don't serve bad traffic

## Changes

- **Prompt character counters on web**: the website and webapp
promptboxes (image + video) now show a `123 / 2000` counter driven by
the API's `text_prompt_max_length`, matching the desktop app. It turns
red when over the limit and generation is blocked with a toast; typing
is never clamped. The webapp's fullscreen prompt editor shows the
counter too.
- **Seedance CJK waiver**: seedance video models normally use the API
limit, but the limit is waived (shows `∞`) while the prompt contains
Chinese characters. Applies on desktop and web via a shared
`effectivePromptMaxLength` helper.
- **API type parity**: added the new `model_creator` field to the web
`OmniGenImageModelInfo` / `OmniGenVideoModelInfo` types (desktop types
were already up to date).
- **Cleanup**: deleted the manual model-to-creator map
(`MODEL_TYPE_TO_CREATOR`, ~150 lines) now that creator info comes from
the API; lightbox model icons use the shared
`getCreatorIconPathForModelId` lookup like every other surface.

crates/service/web/storyteller_web/src/http_server/endpoints/inference_job/get/get_inference_job_status_handler.rs

@@ -4,6 +4,7 @@ use crate::http_server::common_responses::common_web_error::CommonWebError;
 use crate::http_server::common_responses::media::media_domain::MediaDomain;
 use crate::http_server::common_responses::media::media_links_builder::MediaLinksBuilder;
 use crate::http_server::endpoints::inference_job::utils::estimates::estimate_job_progress::estimate_job_progress;
+use crate::http_server::endpoints::tts::enqueue_infer_tts_handler::enqueue_infer_tts_handler::SYNTHETIC_INFERENCE_JOB_TOKEN;
 use crate::http_server::endpoints::inference_job::utils::extractors::extract_lipsync_details::extract_lipsync_details;
 use crate::http_server::endpoints::inference_job::utils::extractors::extract_live_portrait_details::extract_live_portrait_details;
 use crate::http_server::endpoints::inference_job::utils::extractors::extract_polymorphic_inference_args::extract_polymorphic_inference_args;
@@ -154,12 +155,9 @@ pub async fn get_inference_job_status_handler(
   path: Path<GetInferenceJobStatusPathInfo>,
   server_state: web::Data<Arc<ServerState>>,
 ) -> Result<Json<GetInferenceJobStatusSuccessResponse>, CommonWebError> {
-  if path.token.as_str().trim() == "None" {
-    // NB: A bunch of Python clients use our API and can fail in this manner.
-    // This was a large traffic driver during the 2023-03-08 outage.
-    // Presumably, it's this client: https://github.com/shards-7/fakeyou.py
-    return Err(CommonWebError::NotFound);
-  }
+
+  // Don't serve bad traffic
+  blackhole_jobs(&path.token)?;
 
   // NB: Since this is publicly exposed, we don't query sensitive data.
   let maybe_status = get_inference_job_status(
@@ -233,6 +231,24 @@ pub async fn get_inference_job_status_handler(
   }))
 }
 
+/// Reject junk job tokens that abusive or broken clients poll in high volume, before they reach
+/// MySQL or Redis. Deliberately not logged — the whole point is to make these requests as cheap
+/// as possible.
+fn blackhole_jobs(token: &InferenceJobToken) -> Result<(), CommonWebError> {
+  match token.as_str().trim() {
+    // NB: The shut-down legacy TTS enqueue endpoint hands this sentinel token to old clients,
+    // and some of them poll it forever.
+    SYNTHETIC_INFERENCE_JOB_TOKEN => Err(CommonWebError::TooManyRequests),
+    // NB: A bunch of Python clients use our API and can fail in this manner.
+    // This was a large traffic driver during the 2023-03-08 outage.
+    // Presumably, it's this client: https://github.com/shards-7/fakeyou.py
+    "None" => Err(CommonWebError::NotFound),
+    // NB: Broken Javascript clients poll with a stringified `undefined` token.
+    "undefined" => Err(CommonWebError::NotFound),
+    _ => Ok(()),
+  }
+}
+
 fn record_to_payload(
   record: GenericInferenceJobStatus,
   maybe_extra_status_description: Option<String>,

crates/service/web/storyteller_web/src/http_server/endpoints/tts/enqueue_infer_tts_handler/enqueue_infer_tts_handler.rs

@@ -2,8 +2,9 @@ use actix_web::web::Json;
 use utoipa::ToSchema;
 
 /// Sentinel job token returned to legacy clients now that TTS inference is shut down.
-/// There is no real job behind it; status polls for it will never complete.
-const SYNTHETIC_INFERENCE_JOB_TOKEN: &str = "synthetic_too_many_requests";
+/// There is no real job behind it; status polls for it are rejected with a 429 before
+/// touching the database (see `get_inference_job_status_handler`).
+pub const SYNTHETIC_INFERENCE_JOB_TOKEN: &str = "synthetic_too_many_requests";
 
 #[derive(Serialize, ToSchema)]
 pub struct InferTtsSuccessResponse {

frontend/apps/artcraft-webapp/index.html

@@ -26,6 +26,37 @@
       content="ArtCraft, AI, AI Video, AI Images, Open Source"
     />
 
+    <!-- Social share (Open Graph + Twitter Card). Crawlers don't run JS, so
+         these baked-in tags are what Discord/Slack/X show for any webapp URL.
+         Image URLs must be absolute or crawlers ignore them. -->
+    <meta property="og:type" content="website" />
+    <meta property="og:site_name" content="ArtCraft" />
+    <meta
+      property="og:title"
+      content="ArtCraft. AI Video and Images. Fast and Open Desktop App."
+    />
+    <meta
+      property="og:description"
+      content="ArtCraft is an Open Desktop app for generating AI Video and Images. You own ArtCraft!"
+    />
+    <meta
+      property="og:image"
+      content="https://app.getartcraft.com/images/og-image.png"
+    />
+    <meta name="twitter:card" content="summary_large_image" />
+    <meta
+      name="twitter:title"
+      content="ArtCraft. AI Video and Images. Fast and Open Desktop App."
+    />
+    <meta
+      name="twitter:description"
+      content="ArtCraft is an Open Desktop app for generating AI Video and Images. You own ArtCraft!"
+    />
+    <meta
+      name="twitter:image"
+      content="https://app.getartcraft.com/images/og-image.png"
+    />
+
     <base href="/" />
 
     <meta name="viewport" content="width=device-width, initial-scale=1" />

frontend/apps/artcraft-webapp/src/components/lightbox/LightboxDetails.tsx

@@ -28,7 +28,7 @@ import {
   THUMBNAIL_SIZES,
 } from "@storyteller/common";
 import {
-  getModelCreatorIcon,
+  getCreatorIconPathForModelId,
   getModelDisplayName,
   getProviderDisplayName,
   getProviderIconByName,
@@ -361,7 +361,13 @@ export function LightboxDetails({
                       label="Model"
                       value={
                         <>
-                          {getModelCreatorIcon(promptData.modelType)}
+                          <img
+                            src={getCreatorIconPathForModelId(
+                              promptData.modelType,
+                            )}
+                            alt="Model creator logo"
+                            className="h-4 w-4 invert"
+                          />
                           <span>
                             {getModelDisplayName(promptData.modelType)}
                           </span>

frontend/apps/artcraft-webapp/src/components/prompt-box/PromptBox.tsx

@@ -72,6 +72,11 @@ interface PromptBoxProps {
 
   // @-mention support (enables colored prompt overlay + autocomplete)
   mentionItems?: MentionItem[];
+
+  // Soft prompt-length limit from the model API (`text_prompt_max_length`).
+  // Undefined = unlimited (no counter). The limit is not enforced here; the
+  // page's submit handler blocks generation when over.
+  maxPromptLength?: number;
 }
 
 export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
@@ -102,6 +107,7 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
       mediaReferenceRow,
       modelSelector,
       mentionItems,
+      maxPromptLength,
     },
     ref,
   ) => {
@@ -391,7 +397,7 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
                 </Tooltip>
               )}
 
-              <div className="relative flex-1">
+              <div className="promptbox-resize-wrap relative flex-1">
                 {hasMentionItems && mentionItems ? (
                   <MentionTextarea
                     ref={mentionEditorRef}
@@ -519,6 +525,21 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
                   </>
                 )}
                 <PromptFullscreenButton onClick={openFullscreen} />
+
+                {maxPromptLength !== undefined && (
+                  <div
+                    className={twMerge(
+                      // right-4 keeps the counter clear of the textarea's resize grip.
+                      "pointer-events-none absolute -bottom-1 right-4 text-[10px] tabular-nums",
+                      isFinite(maxPromptLength) && prompt.length > maxPromptLength
+                        ? "text-red-500"
+                        : "text-base-fg/40",
+                    )}
+                  >
+                    {prompt.length} /{" "}
+                    {isFinite(maxPromptLength) ? maxPromptLength : "∞"}
+                  </div>
+                )}
               </div>
             </div>
 
@@ -565,6 +586,8 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
         <PromptFullscreenModal
           isOpen={isFullscreen}
           onClose={closeFullscreen}
+          promptLength={prompt.length}
+          maxPromptLength={maxPromptLength}
           footerControls={
             <>
               {modelSelector}

frontend/apps/artcraft-webapp/src/components/prompt-box/PromptFullscreen.tsx

@@ -57,6 +57,10 @@ interface PromptFullscreenModalProps {
   isOpen: boolean;
   onClose: () => void;
   children: ReactNode;
+  /** Current prompt length, shown with the counter when a limit is set. */
+  promptLength?: number;
+  /** Soft prompt-length limit; undefined hides the counter entirely. */
+  maxPromptLength?: number;
   /** Controls rendered in the footer, left of the Done button (e.g. the model
    *  selector + character button). */
   footerControls?: ReactNode;
@@ -68,9 +72,15 @@ export const PromptFullscreenModal = ({
   isOpen,
   onClose,
   children,
+  promptLength,
+  maxPromptLength,
   footerControls,
   imagePromptRow,
 }: PromptFullscreenModalProps) => {
+  const showCounter =
+    promptLength !== undefined && maxPromptLength !== undefined;
+  const overLimit =
+    showCounter && isFinite(maxPromptLength) && promptLength > maxPromptLength;
   const contentRef = useRef<HTMLDivElement>(null);
 
   // Focus the editor and drop the caret at the end once the overlay opens.
@@ -106,6 +116,17 @@ export const PromptFullscreenModal = ({
         >
           {children}
         </div>
+        {showCounter && (
+          <div className="flex shrink-0 items-center justify-end">
+            <span
+              className={`text-[11px] tabular-nums ${
+                overLimit ? "text-red-500" : "text-base-fg/40"
+              }`}
+            >
+              {promptLength} / {isFinite(maxPromptLength) ? maxPromptLength : "∞"}
+            </span>
+          </div>
+        )}
         {imagePromptRow && <div className="shrink-0">{imagePromptRow}</div>}
         <div className="flex shrink-0 items-center justify-between gap-2">
           <div className="flex items-center gap-2">{footerControls}</div>

frontend/apps/artcraft-webapp/src/pages/create-image/create-image.tsx

@@ -121,6 +121,9 @@ export default function CreateImage() {
     return apiModels.find((m) => m.model === DEFAULT_MODEL_ID) ?? apiModels[0];
   }, [apiModels, ui.selectedModelId]);
 
+  // Soft prompt limit from the API; undefined (no model / unset) = unlimited.
+  const maxPromptLength = selectedModel?.text_prompt_max_length ?? undefined;
+
   const prompt = ui.prompt;
   const setPrompt = useCallback((v: string) => setUi({ prompt: v }), [setUi]);
 
@@ -363,6 +366,13 @@ export default function CreateImage() {
     }
     if (!prompt.trim() || isGenerating || !selectedModel) return;
 
+    if (maxPromptLength !== undefined && prompt.length > maxPromptLength) {
+      toast.error(
+        `Prompt exceeds the ${maxPromptLength} character limit for this model`,
+      );
+      return;
+    }
+
     setIsGenerating(true);
     const batchId = startBatch(
       prompt,
@@ -435,6 +445,7 @@ export default function CreateImage() {
     prompt,
     isGenerating,
     selectedModel,
+    maxPromptLength,
     numImages,
     aspectRatio,
     resolution,
@@ -618,6 +629,7 @@ export default function CreateImage() {
             onSubmit={handleGenerate}
             isSubmitting={isGenerating}
             credits={estimatedCredits}
+            maxPromptLength={maxPromptLength}
             placeholder="Describe what you want in the image..."
             supportsImagePrompts={!!selectedModel?.image_refs_supported}
             maxImagePromptCount={maxImageRefs}

frontend/apps/artcraft-webapp/src/pages/create-video/create-video.tsx

@@ -62,6 +62,7 @@ import {
   OMNI_GENERATE_OUTAGE_MESSAGE,
 } from "@storyteller/omni-gen";
 import {
+  effectivePromptMaxLength,
   getCreatorIconPathForModelId,
   getModelDescription,
   getModelInfo,
@@ -273,6 +274,18 @@ export default function CreateVideo() {
 
   const requiresImageInput = selectedModel?.text_to_video_supported === false;
 
+  // Soft prompt limit from the API; undefined = unlimited. Seedance waives
+  // the limit (Infinity) when the prompt contains Chinese characters: the
+  // server counts CJK as more than one unit, so a client-side character
+  // count would false-positive.
+  const maxPromptLength = selectedModel
+    ? effectivePromptMaxLength(
+        selectedModel.model,
+        selectedModel.text_prompt_max_length ?? undefined,
+        ui.prompt,
+      )
+    : undefined;
+
   const prompt = ui.prompt;
   const setPrompt = useCallback((v: string) => setUi({ prompt: v }), [setUi]);
 
@@ -877,6 +890,12 @@ export default function CreateVideo() {
       });
       return;
     }
+    if (maxPromptLength !== undefined && prompt.length > maxPromptLength) {
+      toast.error(
+        `Prompt exceeds the ${maxPromptLength} character limit for this model`,
+      );
+      return;
+    }
     console.log("[generate-video] starting", {
       model: selectedModel.model,
       numVideos,
@@ -1036,6 +1055,7 @@ export default function CreateVideo() {
     needsImage,
     isReferenceMode,
     selectedModel,
+    maxPromptLength,
     selectedSize,
     numVideos,
     effectiveDuration,
@@ -1364,6 +1384,7 @@ export default function CreateVideo() {
             onSubmit={handleGenerate}
             isSubmitting={isGenerating}
             credits={estimatedCredits}
+            maxPromptLength={maxPromptLength}
             placeholder="Describe the video you want to generate..."
             supportsImagePrompts={supportsImagePrompts}
             maxImagePromptCount={

frontend/apps/artcraft-webapp/src/styles.css

@@ -271,6 +271,43 @@
   background-color: rgba(255, 255, 255, 0.35);
 }
 
+/* Hide the native resize grip (Chromium draws different grips for textareas
+   vs contenteditable divs) and draw one consistent grip on the wrapper so the
+   plain and @-mention prompt editors look identical. */
+.promptbox-scrollbar::-webkit-resizer {
+  background: transparent;
+  border: none;
+}
+
+.promptbox-resize-wrap {
+  position: relative;
+}
+
+.promptbox-resize-wrap::after {
+  content: "";
+  position: absolute;
+  bottom: 2px;
+  right: 2px;
+  width: 10px;
+  height: 10px;
+  pointer-events: none;
+  background: linear-gradient(
+    135deg,
+    transparent 40%,
+    rgba(255, 255, 255, 0.5) 40%,
+    rgba(255, 255, 255, 0.5) 45%,
+    transparent 45%,
+    transparent 55%,
+    rgba(255, 255, 255, 0.5) 55%,
+    rgba(255, 255, 255, 0.5) 60%,
+    transparent 60%,
+    transparent 70%,
+    rgba(255, 255, 255, 0.5) 70%,
+    rgba(255, 255, 255, 0.5) 75%,
+    transparent 75%
+  );
+}
+
 @keyframes fade-in-up {
   from {
     opacity: 0;

frontend/apps/artcraft-website/src/components/lightbox/LightboxDetails.tsx

@@ -25,7 +25,7 @@ import {
   THUMBNAIL_SIZES,
 } from "@storyteller/common";
 import {
-  getModelCreatorIcon,
+  getCreatorIconPathForModelId,
   getModelDisplayName,
   getProviderDisplayName,
   getProviderIconByName,
@@ -297,7 +297,13 @@ export function LightboxDetails({
                       label="Model"
                       value={
                         <>
-                          {getModelCreatorIcon(promptData.modelType)}
+                          <img
+                            src={getCreatorIconPathForModelId(
+                              promptData.modelType,
+                            )}
+                            alt="Model creator logo"
+                            className="h-4 w-4 invert"
+                          />
                           <span>
                             {getModelDisplayName(promptData.modelType)}
                           </span>

frontend/apps/artcraft-website/src/components/prompt-box/PromptBox.tsx

@@ -109,6 +109,11 @@ interface PromptBoxProps {
 
   // @-mention support (enables colored prompt overlay + autocomplete)
   mentionItems?: MentionItem[];
+
+  // Soft prompt-length limit from the model API (`text_prompt_max_length`).
+  // Undefined = unlimited (no counter). The limit is not enforced here; the
+  // page's submit handler blocks generation when over.
+  maxPromptLength?: number;
 }
 
 export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
@@ -139,6 +144,7 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
       mediaReferenceRow,
       modelSelector,
       mentionItems,
+      maxPromptLength,
     },
     ref,
   ) => {
@@ -423,7 +429,7 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
                 </Tooltip>
               )}
 
-              <div className="relative flex-1">
+              <div className="promptbox-resize-wrap relative flex-1">
                 {hasMentionItems && mentionItems ? (
                   <MentionTextarea
                     ref={mentionEditorRef}
@@ -545,6 +551,21 @@ export const PromptBox = forwardRef<HTMLDivElement, PromptBoxProps>(
                     )}
                   </>
                 )}
+
+                {maxPromptLength !== undefined && (
+                  <div
+                    className={twMerge(
+                      // right-4 keeps the counter clear of the textarea's resize grip.
+                      "pointer-events-none absolute -bottom-1 right-4 text-[10px] tabular-nums",
+                      isFinite(maxPromptLength) && prompt.length > maxPromptLength
+                        ? "text-red-500"
+                        : "text-base-fg/40",
+                    )}
+                  >
+                    {prompt.length} /{" "}
+                    {isFinite(maxPromptLength) ? maxPromptLength : "∞"}
+                  </div>
+                )}
               </div>
             </div>
 

frontend/apps/artcraft-website/src/config/page-meta.ts

@@ -19,7 +19,7 @@
 export const SITE_ORIGIN = "https://getartcraft.com";
 
 // Default share image, served from /public. Pages may override per-route.
-export const DEFAULT_OG_IMAGE = "/images/3d-interface-preview.jpg";
+export const DEFAULT_OG_IMAGE = "/images/og-image.png";
 
 export interface PageMeta {
   title: string;

frontend/apps/artcraft-website/src/pages/create-image/create-image.tsx

@@ -20,6 +20,7 @@ import {
 } from "../../components/generation-gallery";
 import { Lightbox } from "../../components/lightbox/lightbox";
 import { AppDownloadCta } from "../../components/app-download-cta/AppDownloadCta";
+import { toast } from "../../components/toast/toast";
 import { useCreateImageStore } from "./create-image-store";
 import { enqueueImageGeneration, startPolling } from "./generate-image-api";
 import { AspectRatioPicker } from "./components/AspectRatioPicker";
@@ -83,6 +84,9 @@ export default function CreateImage() {
     return apiModels.find((m) => m.model === DEFAULT_MODEL_ID) ?? apiModels[0];
   }, [apiModels, ui.selectedModelId]);
 
+  // Soft prompt limit from the API; undefined (no model / unset) = unlimited.
+  const maxPromptLength = selectedModel?.text_prompt_max_length ?? undefined;
+
   const prompt = ui.prompt;
   const setPrompt = useCallback((v: string) => setUi({ prompt: v }), [setUi]);
   const aspectRatio = ui.aspectRatio;
@@ -278,6 +282,13 @@ export default function CreateImage() {
   const handleGenerate = useCallback(async () => {
     if (!prompt.trim() || isGenerating || !selectedModel) return;
 
+    if (maxPromptLength !== undefined && prompt.length > maxPromptLength) {
+      toast.error(
+        `Prompt exceeds the ${maxPromptLength} character limit for this model`,
+      );
+      return;
+    }
+
     setIsGenerating(true);
     const batchId = startBatch(
       prompt,
@@ -337,6 +348,7 @@ export default function CreateImage() {
     prompt,
     isGenerating,
     selectedModel,
+    maxPromptLength,
     numImages,
     aspectRatio,
     resolution,
@@ -396,6 +408,7 @@ export default function CreateImage() {
             onSubmit={handleGenerate}
             isSubmitting={isGenerating}
             credits={estimatedCredits}
+            maxPromptLength={maxPromptLength}
             placeholder="Describe what you want in the image..."
             supportsImagePrompts={!!selectedModel?.image_refs_supported}
             maxImagePromptCount={maxImageRefs}

frontend/apps/artcraft-website/src/pages/create-video/create-video.tsx

@@ -46,7 +46,11 @@ import {
 import { GenerationCountPicker } from "../create-image/components/GenerationCountPicker";
 import { useVideoCostEstimate } from "../../lib/cost-estimate-api";
 import { useOmniGenVideoModels } from "@storyteller/omni-gen";
-import { getCreatorIconPathForModelId } from "@storyteller/model-list";
+import {
+  effectivePromptMaxLength,
+  getCreatorIconPathForModelId,
+} from "@storyteller/model-list";
+import { toast } from "../../components/toast/toast";
 
 // ── Constants ────────────────────────────────────────────────────────────
 
@@ -182,6 +186,18 @@ export default function CreateVideo() {
     return apiModels.find((m) => m.model === DEFAULT_MODEL_ID) ?? apiModels[0];
   }, [apiModels, ui.selectedModelId]);
 
+  // Soft prompt limit from the API; undefined = unlimited. Seedance waives
+  // the limit (Infinity) when the prompt contains Chinese characters: the
+  // server counts CJK as more than one unit, so a client-side character
+  // count would false-positive.
+  const maxPromptLength = selectedModel
+    ? effectivePromptMaxLength(
+        selectedModel.model,
+        selectedModel.text_prompt_max_length ?? undefined,
+        ui.prompt,
+      )
+    : undefined;
+
   const prompt = ui.prompt;
   const setPrompt = useCallback((v: string) => setUi({ prompt: v }), [setUi]);
   const selectedSize = ui.selectedSize;
@@ -717,6 +733,12 @@ export default function CreateVideo() {
       });
       return;
     }
+    if (maxPromptLength !== undefined && prompt.length > maxPromptLength) {
+      toast.error(
+        `Prompt exceeds the ${maxPromptLength} character limit for this model`,
+      );
+      return;
+    }
     console.log("[generate-video] starting", {
       model: selectedModel.model,
       numVideos,
@@ -852,6 +874,7 @@ export default function CreateVideo() {
     needsImage,
     isReferenceMode,
     selectedModel,
+    maxPromptLength,
     selectedSize,
     numVideos,
     duration,
@@ -935,6 +958,7 @@ export default function CreateVideo() {
             onSubmit={handleGenerate}
             isSubmitting={isGenerating || needsImage}
             credits={estimatedCredits}
+            maxPromptLength={maxPromptLength}
             placeholder="Describe the video you want to generate..."
             supportsImagePrompts={supportsImagePrompts}
             maxImagePromptCount={

frontend/apps/artcraft-website/src/styles.css

@@ -120,6 +120,43 @@
   background-color: rgba(255, 255, 255, 0.35);
 }
 
+/* Hide the native resize grip (Chromium draws different grips for textareas
+   vs contenteditable divs) and draw one consistent grip on the wrapper so the
+   plain and @-mention prompt editors look identical. */
+.promptbox-scrollbar::-webkit-resizer {
+  background: transparent;
+  border: none;
+}
+
+.promptbox-resize-wrap {
+  position: relative;
+}
+
+.promptbox-resize-wrap::after {
+  content: "";
+  position: absolute;
+  bottom: 2px;
+  right: 2px;
+  width: 10px;
+  height: 10px;
+  pointer-events: none;
+  background: linear-gradient(
+    135deg,
+    transparent 40%,
+    rgba(255, 255, 255, 0.5) 40%,
+    rgba(255, 255, 255, 0.5) 45%,
+    transparent 45%,
+    transparent 55%,
+    rgba(255, 255, 255, 0.5) 55%,
+    rgba(255, 255, 255, 0.5) 60%,
+    transparent 60%,
+    transparent 70%,
+    rgba(255, 255, 255, 0.5) 70%,
+    rgba(255, 255, 255, 0.5) 75%,
+    transparent 75%
+  );
+}
+
 @keyframes fade-in-up {
   from {
     opacity: 0;

frontend/libs/api/src/lib/OmniGenApi.ts

@@ -77,6 +77,8 @@ export interface OmniGenVideoGenerateResponse {
 
 export interface OmniGenImageModelInfo {
   model: string;
+  // ModelCreator enum as a snake_case string, e.g. "bytedance", "google".
+  model_creator?: string | null;
   is_disabled: boolean | null;
   full_name: string | null;
   extra_info_short?: string | null;
@@ -111,6 +113,8 @@ export interface OmniGenImageModelsResponse {
 
 export interface OmniGenVideoModelInfo {
   model: string;
+  // ModelCreator enum as a snake_case string, e.g. "bytedance", "open_ai".
+  model_creator?: string | null;
   is_disabled: boolean | null;
   full_name: string | null;
   extra_info_short?: string | null;

frontend/libs/components/lightbox-modal/src/lib/lightbox-modal.tsx

@@ -43,7 +43,7 @@ import {
   isActionReminderOpen,
 } from "@storyteller/ui-action-reminder-modal";
 import {
-  getModelCreatorIcon,
+  getCreatorIconPathForModelId,
   getModelDisplayName,
   getProviderDisplayName,
   getProviderIconByName,
@@ -1327,7 +1327,11 @@ function InfoSection({
             label="Model"
             value={
               <>
-                {getModelCreatorIcon(modelType)}
+                <img
+                  src={getCreatorIconPathForModelId(modelType)}
+                  alt="Model creator logo"
+                  className="h-4 w-4 invert"
+                />
                 <span>{getModelDisplayName(modelType)}</span>
               </>
             }

frontend/libs/components/promptbox/src/lib/PromptBoxVideo.tsx

@@ -19,6 +19,7 @@ import { FontAwesomeIcon } from "@fortawesome/react-fontawesome";
 import { GalleryItem, GalleryModal } from "@storyteller/ui-gallery-modal";
 import {
   CommonResolution,
+  effectivePromptMaxLength,
   SizeIconOption,
   SizeOption,
   VideoModel,
@@ -697,7 +698,12 @@ export const PromptBoxVideo = ({
     mentionAnchorRef.current = null;
   };
 
-  const maxLen = selectedModel?.maxPromptLength ?? 1000;
+  const maxLen =
+    effectivePromptMaxLength(
+      selectedModel?.tauriId ?? "",
+      selectedModel?.maxPromptLength,
+      prompt,
+    ) ?? 1000;
 
   const handleEnqueue = async () => {
     if (!prompt.trim()) {

frontend/libs/model-list/src/lib/classes/metadata/ModelMapping.ts

@@ -2,165 +2,31 @@ import React from "react";
 import { ModelCreator } from "./ModelCreator.js";
 import { getCreatorIcon } from "./ModelCreatorIcons.js";
 
-// Map from model type strings to ModelCreator enum values
-export const MODEL_TYPE_TO_CREATOR: Record<string, ModelCreator> = {
-  flux_1_dev: ModelCreator.BlackForestLabs,
-  flux_1_schnell: ModelCreator.BlackForestLabs,
-  flux_pro_1p1: ModelCreator.BlackForestLabs,
-  flux_pro_1p1_ultra: ModelCreator.BlackForestLabs,
-  flux_pro_kontext_max: ModelCreator.BlackForestLabs,
-  flux_dev_juggernaut: ModelCreator.BlackForestLabs,
-  flux_pro_1: ModelCreator.BlackForestLabs,
-  // Aliases commonly returned by other services
-  flux_pro_1_1: ModelCreator.BlackForestLabs,
-  flux_pro_1_1_ultra: ModelCreator.BlackForestLabs,
-  gpt_image_1: ModelCreator.OpenAi,
-  gpt_image_1p5: ModelCreator.OpenAi,
-  gpt_image_2: ModelCreator.OpenAi,
-  // Kling — p-style IDs
-  kling_1p6_pro: ModelCreator.Kling,
-  kling_2p1_pro: ModelCreator.Kling,
-  kling_2p1_master: ModelCreator.Kling,
-  kling_2p5_turbo_pro: ModelCreator.Kling,
-  kling_2p6_pro: ModelCreator.Kling,
-  kling_3p0_standard: ModelCreator.Kling,
-  kling_3p0_pro: ModelCreator.Kling,
-  // Kling — dot-normalized aliases (backend sends e.g. "kling_1.6_pro", normalizeModelKey converts . → _)
-  kling_1_6_pro: ModelCreator.Kling,
-  kling_2_1_pro: ModelCreator.Kling,
-  kling_2_1_master: ModelCreator.Kling,
-  // Seedance — p-style IDs
-  seedance_1p0_lite: ModelCreator.Bytedance,
-  seedance_1p5_pro: ModelCreator.Bytedance,
-  seedance_2p0: ModelCreator.Bytedance,
-  seedance_2p0_fast: ModelCreator.Bytedance,
-  seedance_2p0_bp: ModelCreator.Bytedance,
-  seedance_2p0_bp_fast: ModelCreator.Bytedance,
-  seedance_2p0_u: ModelCreator.Bytedance,
-  seedance_2p0_u_fast: ModelCreator.Bytedance,
-  seedance_2p0_bpu: ModelCreator.Bytedance,
-  seedance_2p0_bpu_fast: ModelCreator.Bytedance,
-  // Seedance — dot-normalized aliases
-  seedance_1_0_lite: ModelCreator.Bytedance,
-  // Seedream
-  seedream_4: ModelCreator.Bytedance,
-  seedream_4p5: ModelCreator.Bytedance,
-  seedream_5_lite: ModelCreator.Bytedance,
-  // Sora
-  sora_2: ModelCreator.OpenAi,
-  sora_2_pro: ModelCreator.OpenAi,
-  // Veo
-  veo_2: ModelCreator.Google,
-  veo_3: ModelCreator.Google,
-  veo_3_fast: ModelCreator.Google,
-  veo_3p1: ModelCreator.Google,
-  veo_3p1_fast: ModelCreator.Google,
-  gemini_25_flash: ModelCreator.Google,
-  nano_banana: ModelCreator.Google,
-  nano_banana_2: ModelCreator.Google,
-  nano_banana_pro: ModelCreator.Google,
-  recraft_3: ModelCreator.Recraft,
-  // Hunyuan — p-style IDs
-  hunyuan_3d: ModelCreator.Tencent,
-  hunyuan_3d_2p0: ModelCreator.Tencent,
-  hunyuan_3d_2p1: ModelCreator.Tencent,
-  hunyuan_3d_2: ModelCreator.Tencent,
-  hunyuan_3d_3: ModelCreator.Tencent,
-  // Hunyuan — dot-normalized aliases
-  hunyuan_3d_2_0: ModelCreator.Tencent,
-  hunyuan_3d_2_1: ModelCreator.Tencent,
-  worldlabs_gaussian: ModelCreator.WorldLabs,
-  marble_0p1_mini: ModelCreator.WorldLabs,
-  marble_0p1_plus: ModelCreator.WorldLabs,
-  grok_image: ModelCreator.Grok,
-  grok_imagine_image: ModelCreator.Grok,
-  grok_imagine_image_q: ModelCreator.Grok,
-  grok_video: ModelCreator.Grok,
-  grok_imagine_video: ModelCreator.Grok,
-  grok_imagine_video_1p5: ModelCreator.Grok,
-  midjourney: ModelCreator.Midjourney,
-  midjourney_v6: ModelCreator.Midjourney,
-  midjourney_v6p1: ModelCreator.Midjourney,
-  midjourney_v6p1_raw: ModelCreator.Midjourney,
-  midjourney_v7: ModelCreator.Midjourney,
-  midjourney_v7_raw: ModelCreator.Midjourney,
-  midjourney_v7_draft_raw: ModelCreator.Midjourney,
-  // Underscore-numbered IDs the backend actually sends for current MJ models
-  midjourney_7: ModelCreator.Midjourney,
-  midjourney_7_niji: ModelCreator.Midjourney,
-  midjourney_8: ModelCreator.Midjourney,
-  // Angles
-  flux_2_lora_angles: ModelCreator.BlackForestLabs,
-  qwen_edit_2511_angles: ModelCreator.Alibaba,
-  // Happy Horse (Alibaba wanvideo)
-  happy_horse_1p0: ModelCreator.Alibaba,
-  // Beeble (Background Change / VFX)
-  switch_x: ModelCreator.Beeble,
-};
-
-// Get creator icon for a model type
+// Normalize a model id/type key for lookups (lowercase, dots → underscores)
 const normalizeModelKey = (modelType: string): string =>
   modelType.toLowerCase().replace(/\./g, "_").trim();
 
-export const getModelCreatorIcon = (
-  modelType: string,
-): React.ReactNode | null => {
-  const creator = MODEL_TYPE_TO_CREATOR[normalizeModelKey(modelType)];
-  if (!creator) return null;
-  return getCreatorIcon(creator, "h-4 w-4 invert");
-};
+/** Matches all seedance_* variants (video-only; image models are seedream_*). */
+export const isSeedanceVideoModelId = (modelId: string): boolean =>
+  normalizeModelKey(modelId).startsWith("seedance");
 
-// Get creator name for display
-export const getModelCreatorName = (modelType: string): string | null => {
-  const creator = MODEL_TYPE_TO_CREATOR[normalizeModelKey(modelType)];
+const CHINESE_CHARACTERS_REGEX = /\p{Script=Han}/u;
 
-  // Convert enum value to display name
-  switch (creator) {
-    case ModelCreator.BlackForestLabs:
-      return "Black Forest Labs";
-    case ModelCreator.OpenAi:
-      return "OpenAI";
-    case ModelCreator.Kling:
-      return "Kling AI";
-    case ModelCreator.Bytedance:
-      return "ByteDance";
-    case ModelCreator.Google:
-      return "Google";
-    case ModelCreator.Midjourney:
-      return "Midjourney";
-    case ModelCreator.Stability:
-      return "Stability AI";
-    case ModelCreator.Runway:
-      return "Runway";
-    case ModelCreator.Hailuo:
-      return "Hailuo";
-    case ModelCreator.Recraft:
-      return "Recraft";
-    case ModelCreator.Tencent:
-      return "Tencent";
-    case ModelCreator.Alibaba:
-      return "Alibaba";
-    case ModelCreator.Vidu:
-      return "Vidu";
-    case ModelCreator.Fal:
-      return "Fal";
-    case ModelCreator.Replicate:
-      return "Replicate";
-    case ModelCreator.TensorArt:
-      return "TensorArt";
-    case ModelCreator.OpenArt:
-      return "OpenArt";
-    case ModelCreator.Higgsfield:
-      return "Higgsfield";
-    case ModelCreator.Krea:
-      return "Krea";
-    case ModelCreator.Grok:
-      return "Grok";
-    case ModelCreator.WorldLabs:
-      return "World Labs";
-    default:
-      return creator;
+/**
+ * Effective soft prompt-length limit for a model. Normally the model's
+ * `text_prompt_max_length`, but seedance waives the limit (Infinity) when the
+ * prompt contains Chinese characters: the server counts CJK characters as
+ * more than one unit, so a client-side character count would false-positive.
+ */
+export const effectivePromptMaxLength = (
+  modelId: string,
+  modelMaxLength: number | undefined,
+  prompt: string,
+): number | undefined => {
+  if (isSeedanceVideoModelId(modelId) && CHINESE_CHARACTERS_REGEX.test(prompt)) {
+    return Infinity;
   }
+  return modelMaxLength;
 };
 
 // Convert model type string to human-readable display name. An API-provided