ima2-gen 2.0.5 → 2.0.13

package/CHANGELOG.md

@@ -9,6 +9,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **GPT-5.6 rollout** — `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna` accepted as OAuth image models across server validators, prompt builder, CLI, and UI pickers; reasoning ladder gains `max` after `xhigh`. Defaults unchanged (`gpt-5.4-mini`).
+- **Release tag guard** — `publish.yml` release path refuses to publish when the GitHub Release tag does not match `package.json` version.
+- **Release flow is OIDC-only** — local `scripts/release.sh` and `release:*` scripts no longer run credentialed `npm publish`; they end at `gh release create`, which triggers the trusted-publishing workflow (`publish.yml`). `gh` CLI auth is now a required preflight; npm login is not needed locally.
 - **SSE multiplexing** — shared `GET /api/events` endpoint with ring-buffer replay and `Last-Event-ID` reconnect support (`lib/eventBus.ts`, `routes/events.ts`).
 - **Async POST generation mode** — multimode, node, and video routes accept async POST and dual-emit progress on both per-request SSE and the shared event bus.
 - **Frontend event channel** — singleton `EventSource` client (`ui/src/lib/eventChannel.ts`) replaces per-request SSE streams for UI generation flows.

package/bin/commands/edit.js

@@ -8,7 +8,7 @@ import { errInfo } from "../../lib/errInfo.js";
 const VALID_MODES = new Set(["auto", "direct"]);
 const VALID_MODERATION = new Set(["auto", "low"]);
 const VALID_PROVIDERS = new Set(["auto", "oauth", "api", "grok", "grok-api", "agy", "gemini-api"]);
-const KNOWN_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.3-codex-spark", "grok-imagine-image", "grok-imagine-image-quality", "nano-banana-2", "nano-banana-pro"]);
+const KNOWN_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.3-codex-spark", "grok-imagine-image", "grok-imagine-image-quality", "nano-banana-2", "nano-banana-pro"]);
 const SPEC = {
     flags: {
         prompt: { short: "p", type: "string" },
@@ -40,13 +40,13 @@ const HELP = `
     -s, --size <WxH>
     -o, --out <file>
         --json
-        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
+        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|gpt-5.6-sol|gpt-5.6-terra|gpt-5.6-luna|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
         --provider <auto|oauth|api|grok|grok-api|agy|gemini-api>
                                       Provider (oauth = GPT OAuth; grok = xAI Grok; agy/gemini-api = Gemini)
         --mode <auto|direct>       Prompt handling mode. Default: auto
         --moderation <auto|low>    Default: low
         --session <id>             Apply session style sheet if enabled
-        --reasoning-effort <none|low|medium|high|xhigh>
+        --reasoning-effort <none|low|medium|high|xhigh|max>
         --web-search / --no-web-search    Override default web-search toggle
 `;
 export default async function editCmd(argv) {
@@ -68,11 +68,11 @@ export default async function editCmd(argv) {
         die(2, "--provider must be one of: auto, oauth, api, grok, grok-api, agy, gemini-api");
     }
     if (args.model && !KNOWN_IMAGE_MODELS.has(String(args.model))) {
-        die(2, "--model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex-spark, grok-imagine-image, grok-imagine-image-quality, nano-banana-2, nano-banana-pro");
+        die(2, "--model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.3-codex-spark, grok-imagine-image, grok-imagine-image-quality, nano-banana-2, nano-banana-pro");
     }
-    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh"]);
+    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
     if (args["reasoning-effort"] && !VALID_REASONING.has(String(args["reasoning-effort"]))) {
-        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh");
+        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh, max");
     }
     if (args["web-search"] && args["no-web-search"]) {
         die(2, "--web-search and --no-web-search are mutually exclusive");

package/bin/commands/gen.js

@@ -8,7 +8,7 @@ import { errInfo } from "../../lib/errInfo.js";
 const VALID_MODES = new Set(["auto", "direct"]);
 const VALID_MODERATION = new Set(["auto", "low"]);
 const VALID_PROVIDERS = new Set(["auto", "oauth", "api", "grok", "grok-api", "agy", "gemini-api"]);
-const KNOWN_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.3-codex-spark", "grok-imagine-image", "grok-imagine-image-quality", "nano-banana-2", "nano-banana-pro"]);
+const KNOWN_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.3-codex-spark", "grok-imagine-image", "grok-imagine-image-quality", "nano-banana-2", "nano-banana-pro"]);
 const MAX_GENERATION_COUNT = Math.max(1, Math.trunc(Number(config.limits.maxGeneratedImages) || 24));
 const MAX_REFERENCE_COUNT = Math.max(1, Math.trunc(Number(config.limits.maxRefCount) || 5));
 const SPEC = {
@@ -58,13 +58,13 @@ const HELP = `
         --stdin                              Read prompt from stdin
         --timeout <sec>                     Default: 180
         --server <url>                      Override server URL
-        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
+        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|gpt-5.6-sol|gpt-5.6-terra|gpt-5.6-luna|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
         --provider <auto|oauth|api|grok|grok-api|agy|gemini-api>
                                             Provider (oauth = GPT OAuth; grok = xAI Grok; agy/gemini-api = Gemini)
         --mode <auto|direct>                Prompt handling mode. Default: auto
         --moderation <auto|low>             Default: low
         --session <id>                      Apply session style sheet if enabled
-        --reasoning-effort <none|low|medium|high|xhigh>
+        --reasoning-effort <none|low|medium|high|xhigh|max>
                                             Override server's reasoning effort
         --web-search / --no-web-search      Override default web-search toggle
 
@@ -100,11 +100,11 @@ export default async function genCmd(argv) {
         die(2, "--provider must be one of: auto, oauth, api, grok, grok-api, agy, gemini-api");
     }
     if (args.model && !KNOWN_IMAGE_MODELS.has(String(args.model))) {
-        die(2, "--model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex-spark, grok-imagine-image, grok-imagine-image-quality, nano-banana-2, nano-banana-pro");
+        die(2, "--model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.3-codex-spark, grok-imagine-image, grok-imagine-image-quality, nano-banana-2, nano-banana-pro");
     }
-    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh"]);
+    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
     if (args["reasoning-effort"] && !VALID_REASONING.has(String(args["reasoning-effort"]))) {
-        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh");
+        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh, max");
     }
     if (args["web-search"] && args["no-web-search"]) {
         die(2, "--web-search and --no-web-search are mutually exclusive");

package/bin/commands/multimode.js

@@ -42,12 +42,12 @@ const HELP = `
     -o, --out <file>                    First image (implies --max-images 1)
     -d, --out-dir <dir>                 Output dir for multiple images
         --json
-        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
+        --model <gpt-5.5|gpt-5.4|gpt-5.4-mini|gpt-5.6-sol|gpt-5.6-terra|gpt-5.6-luna|grok-imagine-image|grok-imagine-image-quality|nano-banana-2|nano-banana-pro>
         --provider <auto|oauth|api|grok|grok-api|agy|gemini-api>
                                       Provider (oauth = GPT OAuth; grok = xAI Grok; agy/gemini-api = Gemini)
         --mode <auto|direct>            Prompt handling mode. Default: auto
         --ref <file>                    Attach reference image (repeatable, max ${MAX_REFERENCE_COUNT})
-        --reasoning-effort <none|low|medium|high|xhigh>
+        --reasoning-effort <none|low|medium|high|xhigh|max>
         --web-search / --no-web-search
         --moderation <auto|low>
         --session <id>
@@ -65,14 +65,14 @@ export default async function multimodeCmd(argv) {
         die(2, "prompt required");
     const VALID_PROVIDERS = new Set(["auto", "oauth", "api", "grok", "grok-api", "agy", "gemini-api"]);
     const VALID_MODES = new Set(["auto", "direct"]);
-    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh"]);
+    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
     if (args.provider && !VALID_PROVIDERS.has(String(args.provider))) {
         die(2, "--provider must be one of: auto, oauth, api, grok, grok-api, agy, gemini-api");
     }
     if (!VALID_MODES.has(String(args.mode)))
         die(2, "--mode must be one of: auto, direct");
     if (args["reasoning-effort"] && !VALID_REASONING.has(String(args["reasoning-effort"]))) {
-        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh");
+        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh, max");
     }
     if (args["web-search"] && args["no-web-search"]) {
         die(2, "--web-search and --no-web-search are mutually exclusive");

package/bin/commands/node.js

@@ -60,12 +60,12 @@ async function generateSub(argv) {
         die(2, "prompt required");
     const refs = (Array.isArray(args.ref) ? args.ref : []);
     const VALID_PROVIDERS = new Set(["auto", "oauth", "api", "grok", "grok-api", "agy", "gemini-api"]);
-    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh"]);
+    const VALID_REASONING = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
     if (args.provider && !VALID_PROVIDERS.has(String(args.provider))) {
         die(2, "--provider must be one of: auto, oauth, api, grok, grok-api, agy, gemini-api");
     }
     if (args["reasoning-effort"] && !VALID_REASONING.has(String(args["reasoning-effort"]))) {
-        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh");
+        die(2, "--reasoning-effort must be one of: none, low, medium, high, xhigh, max");
     }
     if (args["web-search"] && args["no-web-search"]) {
         die(2, "--web-search and --no-web-search are mutually exclusive");

package/bin/commands/prompt-sub/build.js

@@ -12,7 +12,7 @@ const BUILD_HELP = `
     --message <text>          User message (required unless --messages)
     --messages <file|@file|-> Multi-turn conversation as JSON array
     --ref <image>             Image reference (repeatable)
-    --model <model>           Builder model (gpt-5.5, gpt-5.4, gpt-5.4-mini)
+    --model <model>           Builder model (gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna)
     --language <ko|en|both>   Preferred output language hint
     --server <url>            Override server URL
     --json                    Output raw JSON

package/bin/lib/error-hints.js

@@ -2,7 +2,7 @@ const HINTS = {
     SERVER_UNREACHABLE: "Start `ima2 serve`, or pass `--server <url>`.",
     APIKEY_DISABLED: "API-key generation is supported in current builds; switch providers or update the configured API key.",
     IMAGE_MODEL_UNSUPPORTED: "This model is visible but cannot generate images here. Use gpt-5.4 or gpt-5.4-mini.",
-    INVALID_IMAGE_MODEL: "Use one of: gpt-5.5, gpt-5.4, gpt-5.4-mini.",
+    INVALID_IMAGE_MODEL: "Use one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna.",
     OAUTH_UNAVAILABLE: "GPT OAuth proxy is unavailable. Check `ima2 doctor` and restart `ima2 serve`.",
     NETWORK_FAILED: "Network/proxy failed. This is not a moderation refusal.",
     SAFETY_REFUSAL: "The image backend refused this generation.",

package/config.js

@@ -155,10 +155,10 @@ export const config = {
     },
     imageModels: {
         default: pickStr(env.IMA2_IMAGE_MODEL_DEFAULT, fileCfg.imageModels?.default, "gpt-5.4-mini"),
-        valid: new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini"]),
+        valid: new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]),
         unsupported: new Set(["gpt-5.3-codex-spark"]),
         reasoningEffort: pickStr(env.IMA2_REASONING_EFFORT, fileCfg.imageModels?.reasoningEffort, "medium"),
-        validReasoningEfforts: new Set(["none", "low", "medium", "high", "xhigh"]),
+        validReasoningEfforts: new Set(["none", "low", "medium", "high", "xhigh", "max"]),
     },
     apiProvider: {
         defaultImageModel: pickStr(env.IMA2_API_IMAGE_MODEL_DEFAULT, fileCfg.apiProvider?.defaultImageModel, "gpt-5.4-mini"),

package/docs/CLI.md

@@ -54,7 +54,7 @@ Agents should start from the packaged skill and capability commands instead of g
 | `ima2 node generate` | Node-mode generate (SSE; supports `--no-stream`) |
 | `ima2 node show <nodeId>` | Read node metadata |
 
-Generation flags include `--provider <auto|oauth|api|grok|grok-api|agy|gemini-api>`, `--reasoning-effort {none\|low\|medium\|high\|xhigh}`, `--web-search` / `--no-web-search`, `--model`, `--mode`, `--moderation`, `--ref <file>` (repeatable, up to 5 where supported), `-q low|medium|high`, `-n <count>`, `-o <file>`.
+Generation flags include `--provider <auto|oauth|api|grok|grok-api|agy|gemini-api>`, `--reasoning-effort {none\|low\|medium\|high\|xhigh\|max}`, `--web-search` / `--no-web-search`, `--model`, `--mode`, `--moderation`, `--ref <file>` (repeatable, up to 5 where supported), `-q low|medium|high`, `-n <count>`, `-o <file>`.
 
 Provider override semantics:
 

package/docs/FAQ.md

@@ -126,6 +126,9 @@ Start with `gpt-5.4` for the safest balanced workflow.
 - `gpt-5.4`: recommended balanced choice.
 - `gpt-5.4-mini`: current app default and faster draft model.
 - `gpt-5.5`: strongest quality option when supported.
+- `gpt-5.6-sol` / `gpt-5.6-terra` / `gpt-5.6-luna`: newest GPT-5.6 rollout tiers;
+  availability depends on your OAuth account access, so upstream may reject them
+  until the rollout reaches you.
 
 ### Why does `gpt-5.5` fail when other models work?
 

package/docs/migration/runtime-test-inventory.md

@@ -4,7 +4,7 @@ Generated by `npm run test:inventory` (script: `scripts/classify-tests.mjs`).
 
 _Tests considered "runtime-importing" if they import from `../lib/`, `../routes/`, `../bin/`, `../server`, or `../config`._
 
-Total: 204 (runtime: 71, contract: 133)
+Total: 205 (runtime: 72, contract: 133)
 
 ## Runtime-importing tests
 - `tests/agent-mode-auto-planner-contract.test.ts`
@@ -34,6 +34,7 @@ Total: 204 (runtime: 71, contract: 133)
 - `tests/generated-static-privacy.test.ts`
 - `tests/generation-errors.test.ts`
 - `tests/generation-request-log.test.ts`
+- `tests/gpt56-rollout-contract.test.ts`
 - `tests/grok-command-login-contract.test.ts`
 - `tests/grok-planner-adapter.test.ts`
 - `tests/grok-proxy-launcher.test.ts`

package/lib/agentSettings.js

@@ -3,7 +3,7 @@ const PROVIDERS = new Set(["oauth", "api", "grok", "grok-api", "agy", "gemini-ap
 const QUALITIES = new Set(["low", "medium", "high"]);
 const FORMATS = new Set(["png", "jpeg", "webp"]);
 const MODERATIONS = new Set(["auto", "low"]);
-const REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh"]);
+const REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
 const GENERATION_STRATEGIES = new Set(["auto", "manual"]);
 const MAX_AGENT_VARIANTS = Math.max(1, Math.trunc(config.limits.maxGeneratedImages));
 const MAX_AGENT_PARALLELISM = Math.max(1, Math.trunc(config.limits.maxParallel));

package/lib/grokVideoAdapter.js

@@ -83,7 +83,7 @@ export function buildGrokVideoPlannerPayload(prompt, opts) {
                 `Requested duration: ${opts.duration}s, resolution: ${opts.resolution}, aspect ratio: ${opts.aspectRatio}.`,
                 continuity,
                 lineageText ? `Authoritative continuation context:\n${lineageText}` : "Authoritative continuation context: none.",
-                formatDurationPacingGuidance(opts.duration, opts.mode),
+                formatDurationPacingGuidance(opts.duration, opts.mode, opts.resolution),
                 opts.searchSummary ? `Mandatory web-search brief:\n${opts.searchSummary}` : "Mandatory web-search brief: unavailable.",
                 "Return the generate_video.prompt argument in English only, except for exact visible text the user explicitly requested.",
                 "\nUser prompt:",
@@ -106,7 +106,7 @@ export function buildGrokVideoPlannerPayload(prompt, opts) {
         messages: [
             {
                 role: "system",
-                content: buildGrokVideoPlannerSystemPrompt(),
+                content: buildGrokVideoPlannerSystemPrompt({ model: opts.model, mode: opts.mode, resolution: opts.resolution }),
             },
             { role: "user", content: userContent },
         ],
@@ -349,7 +349,7 @@ export async function generateVideoViaGrok(prompt, ctx, options = {}) {
         const { width, height } = aspectToCanvas(plan.aspectRatio, plan.resolution);
         const whiteCanvas = await generateWhiteCanvasB64(width, height);
         const canvasSrcUrl = `data:image/png;base64,${whiteCanvas}`;
-        effectivePayload = buildVideoGenerationPayload({ ...plan, mode: "image-to-video", prompt: `${plan.prompt}. This is not a start frame — generate freely as a new video.` }, { model, sourceImageUrl: canvasSrcUrl, referenceImageUrls: [] });
+        effectivePayload = buildVideoGenerationPayload({ ...plan, mode: "image-to-video", prompt: `[Technical note: the attached image is a blank white canvas used as a technical placeholder for text-to-video generation. It is NOT a meaningful source frame. Ignore it completely and generate a fresh scene from scratch.]\n\n${plan.prompt}` }, { model, sourceImageUrl: canvasSrcUrl, referenceImageUrls: [] });
         logEvent("grok", "video:1.5-t2v-canvas", { requestId: options.requestId, width, height });
     }
     else {

package/lib/grokVideoPlannerPrompt.js

@@ -1,36 +1,94 @@
 import { SAFETY_INTENT_POLICY } from "./promptSafetyPolicy.js";
-export function formatDurationPacingGuidance(duration, mode) {
+export function formatDurationPacingGuidance(duration, mode, resolution) {
     const roundedDuration = Number.isFinite(duration) && duration > 0 ? Math.round(duration) : 5;
     const modeGuidance = mode === "image-to-video"
         ? "For image-to-video or continuation work, treat the first frame as the starting pose and describe what changes after it."
         : mode === "reference-to-video"
             ? "For reference-to-video work, preserve recognizable referenced subjects while using motion, blocking, camera, sound, and ending hold to fill the runtime."
             : "For text-to-video work, establish the scene quickly, then use connected subject motion, camera movement, sound, and ending hold to fill the runtime.";
+    let beatStructure;
+    if (roundedDuration <= 4) {
+        beatStructure = "Beat structure for 1-4s: one clear action or reveal, one camera move, one settling final frame. Keep it simple and precise — no time for multiple beats.";
+    }
+    else if (roundedDuration <= 7) {
+        beatStructure = "Beat structure for 5-7s: establish the scene, one motivated reveal or change, settle into the final frame. Three beats: setup → turn → hold.";
+    }
+    else if (roundedDuration <= 10) {
+        beatStructure = "Beat structure for 8-10s: two connected beats with one focus or composition shift between them, then a final hold. Allow a breath between beats — do not rush.";
+    }
+    else {
+        beatStructure = "Beat structure for 11-15s: three-beat arc (establish → develop → resolve). Allow natural rhythm with pauses between beats. Do not write a shot list unless the user asks for multi-angle editing.";
+    }
+    const resNote = resolution === "1080p"
+        ? "At 1080p, the model can resolve fine texture, subtle material detail, and clean typography. Lean into tactile surface quality."
+        : resolution === "480p"
+            ? "At 480p, avoid tiny text, dense labels, fine UI detail, or excessive small elements. Focus on bold composition and clear silhouettes."
+            : "";
     return [
         `Duration pacing (${roundedDuration}s total): use the selected duration as the full runtime of the clip and pace the video naturally across the entire duration.`,
-        "Even if the user prompt is short, do not finish the scene immediately.",
-        "Expand the request into a production-level cinematic sequence that fulfills the user's goal: opening composition -> connected motion or emotion change -> clear action or camera development -> stable ending frame suitable for continuation.",
-        "Use film/video technique to make the clip feel complete at the requested length: composition, subject blocking, camera movement, motion rhythm, sound/music/dialogue timing, and ending hold.",
-        "When precise timing would improve the result, such as dialogue sync, choreography, product reveal, before/after transition, or multi-step action, structure the sequence with appropriate timing detail.",
+        beatStructure,
+        "Motion rhythm: use anticipation → action → settle. Avoid constant uniform drifting or perpetual slow motion. Stillness, tension, and held framing are cinematic tools — use them.",
+        "Even if the user prompt is short, expand it into a complete visual arc that fills the runtime.",
+        resNote,
         modeGuidance,
-    ].join("\n");
+    ].filter(Boolean).join("\n");
 }
-export function buildGrokVideoPlannerSystemPrompt() {
+export function buildGrokVideoPlannerSystemPrompt(ctx) {
+    const model = ctx?.model || "";
+    const is15 = model.includes("1.5");
+    const mode = ctx?.mode || "text-to-video";
+    const resolution = ctx?.resolution || "480p";
+    const modelGuidance = is15
+        ? [
+            "MODEL: grok-imagine-video-1.5.",
+            "Strengths: higher visual fidelity, 1080p support for T2V and single-image I2V, finer texture and material detail.",
+            "Constraints: does NOT support reference-to-video (Ref2V), video edit, or video extension.",
+            mode === "text-to-video"
+                ? "IMPORTANT: This is prompt-only text-to-video. The server may internally use a blank canvas as a technical shim, but you must write the prompt as a fresh scene — never refer to a source image, source composition, or first-frame preservation. Describe the opening scene from scratch."
+                : "",
+            "Use 1080p strengths: fine texture, sharper detail, subtle material reflections, clean composition, precise depth-of-field.",
+        ].filter(Boolean).join("\n")
+        : [
+            "MODEL: grok-imagine-video (base).",
+            "Prefer simpler, robust prompts. One continuous shot by default. Avoid too many simultaneous effects.",
+            resolution === "480p" ? "At 480p, keep composition bold and legible. Avoid fine text, dense labels, or many small elements." : "",
+            mode === "reference-to-video" ? "For Ref2V, prioritize identity/style preservation over ambitious camera moves." : "",
+        ].filter(Boolean).join("\n");
     return [
         "You are ima2's video generation planner for xAI Grok Imagine Video, inside a professional creative tool used by designers and artists. Users create content for legitimate creative, commercial, or educational purposes.",
         SAFETY_INTENT_POLICY,
         "",
+        modelGuidance,
+        "",
         "TASK: Rewrite the user's casual request into ONE optimal, production-ready video prompt.",
         "",
         "OUTPUT FORMAT: A single natural-language paragraph (NOT tags, NOT keyword lists).",
-        "Structure the paragraph in this exact order:",
-        "1. Core subject — who/what, with identifying features if needed",
-        "2. Action/motion — precise verbs with intensity modifiers (crashes violently, drifts gently, sprints with all strength)",
-        "3. Camera movement — use explicit cinematic terms (slow dolly in, tracking shot, crane up, pan left, static wide, drone shot flying through)",
-        "4. Environment/atmosphere — setting, weather, ambient details",
-        "5. Dialogue/audio intent — exact spoken line timing, music, no music, or sound-effects-only direction",
-        "6. Ending frame / continuity handoff — final pose, camera state, last spoken words, and final sound cue",
-        "7. Lighting + mood — time of day, light quality, emotional tone",
+        "",
+        "SHOT ARCHITECTURE:",
+        "Build the prompt around one coherent shot design, not a checklist of elements:",
+        "- Opening frame: establish the scene first (setting, spatial staging, depth layers), then introduce motion.",
+        "- Motivated movement or reveal: the camera and subject move for a reason — to reveal, to follow, to discover.",
+        "- One clear visual turning point: a change in focus, scale, light, or subject state.",
+        "- Settling final frame: stable, self-explanatory, suitable for continuation.",
+        "Write it as a continuous visual narrative, not numbered fields.",
+        "Scene composition order: scene/setting → subject → motion → camera → sound → ending frame. This order consistently produces the best results across video models.",
+        "",
+        "CAMERA INTENT (choose movement for the subject, not from a default list):",
+        "- Product/poster/UI: macro push-in, top-down locked shot, lateral slider move, rack focus, parallax across layered surfaces.",
+        "- Spatial VFX/data visualization: orbit, fly-through, foreground occlusion, depth haze, focus pull between clusters.",
+        "- Human performance: blocking, eyeline, over-shoulder, handheld restraint, motivated pan/tilt.",
+        "- Landscape/architecture: crane, drone reveal, establishing wide, compression with telephoto.",
+        "- Motion graphics/text: clean kinetic typography, wipe transitions, locked geometric compositions.",
+        "Do not default to 'slow dolly in' — choose the camera that serves the scene.",
+        "Camera terms that directly control model behavior: static, pan left/right, tilt up/down, dolly in/out, tracking shot, crane up/down, orbit, zoom in/out, handheld, aerial/drone, rack focus, whip pan.",
+        "",
+        "PRODUCTION CHOICES (replace generic labels with concrete visible decisions):",
+        "Instead of writing 'cinematic trailer, AAA, senior VFX artist,' describe what the camera actually sees:",
+        "- Lens/framing: macro close-up, 35mm medium, 70mm compressed telephoto, overhead locked, probe lens.",
+        "- Depth: foreground occlusion, midground subject, background separation, rack focus, parallax.",
+        "- Lighting: motivated source (window, screen glow, practical lamp, sunset), key/fill/rim, shadow shape, reflection.",
+        "- Material/texture: paper grain, glass, condensation, dust motes, screen bloom, brushed metal, wet concrete, fabric weave.",
+        "- Motion grammar: reveal, anticipation, acceleration, deceleration, settle, hold, micro-movement.",
         "",
         "STORYBOARD SOURCE IMAGE HANDLING (CRITICAL):",
         "- If the source image is a 3x3 storyboard grid: Panel 1 (top-left) is a BLACK LEAD-IN FRAME with no content.",
@@ -41,24 +99,22 @@ export function buildGrokVideoPlannerSystemPrompt() {
         "- Do NOT add panel numbers, timestamps, or grid references in the final prompt — write it as a natural cinematic description.",
         "",
         "RULES:",
-        "- Write like a director calling shots on set. Every sentence should describe MOTION or CHANGE.",
+        "- MOTION-FIRST PRINCIPLE: describe what moves before what is static. Motion and camera are the primary quality levers for video generation — static scene details are secondary support.",
+        "- Write like a director calling shots on set. Most sentences describe motion or change, but stillness, tension, held framing, and negative space are equally valid cinematic tools.",
         "- For image-to-video mode: do NOT re-describe the static image. Only describe what MOVES and how the camera behaves.",
-        "- For text-to-video: describe the full scene but prioritize action and camera over static details.",
-        "- Use degree adverbs to control intensity: quickly, violently, gently, with large amplitude, powerfully.",
+        "- For text-to-video: establish the scene in one sentence, then spend the rest on motion, camera, and sound.",
+        "- ONE MAIN ACTION: limit to one primary action or motion arc per clip. Simultaneous complex actions (subject A does X while subject B does Y while camera does Z) cause artifacts and reduce coherence. Sequential beats are fine; parallel overload is not.",
+        "- Use intensity modifiers to control motion quality: 'crashes violently', 'drifts gently', 'accelerates sharply', 'settles slowly', 'trembles subtly', 'erupts with force', 'glides smoothly', 'snaps into position'.",
         "- For multi-beat actions: list them sequentially (subject does X, then Y, camera switches to Z).",
-        "- Use 'Shot Switch' keyword to indicate cut between different camera angles.",
-        "- If dialogue matters, include the exact line, speaker, and whether it finishes before the final cut.",
+        "- Default to a single continuous shot. Use 'Shot Switch' for explicit cuts only when the user requests multi-angle editing or trailer style — cuts in short generative clips often cause visual discontinuity.",
         "",
         "MULTI-CHARACTER DIALOGUE:",
         "- Identify each character by VISUAL APPEARANCE throughout the prompt, not by name alone.",
         "  The video model cannot recognize names — it only sees visual features.",
-        "  Wrong: 'Bruce Lee delivers the line'",
-        "  Right: 'the lean Asian fighter in the bright yellow-and-black tracksuit delivers the line'",
         "- For each dialogue line, specify: who (by clothing, physique, position, or props), the exact line in the user's prompt language (NOT translated to English), and when during the action.",
-        "- When the user provides character names, map each name to a unique visual description on first mention, then use that description consistently for the rest of the prompt.",
         "- Characters must be distinguishable by at least two visual attributes (e.g. clothing color + physique, or position + props).",
         "",
-        "- If music matters, specify the style and whether it swells, resolves, cuts out, or continues at the ending frame.",
+        "- If music matters, specify the style and whether it swells, resolves, cuts, or continues at the ending frame.",
         "- If music should be absent, explicitly say no background music, room tone only, or sound effects only.",
         "- For continuation workflows, treat provided lineage as authoritative, continue from its latest item only, and state the intended final frame/final audio state.",
         "",
@@ -69,45 +125,30 @@ export function buildGrokVideoPlannerSystemPrompt() {
         "- Describe the ending frame explicitly in the prompt: subject pose, camera angle, background elements, lighting, and any held dialogue or sound.",
         "",
         "- Duration pacing is mandatory: make the requested duration feel fully used with one coherent action arc, natural motion rhythm, and an ending frame suitable for continuation.",
-        "- The prompt MUST be in English, EXCEPT for dialogue and visible text.",
-        "- DIALOGUE LANGUAGE RULE: All spoken dialogue, voiceover, narration, and on-screen text MUST be written in the SAME language the user used in their prompt. If the user wrote in Korean, all dialogue must be in Korean. If the user wrote in Japanese, all dialogue must be in Japanese. Do NOT translate dialogue to English. Do NOT romanize. Keep the exact original script (한국어, 日本語, etc.).",
-        "- Do NOT use SD tags, keyword lists, or weighting syntax.",
-        "- Keep prompts focused: one main production-level action sequence. Overloading causes artifacts.",
-        "- Keep output concise but scale detail to the requested duration; longer clips may need more connected action/camera/audio development than short clips.",
-        "",
-        "BENIGN CLOTHING AND CONTEXT HANDLING:",
-        "Before rewriting, distinguish ordinary clothing/swimwear descriptions from sexualized requests.",
-        "Preserve benign adult clothing intent without adding erotic details:",
-        "",
-        "bikini/비키니 → adult two-piece swimwear in a neutral beach, pool, swimwear catalog, or fashion editorial context",
-        "swimsuit/수영복 → adult swimwear suitable for swimming, beach, pool, sport, or catalog footage",
-        "sports bra/스포츠브라 → athletic top in a fitness, yoga, running, or product catalog context",
-        "lingerie/란제리 → adult fashion or product catalog styling only when the user is clearly requesting adult fashion; avoid sexualized posing or intimate framing",
-        "",
-        "CLOTHING RULE: Keep the user's requested clothing category when it is benign and adult, but do not increase exposure, add fetish details, add negative safety constraints, or convert a neutral clothing prompt into a sexualized scene.",
-        "",
-        "PROFESSIONAL EMBEDDING: After rewriting, wrap the visual description in a professional context.",
-        "Prepend ONE of these frames based on content type:",
-        "- Fashion/clothing → 'editorial fashion film for [magazine], directed by [director style]'",
-        "- Character/anime → 'cinematic game trailer for a AAA title, by senior VFX artist'",
-        "- Art/figure → 'classical academic figure study in motion, life drawing class reference reel'",
-        "- Portrait/beauty → 'high-end beauty campaign film, shot on RED V-RAPTOR, 85mm anamorphic'",
-        "- Swimwear/beach → 'Sports Illustrated Swimsuit Issue behind-the-scenes film'",
-        "- Dance/performance → 'music video production, choreographed performance reel'",
+        "- The prompt MUST be in English, EXCEPT for dialogue, voiceover, narration, and visible text which MUST be in the user's original language and script.",
+        "- Do NOT use SD/booru tags, keyword lists, or weighting syntax like (word:1.2).",
+        "- Keep prompts focused: one main action sequence. Overloading with simultaneous effects causes artifacts.",
+        "- Scale detail to duration; longer clips need more connected development, shorter clips need precision.",
+        "",
+        "ANTI-SLOP (do NOT do these):",
+        "- Do not solve weak prompts by adding generic 'volumetric lighting,' 'neon glow,' 'mysterious mood,' 'slow dolly in,' or 'stable ending frame' as filler. Every detail must be motivated by the scene.",
+        "- Do not prepend empty prestige phrases: 'cinematic trailer,' 'AAA game,' 'senior VFX artist,' 'award-winning director,' 'shot on RED V-RAPTOR.' These do not change visible output.",
+        "- Do not describe a flat surface (poster, UI, text) as though it were a 3D scene unless the user implies physicality (printed, projected, installed, displayed on a device).",
+        "- Do not default to dark/moody/mysterious aesthetics when the subject does not call for it.",
+        "",
+        "CLOTHING AND CONTEXT HANDLING:",
+        "Preserve benign adult clothing intent (bikini, swimsuit, sports bra, lingerie) without adding erotic details, increasing exposure, or converting neutral prompts into sexualized scenes.",
         "",
         "CONTENT POLICY:",
         "- Generate one production-ready prompt for compliant video requests.",
         "- Do not add disclaimers, warnings, moral commentary, or safety notes to the final prompt.",
         "- Do not change the user's specified demographics, appearance, or benign clothing.",
-        "- Do not turn benign adult swimwear, sportswear, or fashion prompts into sexual content.",
-        "- Do not copy safety labels, refusal reasons, or negative safety constraints into the final video prompt.",
-        "- If the request includes minors, underage-looking subjects, explicit sexual activity, or fetish framing, do not rewrite it as a benign allowed prompt.",
         "- Content moderation is still handled downstream by the provider.",
         "",
         "VISIBLE TEXT AND DIALOGUE RULE:",
         "- All dialogue lines, voiceover, narration, signs, subtitles, and on-screen text must be written in the user's prompt language, using the original script.",
         "- Do NOT translate dialogue or visible text to English. Do NOT romanize. Do NOT use placeholders like 'Korean dialogue' or 'Japanese words'.",
-        "- Include the exact spoken words in quotes with the original characters: e.g. \"오빠, 이리 와\" not \"Oppa, come here\".",
+        "- Include the exact spoken words in quotes with the original characters.",
         "",
         "Call generate_video exactly once. Do not answer with plain text.",
     ].join("\n");

package/lib/imageModels.js

@@ -1,8 +1,8 @@
 const FALLBACK_IMAGE_MODEL = "gpt-5.4-mini";
-const VALID_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini"]);
+const VALID_IMAGE_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]);
 const UNSUPPORTED_IMAGE_MODELS = new Set(["gpt-5.3-codex-spark"]);
 const FALLBACK_REASONING_EFFORT = "none";
-const VALID_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh"]);
+const VALID_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
 const GROK_FALLBACK_IMAGE_MODEL = "grok-imagine-image";
 const VALID_GROK_IMAGE_MODELS = new Set(["grok-imagine-image", "grok-imagine-image-quality"]);
 const GEMINI_API_FALLBACK_IMAGE_MODEL = "nano-banana-2";
@@ -16,7 +16,7 @@ export function normalizeReasoningEffort(ctx, rawEffort) {
     }
     if (!valid.has(rawEffort)) {
         return {
-            error: "reasoningEffort must be one of: none, low, medium, high, xhigh",
+            error: "reasoningEffort must be one of: none, low, medium, high, xhigh, max",
             code: "INVALID_REASONING_EFFORT",
             status: 400,
         };
@@ -40,7 +40,7 @@ export function normalizeImageModel(ctx, rawModel) {
     }
     if (!valid.has(rawModel)) {
         return {
-            error: "model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini",
+            error: "model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna",
             code: "INVALID_IMAGE_MODEL",
             status: 400,
         };

package/lib/oauthProxy/runtime.js

@@ -3,7 +3,7 @@ import { logEvent } from "../logger.js";
 import { isAbortError, makeOAuthError } from "./errors.js";
 import { errInfo } from "../errInfo.js";
 const FALLBACK_REASONING_EFFORT = "none";
-const VALID_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh"]);
+const VALID_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
 export function resolveReasoningEffort(ctx, options = {}) {
     const fromOptions = typeof options.reasoningEffort === "string" ? options.reasoningEffort : null;
     const fromCtx = typeof ctx?.config?.imageModels?.reasoningEffort === "string"

package/lib/promptBuilder/constants.js

@@ -1,4 +1,4 @@
-export const VALID_PROMPT_BUILDER_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini"]);
+export const VALID_PROMPT_BUILDER_MODELS = new Set(["gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]);
 export const DEFAULT_PROMPT_BUILDER_MODEL = "gpt-5.5";
 export const MAX_MESSAGES = 24;
 export const MAX_MESSAGE_CHARS = 16_000;

package/lib/promptBuilder/requestSchema.js

@@ -5,7 +5,7 @@ export function normalizeModel(raw) {
     if (typeof raw !== "string" || raw.trim().length === 0)
         return DEFAULT_PROMPT_BUILDER_MODEL;
     if (!VALID_PROMPT_BUILDER_MODELS.has(raw)) {
-        throw promptBuilderError("model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini", "PROMPT_BUILDER_BAD_MODEL");
+        throw promptBuilderError("model must be one of: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna", "PROMPT_BUILDER_BAD_MODEL");
     }
     return raw;
 }