
                <VideoGenerationCountPicker
-                  maxCount={4}
-                  currentCount={generationCount}
+                  counts={generationCounts}
+                  currentCount={effectiveGenerationCount}
                  handleCountChange={setGenerationCount}
                />
              )}
@@ -1541,9 +1572,7 @@ export const PromptBoxVideo = ({
                    onClick={handleEnqueue}
                    disabled={!prompt.trim()}
                    loading={isEnqueueing}
-                    credits={
-                      credits != null ? credits * generationCount : credits
-                    }
+                    credits={credits}
                  />
                </div>
              </Tooltip>
@@ -10,6 +10,9 @@ interface QualityPickerProps {
 }

 const QUALITY_LABELS: Record<CommonQuality, string> = {
+  [CommonQuality.Auto]: "Auto",
+  [CommonQuality.Max]: "Max",
+  [CommonQuality.XHigh]: "Extra high",
  [CommonQuality.High]: "High",
  [CommonQuality.Medium]: "Medium",
  [CommonQuality.Low]: "Low",
@@ -17,7 +20,7 @@ const QUALITY_LABELS: Record<CommonQuality, string> = {

 /**
 * Stateless picker for image generation "quality" (used by OpenAI image
- * models — gpt_image_1, gpt_image_1p5, gpt_image_2). Models that don't
+ * models, including GPT Image 2.5). Models that don't
 * declare `qualities` should not render this picker.
 */
 export const QualityPicker = ({
@@ -25,7 +28,7 @@ export const QualityPicker = ({
  currentQuality,
  handleCommonQualitySelect,
 }: QualityPickerProps) => {
-  const useQuality = currentQuality ?? model.defaultQuality ?? undefined;
+  const useQuality = model.resolveQuality(currentQuality);

  const handleSelectAdapter = (item: PopoverItem) => {
    const quality = (item.action ?? item.label) as CommonQuality;
@@ -4,27 +4,24 @@ import { Tooltip } from "@storyteller/ui-tooltip";
 import { PROMPT_TOOLBAR_ICON_BUTTON_CLASSES } from "../PromptClearAllButton";

 interface VideoGenerationCountPickerProps {
-  maxCount: number;
+  counts: number[];
  currentCount: number;
  handleCountChange: (count: number) => void;
 }

 export const VideoGenerationCountPicker = ({
-  maxCount,
+  counts,
  currentCount,
  handleCountChange,
 }: VideoGenerationCountPickerProps) => {
-  const options: PopoverItem[] = [];
-  for (let i = 1; i <= maxCount; i++) {
-    options.push({
-      label: String(i),
-      selected: i === currentCount,
-    });
-  }
+  const options: PopoverItem[] = counts.map((count) => ({
+    label: String(count),
+    selected: count === currentCount,
+  }));

  const onSelect = (item: PopoverItem) => {
    const count = parseInt(item.label, 10);
-    if (!isNaN(count) && count >= 1 && count <= maxCount) {
+    if (counts.includes(count)) {
      handleCountChange(count);
    }
  };
@@ -165,6 +165,7 @@ interface PromptVideoStore {
  generateWithSound: boolean;
  duration: number | null;
  bitrate: string | null;
+  outputFormat: string | null;
  inputMode: VideoInputMode;
  generationCount: number;
  setPrompt: (prompt: string) => void;
@@ -178,6 +179,7 @@ interface PromptVideoStore {
  setGenerateWithSound: (value: boolean) => void;
  setDuration: (duration: number | null) => void;
  setBitrate: (bitrate: string | null) => void;
+  setOutputFormat: (outputFormat: string | null) => void;
  setInputMode: (mode: VideoInputMode) => void;
  setGenerationCount: (count: number) => void;
 }
@@ -194,6 +196,7 @@ export const usePromptVideoStore = create<PromptVideoStore>()((set) => ({
  generateWithSound: true,
  duration: null,
  bitrate: null,
+  outputFormat: null,
  inputMode: "reference",
  generationCount: 1,
  setPrompt: (prompt) => set({ prompt }),
@@ -207,6 +210,7 @@ export const usePromptVideoStore = create<PromptVideoStore>()((set) => ({
  setGenerateWithSound: (generateWithSound) => set({ generateWithSound }),
  setDuration: (duration) => set({ duration }),
  setBitrate: (bitrate) => set({ bitrate }),
+  setOutputFormat: (outputFormat) => set({ outputFormat }),
  setInputMode: (inputMode) => set({ inputMode }),
  setGenerationCount: (generationCount) => set({ generationCount }),
 }));
@@ -189,6 +189,13 @@ export class ImageModel extends Model {
    return this.qualityOptions && this.qualityOptions.length > 0;
  }

+  // A saved tier from another model may not be supported by this model.
+  resolveQuality(current?: CommonQuality): CommonQuality | undefined {
+    if (!this.supportsQuality()) return undefined;
+    return current && this.qualityOptions.includes(current)
+      ? current : this.defaultQuality ?? this.qualityOptions[0];
+  }
+
  // Return whether the count of generations is valid for this model
  isValidGenerationCount(count: number): boolean {
    if (this.predefinedGenerationCounts) {
@@ -41,6 +41,10 @@ export class VideoModel extends Model {
  readonly maxDurationWithImageReferences?: number;
  readonly bitrateOptions?: string[];
  readonly defaultBitrate?: string;
+  readonly outputFormatOptions?: string[];
+  readonly defaultOutputFormat?: string;
+  readonly generationCountOptions?: number[];
+  readonly defaultGenerationCount?: number;
  readonly defaultAspectRatio?: string;


@@ -102,6 +106,10 @@ export class VideoModel extends Model {
    maxDurationWithImageReferences?: number;
    bitrateOptions?: string[];
    defaultBitrate?: string;
+    outputFormatOptions?: string[];
+    defaultOutputFormat?: string;
+    generationCountOptions?: number[];
+    defaultGenerationCount?: number;
    defaultAspectRatio?: string;

    supportsReferenceMode?: boolean;
@@ -130,6 +138,10 @@ export class VideoModel extends Model {
    this.maxDurationWithImageReferences = args.maxDurationWithImageReferences;
    this.bitrateOptions = args.bitrateOptions;
    this.defaultBitrate = args.defaultBitrate;
+    this.outputFormatOptions = args.outputFormatOptions;
+    this.defaultOutputFormat = args.defaultOutputFormat;
+    this.generationCountOptions = args.generationCountOptions;
+    this.defaultGenerationCount = args.defaultGenerationCount;
    this.defaultAspectRatio = args.defaultAspectRatio;

    this.supportsReferenceMode = args.supportsReferenceMode;
@@ -24,6 +24,7 @@ export const MODEL_ID_PREFIX_CREATORS: Array<[string, ModelCreator]> = [
  ["vidu", ModelCreator.Vidu],
  ["recraft", ModelCreator.Recraft],
  ["happy_horse", ModelCreator.Alibaba],
+  ["wan", ModelCreator.Alibaba],
  ["minimax", ModelCreator.Hailuo],
  ["qwen", ModelCreator.Alibaba],
  ["suno", ModelCreator.Suno],
@@ -13,6 +13,7 @@ const MODEL_ID_PREFIX_FAMILIES: Array<[string, string]> = [
  ["minimax", "MiniMax"],
  ["hailuo", "MiniMax"],
  ["happy_horse", "Happy Horse"],
+  ["wan", "Wan"],
  ["vidu", "Vidu"],
  // ── Image ── (flux also covers the Flux 3 video model)
  ["flux", "Flux"],
@@ -50,6 +51,7 @@ export const FAMILY_ORDER: string[] = [
  "Grok",
  "Vidu",
  "Happy Horse",
+  "Wan",
  // Image families.
  "Nano Banana",
  "Seedream",
@@ -96,7 +96,10 @@ export const getModelDisplayName = (
    seedance_2p0_bpu_fast: "Seedance 2.0 Plus Ultra Fast",
    seedance_2p0_bpu_mini: "Seedance 2.0 Plus Ultra Mini",
    seedance_2p5: "Seedance 2.5",
+    seedance_2p5_u: "Seedance 2.5 Ultra",
    seedance_2p5_preview: "Seedance 2.5 Preview",
+    wan_3p0: "Wan 3.0",
+    wan_3p0_prime: "Wan 3.0 Prime",
    // Seedance — dot-normalized aliases
    seedance_1_0_lite: "Seedance 1.0 Lite",
    // Anonymous preview aliases the backend used pre-announce
@@ -2,6 +2,9 @@
 /// Not all models have each of these qualities,
 /// but Tauri will interpolate if a mistake is sent.
 export enum CommonQuality {
+  Auto = "auto",
+  Max = "max",
+  XHigh = "xhigh",
  High = "high",
  Medium = "medium",
  Low = "low",
@@ -1,5 +1,23 @@
 import type { VideoModel } from "./VideoModel.js";

+export function resolveVideoOutputFormat(model: VideoModel, current?: string | null): string | undefined {
+  const options = model.outputFormatOptions;
+  if (!options?.length) return undefined;
+  if (current && options.includes(current)) return current;
+  return model.defaultOutputFormat ?? options[0];
+}
+
+export function videoGenerationCounts(model: VideoModel): number[] {
+  return model.generationCountOptions?.length ? model.generationCountOptions : [1];
+}
+
+export function resolveVideoGenerationCount(model: VideoModel, current: number): number {
+  const options = videoGenerationCounts(model);
+  if (options.includes(current)) return current;
+  return model.defaultGenerationCount != null && options.includes(model.defaultGenerationCount)
+    ? model.defaultGenerationCount : options[0];
+}
+
 /** Normalize old display values while preserving future API values. */
 export function videoResolutionValue(value: string): string {
  const legacy: Record<string, string> = {
@@ -403,7 +403,7 @@ export const IMAGE_MODELS: ImageModel[] = [
    progressBarTime: 60000,
    maxPromptLength: Infinity,
    canChangeAspectRatio: true,
-    defaultAspectRatio: CommonAspectRatio.WideSixteenByNine,
+    defaultAspectRatio: CommonAspectRatio.Square,
    aspectRatios: [
      CommonAspectRatio.Auto,
      CommonAspectRatio.Square,
@@ -413,8 +413,19 @@ export const IMAGE_MODELS: ImageModel[] = [
      CommonAspectRatio.WideFourByThree,
      CommonAspectRatio.WideSixteenByNine,
    ],
+    canChangeResolution: true,
+    resolutions: [
+      CommonResolution.OneK,
+      CommonResolution.TwoK,
+      CommonResolution.ThreeK,
+      CommonResolution.FourK,
+    ],
+    defaultResolution: CommonResolution.OneK,
    defaultQuality: CommonQuality.High,
    qualityOptions: [
+      CommonQuality.Auto,
+      CommonQuality.Max,
+      CommonQuality.XHigh,
      CommonQuality.High,
      CommonQuality.Medium,
      CommonQuality.Low,
@@ -438,7 +449,7 @@ export const IMAGE_MODELS: ImageModel[] = [
    progressBarTime: 120000,
    maxPromptLength: Infinity,
    canChangeAspectRatio: true,
-    defaultAspectRatio: CommonAspectRatio.WideSixteenByNine,
+    defaultAspectRatio: CommonAspectRatio.Square,
    aspectRatios: [
      CommonAspectRatio.Auto,
      CommonAspectRatio.Square,
@@ -448,8 +459,19 @@ export const IMAGE_MODELS: ImageModel[] = [
      CommonAspectRatio.WideFourByThree,
      CommonAspectRatio.WideSixteenByNine,
    ],
+    canChangeResolution: true,
+    resolutions: [
+      CommonResolution.OneK,
+      CommonResolution.TwoK,
+      CommonResolution.ThreeK,
+      CommonResolution.FourK,
+    ],
+    defaultResolution: CommonResolution.OneK,
    defaultQuality: CommonQuality.High,
    qualityOptions: [
+      CommonQuality.Auto,
+      CommonQuality.Max,
+      CommonQuality.XHigh,
      CommonQuality.High,
      CommonQuality.Medium,
      CommonQuality.Low,
@@ -345,6 +345,8 @@ export const VIDEO_MODELS: VideoModel[] = [
  new VideoModel({
    id: "seedance_2p0",
    tauriId: "seedance_2p0",
+    generationCountOptions: [1, 2, 3, 4],
+    defaultGenerationCount: 1,
    fullName: "Seedance 2.0",
    category: "video",
    creator: ModelCreator.Bytedance,
@@ -7,10 +7,14 @@ import {
 } from "./buildModelsFromListing.js";
 import {
  resolveVideoDuration,
+  resolveVideoGenerationCount,
+  resolveVideoOutputFormat,
+  videoGenerationCounts,
  videoDurationRange,
  videoResolutionValue,
 } from "../classes/videoOptions.js";
 import { IMAGE_MODELS } from "../lists/ImageModels.js";
+import { CommonQuality } from "../classes/properties/CommonQuality.js";
 import { VIDEO_MODELS } from "../lists/VideoModels.js";
 import { SPLAT_MODELS } from "../lists/SplatModels.js";

@@ -24,6 +28,8 @@ describe("desktop catalog", () => {
      "flux_3",
      "minimax_h3",
      "vidu_q3",
+      "wan_3p0",
+      "wan_3p0_prime",
      "future_video_v99",
    ];
    const models = buildVideoModelsFromListing(
@@ -97,6 +103,61 @@ describe("desktop catalog", () => {
 });

 describe("API video options", () => {
+  it.each(["wan_3p0", "wan_3p0_prime"])("hydrates %s with Wan's reference and duration capabilities", (id) => {
+    const [model] = buildVideoModelsFromListing([], [{
+      model: id,
+      model_creator: "alibaba",
+      text_to_video_supported: true,
+      text_prompt_max_length: 20000,
+      starting_keyframe_supported: true,
+      ending_keyframe_supported: true,
+      image_references_supported: true,
+      image_references_max: 10,
+      video_references_max: 5,
+      video_references_max_total_duration_seconds: 15,
+      audio_references_max: 5,
+      audio_references_max_total_duration_seconds: 15,
+      show_generate_with_sound_toggle: true,
+      resolution_options: ["four_eighty_p", "seven_twenty_p", "ten_eighty_p"],
+      resolution_default: "seven_twenty_p",
+      duration_seconds_min: 2,
+      duration_seconds_max: 30,
+      duration_seconds_default: 5,
+      batch_size_options: [1],
+      batch_size_default: 1,
+    }]);
+    expect(model.getProviders()).toEqual([GenerationProvider.Artcraft]);
+    expect(model).toMatchObject({
+      startFrame: true, endFrame: true, textToVideoSupported: true,
+      supportsReferenceMode: true, maxReferenceImages: 10,
+      maxReferenceVideos: 5, maxVideoRefDuration: 15,
+      maxReferenceAudios: 5, maxAudioRefDuration: 15,
+      generateWithSound: true, maxPromptLength: 20000,
+      resolutionOptions: ["480p", "720p", "1080p"], defaultResolution: "720p",
+    });
+    expect(videoDurationRange(model, true)).toEqual({ min: 2, max: 30 });
+    expect(resolveVideoDuration(model, null, false)).toBe(5);
+    expect(videoGenerationCounts(model)).toEqual([1]);
+    expect(resolveVideoGenerationCount(model, 4)).toBe(1);
+  });
+
+  it.each(["seedance_2p5", "seedance_2p5_u"])("preserves %s output formats and batch choices", (id) => {
+    const [model] = buildVideoModelsFromListing([], [{
+      model: id,
+      output_format_options: ["mp4", "mov", "future_format"],
+      output_format_default: "mp4",
+      batch_size_options: [1, 2, 3, 4],
+      batch_size_default: 1,
+    }]);
+    expect(model.outputFormatOptions).toEqual(["mp4", "mov", "future_format"]);
+    expect(model.defaultOutputFormat).toBe("mp4");
+    expect(resolveVideoOutputFormat(model, "mov")).toBe("mov");
+    expect(resolveVideoOutputFormat(model, "unsupported")).toBe("mp4");
+    expect(videoGenerationCounts(model)).toEqual([1, 2, 3, 4]);
+    expect(resolveVideoGenerationCount(model, 3)).toBe(3);
+    expect(resolveVideoGenerationCount(model, 5)).toBe(1);
+  });
+
  it("uses range durations, reference-mode caps, bitrate, and future resolutions", () => {
    const [model] = buildVideoModelsFromListing(
      [],
@@ -134,6 +195,37 @@ describe("API video options", () => {
  });
 });

+it.each(["gpt_image_2p5_flare", "gpt_image_2p5_sunburst"])("preserves every %s quality and resolution option", (id) => {
+  const qualities = ["auto", "max", "xhigh", "high", "medium", "low"];
+  const resolutions = ["one_k", "two_k", "three_k", "four_k"];
+  const model = buildImageModelsFromListing(IMAGE_MODELS, [{
+    model: id,
+    quality_options: qualities,
+    default_quality: "high",
+    resolution_options: resolutions,
+    resolution_default: "one_k",
+    image_refs_supported: true,
+  }]).find((m) => m.tauriId === id)!;
+  expect(model.qualityOptions).toEqual(qualities);
+  expect(model.defaultQuality).toBe("high");
+  expect(model.resolutions).toEqual(resolutions);
+  expect(model.defaultResolution).toBe("one_k");
+  expect(model.canEditImages).toBe(true);
+  expect(model.resolveQuality(CommonQuality.Max)).toBe("max");
+  expect(model.resolveQuality(CommonQuality.Auto)).toBe("auto");
+});
+
+it("uses the new model's default when a saved quality tier is unsupported", () => {
+  const [model] = buildImageModelsFromListing([], [{
+    model: "gpt_image_2",
+    quality_options: ["high", "medium", "low"],
+    default_quality: "high",
+  }]);
+  expect(model.resolveQuality(CommonQuality.Max)).toBe("high");
+  expect(model.resolveQuality(CommonQuality.XHigh)).toBe("high");
+  expect(model.resolveQuality(CommonQuality.Low)).toBe("low");
+});
+
 it("offers ArtCraft and direct Midjourney only for supported Midjourney models", () => {
  const models = buildImageModelsFromListing([], [
    { model: "midjourney_7" }, { model: "midjourney_7_niji" },
@@ -76,6 +76,8 @@ export interface ListingVideoModel extends ListingModelBase {
  duration_seconds_max_with_image_references?: number | null;
  bitrate_options?: string[] | null;
  bitrate_default?: string | null;
+  output_format_options?: string[] | null;
+  output_format_default?: string | null;
  duration_seconds_default?: number | null;
 }

@@ -269,6 +271,10 @@ const mergedVideoModel = (
    maxDurationWithImageReferences: m.duration_seconds_max_with_image_references ?? o?.maxDurationWithImageReferences,
    bitrateOptions: m.bitrate_options ?? o?.bitrateOptions,
    defaultBitrate: m.bitrate_default ?? o?.defaultBitrate,
+    outputFormatOptions: m.output_format_options ?? o?.outputFormatOptions,
+    defaultOutputFormat: m.output_format_default ?? o?.defaultOutputFormat,
+    generationCountOptions: m.batch_size_options ?? o?.generationCountOptions,
+    defaultGenerationCount: m.batch_size_default ?? o?.defaultGenerationCount,
    defaultAspectRatio: m.aspect_ratio_default ?? o?.defaultAspectRatio,
    supportsReferenceMode: m.image_references_supported ?? o?.supportsReferenceMode,
    maxReferenceImages: m.image_references_max ?? o?.maxReferenceImages,
@@ -16,6 +16,7 @@ export interface EstimateVideoCostRequest {
  video_batch_count?: number;
  generate_audio?: boolean;
  bitrate?: string;
+  output_format?: string;
  start_frame_image_media_token?: string;
  end_frame_image_media_token?: string;
  reference_image_media_tokens?: string[];
@@ -24,6 +24,7 @@ it("passes unknown model and option strings through image and video commands", a
  const video = {
    model: "future_video",
    bitrate: "high",
+    output_format: "mov",
    resolution: "eight_k",
    duration_seconds: 30,
    generate_audio: false,
@@ -54,6 +55,25 @@ it("dispatches every added modality and its pricing through Tauri", async () =>
  ]);
 });

+it.each(["gpt_image_2p5_flare", "gpt_image_2p5_sunburst"])("forwards every %s quality tier", async (model) => {
+  for (const quality of ["auto", "max", "xhigh", "high", "medium", "low"]) {
+    const request = { model, quality, resolution: "four_k" };
+    await GenerateImage(request);
+    expect(invoke).toHaveBeenLastCalledWith("generate_image_command", { request });
+  }
+});
+
+it.each(["wan_3p0", "wan_3p0_prime"])("forwards %s with reference media", async (model) => {
+  const request = {
+    model, duration_seconds: 15, resolution: "ten_eighty_p", generate_audio: true,
+    reference_image_media_tokens: ["mf_image"],
+    reference_video_media_tokens: ["mf_video"],
+    reference_audio_media_tokens: ["mf_audio"],
+  };
+  await GenerateVideo(request);
+  expect(invoke).toHaveBeenLastCalledWith("generate_video_command", { request });
+});
+
 it("preserves reference duration hints in video pricing", async () => {
  const request = {
    model: "future_video",
@@ -38,6 +38,7 @@ export interface GenerateVideoRequest {
  resolution?: CommonResolution | string;

  bitrate?: string;
+  output_format?: string;
  duration_seconds?: number;
  generate_audio?: boolean;
  video_batch_count?: number;
@@ -55,6 +55,8 @@ export interface OmniGenVideoModelDetails {
  resolution_default?: string;
  bitrate_options?: string[];
  bitrate_default?: string;
+  output_format_options?: string[];
+  output_format_default?: string;
  quality_options?: string[];
  default_quality?: string;
  duration_seconds_min?: number;