From ac2ce0b8b6aade69eaffa1fed95cf2d2a6c5bc59 Mon Sep 17 00:00:00 2001 From: JianJang2017 Date: Fri, 13 Mar 2026 19:09:54 -0500 Subject: [PATCH] Add qwen-image-2.0-pro support for baoyu-image-gen --- README.md | 4 +- README.zh.md | 4 +- skills/baoyu-image-gen/SKILL.md | 56 ++- .../references/config/first-time-setup.md | 22 +- .../references/config/preferences-schema.md | 4 +- skills/baoyu-image-gen/scripts/main.ts | 2 +- .../scripts/providers/dashscope.test.ts | 126 ++++- .../scripts/providers/dashscope.ts | 462 ++++++++++++++---- 8 files changed, 583 insertions(+), 97 deletions(-) diff --git a/README.md b/README.md index c5c6668..cb7669e 100644 --- a/README.md +++ b/README.md @@ -726,7 +726,7 @@ AI SDK-based image generation using OpenAI, Google, OpenRouter, DashScope (Aliyu | `OPENAI_IMAGE_MODEL` | OpenAI model | `gpt-image-1.5` | | `OPENROUTER_IMAGE_MODEL` | OpenRouter model | `google/gemini-3.1-flash-image-preview` | | `GOOGLE_IMAGE_MODEL` | Google model | `gemini-3-pro-image-preview` | -| `DASHSCOPE_IMAGE_MODEL` | DashScope model | `z-image-turbo` | +| `DASHSCOPE_IMAGE_MODEL` | DashScope model | `qwen-image-2.0-pro` | | `REPLICATE_IMAGE_MODEL` | Replicate model | `google/nano-banana-pro` | | `JIMENG_IMAGE_MODEL` | Jimeng model | `jimeng_t2i_v40` | | `SEEDREAM_IMAGE_MODEL` | Seedream model | `doubao-seedream-5-0-260128` | @@ -996,7 +996,7 @@ GOOGLE_IMAGE_MODEL=gemini-3-pro-image-preview # DashScope (Aliyun Tongyi Wanxiang) DASHSCOPE_API_KEY=sk-xxx -DASHSCOPE_IMAGE_MODEL=z-image-turbo +DASHSCOPE_IMAGE_MODEL=qwen-image-2.0-pro # DASHSCOPE_BASE_URL=https://dashscope.aliyuncs.com/api/v1 # Replicate diff --git a/README.zh.md b/README.zh.md index 8a15cd4..2f70c64 100644 --- a/README.zh.md +++ b/README.zh.md @@ -726,7 +726,7 @@ AI 驱动的生成后端。 | `OPENAI_IMAGE_MODEL` | OpenAI 模型 | `gpt-image-1.5` | | `OPENROUTER_IMAGE_MODEL` | OpenRouter 模型 | `google/gemini-3.1-flash-image-preview` | | `GOOGLE_IMAGE_MODEL` | Google 模型 | `gemini-3-pro-image-preview` | -| `DASHSCOPE_IMAGE_MODEL` | DashScope 模型 | `z-image-turbo` | +| `DASHSCOPE_IMAGE_MODEL` | DashScope 模型 | `qwen-image-2.0-pro` | | `REPLICATE_IMAGE_MODEL` | Replicate 模型 | `google/nano-banana-pro` | | `JIMENG_IMAGE_MODEL` | 即梦模型 | `jimeng_t2i_v40` | | `SEEDREAM_IMAGE_MODEL` | 豆包模型 | `doubao-seedream-5-0-260128` | @@ -996,7 +996,7 @@ GOOGLE_IMAGE_MODEL=gemini-3-pro-image-preview # DashScope（阿里通义万相） DASHSCOPE_API_KEY=sk-xxx -DASHSCOPE_IMAGE_MODEL=z-image-turbo +DASHSCOPE_IMAGE_MODEL=qwen-image-2.0-pro # DASHSCOPE_BASE_URL=https://dashscope.aliyuncs.com/api/v1 # Replicate diff --git a/skills/baoyu-image-gen/SKILL.md b/skills/baoyu-image-gen/SKILL.md index 1bc7d09..53d04e9 100644 --- a/skills/baoyu-image-gen/SKILL.md +++ b/skills/baoyu-image-gen/SKILL.md @@ -92,6 +92,12 @@ ${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider o # DashScope (阿里通义万象) ${BUN_X} {baseDir}/scripts/main.ts --prompt "一只可爱的猫" --image out.png --provider dashscope +# DashScope Qwen-Image 2.0 Pro (recommended for custom sizes and text rendering) +${BUN_X} {baseDir}/scripts/main.ts --prompt "为咖啡品牌设计一张 21:9 横幅海报，包含清晰中文标题" --image out.png --provider dashscope --model qwen-image-2.0-pro --size 2048x872 + +# DashScope legacy Qwen fixed-size model +${BUN_X} {baseDir}/scripts/main.ts --prompt "一张电影感海报" --image out.png --provider dashscope --model qwen-image-max --size 1664x928 + # Replicate (google/nano-banana-pro) ${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider replicate @@ -142,7 +148,7 @@ Paths in `promptFiles`, `image`, and `ref` are resolved relative to the batch fi | `--batchfile ` | JSON batch file for multi-image generation | | `--jobs ` | Worker count for batch mode (default: auto, max from config, built-in default 10) | | `--provider google\|openai\|openrouter\|dashscope\|jimeng\|seedream\|replicate` | Force provider (default: auto-detect) | -| `--model `, `-m` | Model ID (Google: `gemini-3-pro-image-preview`; OpenAI: `gpt-image-1.5`; OpenRouter: `google/gemini-3.1-flash-image-preview`) | +| `--model `, `-m` | Model ID (Google: `gemini-3-pro-image-preview`; OpenAI: `gpt-image-1.5`; OpenRouter: `google/gemini-3.1-flash-image-preview`; DashScope: `qwen-image-2.0-pro`) | | `--ar ` | Aspect ratio (e.g., `16:9`, `1:1`, `4:3`) | | `--size ` | Size (e.g., `1024x1024`) | | `--quality normal\|2k` | Quality preset (default: `2k`) | @@ -166,7 +172,7 @@ Paths in `promptFiles`, `image`, and `ref` are resolved relative to the batch fi | `OPENAI_IMAGE_MODEL` | OpenAI model override | | `OPENROUTER_IMAGE_MODEL` | OpenRouter model override (default: `google/gemini-3.1-flash-image-preview`) | | `GOOGLE_IMAGE_MODEL` | Google model override | -| `DASHSCOPE_IMAGE_MODEL` | DashScope model override (default: z-image-turbo) | +| `DASHSCOPE_IMAGE_MODEL` | DashScope model override (default: `qwen-image-2.0-pro`) | | `REPLICATE_IMAGE_MODEL` | Replicate model override (default: google/nano-banana-pro) | | `JIMENG_IMAGE_MODEL` | Jimeng model override (default: jimeng_t2i_v40) | | `SEEDREAM_IMAGE_MODEL` | Seedream model override (default: doubao-seedream-5-0-260128) | @@ -201,6 +207,52 @@ Model priority (highest → lowest), applies to all providers: - Show: `Using [provider] / [model]` - Show switch hint: `Switch model: --model | EXTEND.md default_model.[provider] | env _IMAGE_MODEL` +### DashScope Models + +Use `--model qwen-image-2.0-pro` or set `default_model.dashscope` / `DASHSCOPE_IMAGE_MODEL` when the user wants official Qwen-Image behavior. + +Official DashScope model families: + +- `qwen-image-2.0-pro`, `qwen-image-2.0-pro-2026-03-03`, `qwen-image-2.0`, `qwen-image-2.0-2026-03-03` + - Free-form `size` in `宽*高` format + - Total pixels must stay between `512*512` and `2048*2048` + - Default size is approximately `1024*1024` + - Best choice for custom ratios such as `21:9` and text-heavy Chinese/English layouts +- `qwen-image-max`, `qwen-image-max-2025-12-30`, `qwen-image-plus`, `qwen-image-plus-2026-01-09`, `qwen-image` + - Fixed sizes only: `1664*928`, `1472*1104`, `1328*1328`, `1104*1472`, `928*1664` + - Default size is `1664*928` + - `qwen-image` currently has the same capability as `qwen-image-plus` +- Legacy DashScope models such as `z-image-turbo`, `z-image-ultra`, `wanx-v1` + - Keep using them only when the user explicitly asks for legacy behavior or compatibility + +When translating CLI args into DashScope behavior: + +- `--size` wins over `--ar` +- For `qwen-image-2.0*`, prefer explicit `--size`; otherwise infer from `--ar` and use the official recommended resolutions below +- For `qwen-image-max/plus/image`, only use the five official fixed sizes; if the requested ratio is not covered, switch to `qwen-image-2.0-pro` +- `--quality` is a baoyu-image-gen compatibility preset, not a native DashScope API field. Mapping `normal` / `2k` onto the `qwen-image-2.0*` table below is an implementation inference, not an official API guarantee + +Recommended `qwen-image-2.0*` sizes for common aspect ratios: + +| Ratio | `normal` | `2k` | +|-------|----------|------| +| `1:1` | `1024*1024` | `1536*1536` | +| `2:3` | `768*1152` | `1024*1536` | +| `3:2` | `1152*768` | `1536*1024` | +| `3:4` | `960*1280` | `1080*1440` | +| `4:3` | `1280*960` | `1440*1080` | +| `9:16` | `720*1280` | `1080*1920` | +| `16:9` | `1280*720` | `1920*1080` | +| `21:9` | `1344*576` | `2048*872` | + +DashScope official APIs also expose `negative_prompt`, `prompt_extend`, and `watermark`, but `baoyu-image-gen` does not expose them as dedicated CLI flags today. + +Official references: + +- [Qwen-Image API](https://help.aliyun.com/zh/model-studio/qwen-image-api) +- [Text-to-image guide](https://help.aliyun.com/zh/model-studio/text-to-image) +- [Qwen-Image Edit API](https://help.aliyun.com/zh/model-studio/qwen-image-edit-api) + ### OpenRouter Models Use full OpenRouter model IDs, e.g.: diff --git a/skills/baoyu-image-gen/references/config/first-time-setup.md b/skills/baoyu-image-gen/references/config/first-time-setup.md index ef11cba..78489fe 100644 --- a/skills/baoyu-image-gen/references/config/first-time-setup.md +++ b/skills/baoyu-image-gen/references/config/first-time-setup.md @@ -50,7 +50,7 @@ options: - label: "OpenRouter" description: "Router for Gemini/FLUX/OpenAI-compatible image models" - label: "DashScope" - description: "Alibaba Cloud - z-image-turbo, good for Chinese content" + description: "Alibaba Cloud - Qwen-Image, strong Chinese/English text rendering" - label: "Replicate" description: "Community models - nano-banana-pro, flexible model selection" ``` @@ -186,12 +186,26 @@ options: header: "DashScope Model" question: "Choose a default DashScope image generation model?" options: - - label: "z-image-turbo (Recommended)" - description: "Fast generation, good quality" + - label: "qwen-image-2.0-pro (Recommended)" + description: "Best DashScope model for text rendering and custom sizes" + - label: "qwen-image-2.0" + description: "Faster 2.0 variant with flexible output size" + - label: "qwen-image-max" + description: "Legacy Qwen model with five fixed output sizes" + - label: "qwen-image-plus" + description: "Legacy Qwen model, same current capability as qwen-image" + - label: "z-image-turbo" + description: "Legacy DashScope model for compatibility" - label: "z-image-ultra" - description: "Higher quality, slower generation" + description: "Legacy DashScope model, higher quality but slower" ``` +Notes for DashScope setup: + +- Prefer `qwen-image-2.0-pro` when the user needs custom `--size`, uncommon ratios like `21:9`, or strong Chinese/English text rendering. +- `qwen-image-max` / `qwen-image-plus` / `qwen-image` only support five fixed sizes: `1664*928`, `1472*1104`, `1328*1328`, `1104*1472`, `928*1664`. +- In `baoyu-image-gen`, `quality` is a compatibility preset. It is not a native DashScope parameter. + ### Replicate Model Selection ```yaml diff --git a/skills/baoyu-image-gen/references/config/preferences-schema.md b/skills/baoyu-image-gen/references/config/preferences-schema.md index ec252a1..ad2c045 100644 --- a/skills/baoyu-image-gen/references/config/preferences-schema.md +++ b/skills/baoyu-image-gen/references/config/preferences-schema.md @@ -23,7 +23,7 @@ default_model: google: null # e.g., "gemini-3-pro-image-preview", "gemini-3.1-flash-image-preview" openai: null # e.g., "gpt-image-1.5", "gpt-image-1" openrouter: null # e.g., "google/gemini-3.1-flash-image-preview" - dashscope: null # e.g., "z-image-turbo" + dashscope: null # e.g., "qwen-image-2.0-pro" replicate: null # e.g., "google/nano-banana-pro" batch: @@ -88,7 +88,7 @@ default_model: google: "gemini-3-pro-image-preview" openai: "gpt-image-1.5" openrouter: "google/gemini-3.1-flash-image-preview" - dashscope: "z-image-turbo" + dashscope: "qwen-image-2.0-pro" replicate: "google/nano-banana-pro" batch: max_workers: 10 diff --git a/skills/baoyu-image-gen/scripts/main.ts b/skills/baoyu-image-gen/scripts/main.ts index f64b41c..d31e18a 100644 --- a/skills/baoyu-image-gen/scripts/main.ts +++ b/skills/baoyu-image-gen/scripts/main.ts @@ -116,7 +116,7 @@ Environment variables: OPENAI_IMAGE_MODEL Default OpenAI model (gpt-image-1.5) OPENROUTER_IMAGE_MODEL Default OpenRouter model (google/gemini-3.1-flash-image-preview) GOOGLE_IMAGE_MODEL Default Google model (gemini-3-pro-image-preview) - DASHSCOPE_IMAGE_MODEL Default DashScope model (z-image-turbo) + DASHSCOPE_IMAGE_MODEL Default DashScope model (qwen-image-2.0-pro) REPLICATE_IMAGE_MODEL Default Replicate model (google/nano-banana-pro) JIMENG_IMAGE_MODEL Default Jimeng model (jimeng_t2i_v40) SEEDREAM_IMAGE_MODEL Default Seedream model (doubao-seedream-5-0-260128) diff --git a/skills/baoyu-image-gen/scripts/providers/dashscope.test.ts b/skills/baoyu-image-gen/scripts/providers/dashscope.test.ts index f7c84b7..ab6100a 100644 --- a/skills/baoyu-image-gen/scripts/providers/dashscope.test.ts +++ b/skills/baoyu-image-gen/scripts/providers/dashscope.test.ts @@ -1,25 +1,147 @@ import assert from "node:assert/strict"; -import test from "node:test"; +import test, { type TestContext } from "node:test"; import { + getDefaultModel, + getModelFamily, + getQwen2SizeFromAspectRatio, getSizeFromAspectRatio, normalizeSize, parseAspectRatio, + parseSize, + resolveSizeForModel, } from "./dashscope.ts"; +function useEnv( + t: TestContext, + values: Record, +): void { + const previous = new Map(); + for (const [key, value] of Object.entries(values)) { + previous.set(key, process.env[key]); + if (value == null) { + delete process.env[key]; + } else { + process.env[key] = value; + } + } + + t.after(() => { + for (const [key, value] of previous.entries()) { + if (value == null) { + delete process.env[key]; + } else { + process.env[key] = value; + } + } + }); +} + +test("DashScope default model prefers env override and otherwise uses qwen-image-2.0-pro", (t) => { + useEnv(t, { DASHSCOPE_IMAGE_MODEL: null }); + assert.equal(getDefaultModel(), "qwen-image-2.0-pro"); + + process.env.DASHSCOPE_IMAGE_MODEL = "qwen-image-max"; + assert.equal(getDefaultModel(), "qwen-image-max"); +}); + test("DashScope aspect-ratio parsing accepts numeric ratios only", () => { assert.deepEqual(parseAspectRatio("3:2"), { width: 3, height: 2 }); assert.equal(parseAspectRatio("square"), null); assert.equal(parseAspectRatio("-1:2"), null); }); -test("DashScope size selection picks the closest supported size per quality preset", () => { +test("DashScope model family routing distinguishes qwen-2.0, fixed-size qwen, and legacy models", () => { + assert.equal(getModelFamily("qwen-image-2.0-pro"), "qwen2"); + assert.equal(getModelFamily("qwen-image"), "qwenFixed"); + assert.equal(getModelFamily("z-image-turbo"), "legacy"); + assert.equal(getModelFamily("wanx-v1"), "legacy"); +}); + +test("Legacy DashScope size selection keeps the previous quality-based heuristic", () => { assert.equal(getSizeFromAspectRatio(null, "normal"), "1024*1024"); assert.equal(getSizeFromAspectRatio("16:9", "normal"), "1280*720"); assert.equal(getSizeFromAspectRatio("16:9", "2k"), "2048*1152"); assert.equal(getSizeFromAspectRatio("invalid", "2k"), "1536*1536"); }); +test("Qwen 2.0 recommended sizes follow the official common-ratio table", () => { + assert.equal(getQwen2SizeFromAspectRatio(null, "normal"), "1024*1024"); + assert.equal(getQwen2SizeFromAspectRatio(null, "2k"), "1536*1536"); + assert.equal(getQwen2SizeFromAspectRatio("16:9", "normal"), "1280*720"); + assert.equal(getQwen2SizeFromAspectRatio("21:9", "2k"), "2048*872"); +}); + +test("Qwen 2.0 derives free-form sizes within pixel budget for uncommon ratios", () => { + const size = getQwen2SizeFromAspectRatio("5:2", "normal"); + const parsed = parseSize(size); + assert.ok(parsed); + assert.ok(parsed.width * parsed.height >= 512 * 512); + assert.ok(parsed.width * parsed.height <= 2048 * 2048); + assert.ok(Math.abs(parsed.width / parsed.height - 2.5) < 0.08); +}); + +test("resolveSizeForModel validates explicit qwen-image-2.0 sizes by total pixels", () => { + assert.equal( + resolveSizeForModel("qwen-image-2.0-pro", { + size: "2048x872", + aspectRatio: null, + quality: "2k", + }), + "2048*872", + ); + + assert.throws( + () => + resolveSizeForModel("qwen-image-2.0-pro", { + size: "4096x4096", + aspectRatio: null, + quality: "2k", + }), + /total pixels between/, + ); +}); + +test("resolveSizeForModel enforces fixed sizes for qwen-image-max/plus/image", () => { + assert.equal( + resolveSizeForModel("qwen-image-max", { + size: null, + aspectRatio: "1:1", + quality: "2k", + }), + "1328*1328", + ); + + assert.equal( + resolveSizeForModel("qwen-image", { + size: "1664x928", + aspectRatio: "9:16", + quality: "normal", + }), + "1664*928", + ); + + assert.throws( + () => + resolveSizeForModel("qwen-image-max", { + size: null, + aspectRatio: "21:9", + quality: "2k", + }), + /supports only fixed ratios/, + ); + + assert.throws( + () => + resolveSizeForModel("qwen-image-plus", { + size: "1024x1024", + aspectRatio: null, + quality: "2k", + }), + /support only these sizes/, + ); +}); + test("DashScope size normalization converts WxH into provider format", () => { assert.equal(normalizeSize("1024x1024"), "1024*1024"); assert.equal(normalizeSize("2048*1152"), "2048*1152"); diff --git a/skills/baoyu-image-gen/scripts/providers/dashscope.ts b/skills/baoyu-image-gen/scripts/providers/dashscope.ts index f854e33..60228c1 100644 --- a/skills/baoyu-image-gen/scripts/providers/dashscope.ts +++ b/skills/baoyu-image-gen/scripts/providers/dashscope.ts @@ -1,28 +1,46 @@ -import type { CliArgs } from "../types"; +import type { CliArgs, Quality } from "../types"; -export function getDefaultModel(): string { - return process.env.DASHSCOPE_IMAGE_MODEL || "z-image-turbo"; -} +type DashScopeModelFamily = "qwen2" | "qwenFixed" | "legacy"; -function getApiKey(): string | null { - return process.env.DASHSCOPE_API_KEY || null; -} +type DashScopeModelSpec = { + family: DashScopeModelFamily; + defaultSize: string; +}; -function getBaseUrl(): string { - const base = process.env.DASHSCOPE_BASE_URL || "https://dashscope.aliyuncs.com"; - return base.replace(/\/+$/g, ""); -} +const DEFAULT_MODEL = "qwen-image-2.0-pro"; +const MIN_QWEN_2_TOTAL_PIXELS = 512 * 512; +const MAX_QWEN_2_TOTAL_PIXELS = 2048 * 2048; +const SIZE_STEP = 16; +const QWEN_NEGATIVE_PROMPT = + "低分辨率，低画质，肢体畸形，手指畸形，画面过饱和，蜡像感，人脸无细节，过度光滑，画面具有AI感，构图混乱，文字模糊，扭曲"; -export function parseAspectRatio(ar: string): { width: number; height: number } | null { - const match = ar.match(/^(\d+(?:\.\d+)?):(\d+(?:\.\d+)?)$/); - if (!match) return null; - const w = parseFloat(match[1]!); - const h = parseFloat(match[2]!); - if (w <= 0 || h <= 0) return null; - return { width: w, height: h }; -} +const QWEN_2_TARGET_PIXELS: Record = { + normal: 1024 * 1024, + "2k": 1536 * 1536, +}; -const STANDARD_SIZES: [number, number][] = [ +const QWEN_2_RECOMMENDED: Record> = { + "1:1": { normal: "1024*1024", "2k": "1536*1536" }, + "2:3": { normal: "768*1152", "2k": "1024*1536" }, + "3:2": { normal: "1152*768", "2k": "1536*1024" }, + "3:4": { normal: "960*1280", "2k": "1080*1440" }, + "4:3": { normal: "1280*960", "2k": "1440*1080" }, + "9:16": { normal: "720*1280", "2k": "1080*1920" }, + "16:9": { normal: "1280*720", "2k": "1920*1080" }, + "21:9": { normal: "1344*576", "2k": "2048*872" }, +}; + +const QWEN_FIXED_SIZES_BY_RATIO: Record = { + "16:9": "1664*928", + "4:3": "1472*1104", + "1:1": "1328*1328", + "3:4": "1104*1472", + "9:16": "928*1664", +}; + +const QWEN_FIXED_SIZES = Object.values(QWEN_FIXED_SIZES_BY_RATIO); + +const LEGACY_STANDARD_SIZES: [number, number][] = [ [1024, 1024], [1280, 720], [720, 1280], @@ -34,7 +52,7 @@ const STANDARD_SIZES: [number, number][] = [ [864, 1536], ]; -const STANDARD_SIZES_2K: [number, number][] = [ +const LEGACY_STANDARD_SIZES_2K: [number, number][] = [ [1536, 1536], [2048, 1152], [1152, 2048], @@ -45,9 +63,167 @@ const STANDARD_SIZES_2K: [number, number][] = [ [2048, 2048], ]; +const QWEN_2_SPEC: DashScopeModelSpec = { + family: "qwen2", + defaultSize: "1024*1024", +}; + +const QWEN_FIXED_SPEC: DashScopeModelSpec = { + family: "qwenFixed", + defaultSize: QWEN_FIXED_SIZES_BY_RATIO["16:9"], +}; + +const LEGACY_SPEC: DashScopeModelSpec = { + family: "legacy", + defaultSize: "1536*1536", +}; + +const MODEL_SPEC_ALIASES: Record = { + "qwen-image-2.0-pro": QWEN_2_SPEC, + "qwen-image-2.0-pro-2026-03-03": QWEN_2_SPEC, + "qwen-image-2.0": QWEN_2_SPEC, + "qwen-image-2.0-2026-03-03": QWEN_2_SPEC, + "qwen-image-max": QWEN_FIXED_SPEC, + "qwen-image-max-2025-12-30": QWEN_FIXED_SPEC, + "qwen-image-plus": QWEN_FIXED_SPEC, + "qwen-image-plus-2026-01-09": QWEN_FIXED_SPEC, + "qwen-image": QWEN_FIXED_SPEC, +}; + +export function getDefaultModel(): string { + return process.env.DASHSCOPE_IMAGE_MODEL || DEFAULT_MODEL; +} + +function getApiKey(): string | null { + return process.env.DASHSCOPE_API_KEY || null; +} + +function getBaseUrl(): string { + const base = process.env.DASHSCOPE_BASE_URL || "https://dashscope.aliyuncs.com"; + return base.replace(/\/+$/g, ""); +} + +function getModelSpec(model: string): DashScopeModelSpec { + return MODEL_SPEC_ALIASES[model.trim().toLowerCase()] || LEGACY_SPEC; +} + +export function getModelFamily(model: string): DashScopeModelFamily { + return getModelSpec(model).family; +} + +function normalizeQuality(quality: CliArgs["quality"]): Quality { + return quality === "normal" ? "normal" : "2k"; +} + +export function parseAspectRatio(ar: string): { width: number; height: number } | null { + const match = ar.match(/^(\d+(?:\.\d+)?):(\d+(?:\.\d+)?)$/); + if (!match) return null; + const w = parseFloat(match[1]!); + const h = parseFloat(match[2]!); + if (w <= 0 || h <= 0) return null; + return { width: w, height: h }; +} + +export function normalizeSize(size: string): string { + return size.replace("x", "*"); +} + +export function parseSize(size: string): { width: number; height: number } | null { + const match = normalizeSize(size).match(/^(\d+)\*(\d+)$/); + if (!match) return null; + const width = Number(match[1]); + const height = Number(match[2]); + if (!Number.isFinite(width) || !Number.isFinite(height) || width <= 0 || height <= 0) { + return null; + } + return { width, height }; +} + +function formatSize(width: number, height: number): string { + return `${width}*${height}`; +} + +function getRatioValue(ar: string): number | null { + const parsed = parseAspectRatio(ar); + if (!parsed) return null; + return parsed.width / parsed.height; +} + +function findKnownRatioKey(ar: string, candidates: string[], tolerance = 0.02): string | null { + const targetRatio = getRatioValue(ar); + if (targetRatio == null) return null; + + let bestKey: string | null = null; + let bestDiff = Infinity; + + for (const candidate of candidates) { + const candidateRatio = getRatioValue(candidate); + if (candidateRatio == null) continue; + const diff = Math.abs(candidateRatio - targetRatio); + if (diff < bestDiff) { + bestDiff = diff; + bestKey = candidate; + } + } + + return bestDiff <= tolerance ? bestKey : null; +} + +function roundToStep(value: number): number { + return Math.max(SIZE_STEP, Math.round(value / SIZE_STEP) * SIZE_STEP); +} + +function fitToPixelBudget( + width: number, + height: number, + minPixels: number, + maxPixels: number, +): { width: number; height: number } { + let nextWidth = width; + let nextHeight = height; + let pixels = nextWidth * nextHeight; + + if (pixels > maxPixels) { + const scale = Math.sqrt(maxPixels / pixels); + nextWidth *= scale; + nextHeight *= scale; + } else if (pixels < minPixels) { + const scale = Math.sqrt(minPixels / pixels); + nextWidth *= scale; + nextHeight *= scale; + } + + let roundedWidth = roundToStep(nextWidth); + let roundedHeight = roundToStep(nextHeight); + pixels = roundedWidth * roundedHeight; + + while (pixels > maxPixels && (roundedWidth > SIZE_STEP || roundedHeight > SIZE_STEP)) { + if (roundedWidth >= roundedHeight && roundedWidth > SIZE_STEP) { + roundedWidth -= SIZE_STEP; + } else if (roundedHeight > SIZE_STEP) { + roundedHeight -= SIZE_STEP; + } else { + break; + } + pixels = roundedWidth * roundedHeight; + } + + while (pixels < minPixels) { + if (roundedWidth <= roundedHeight) { + roundedWidth += SIZE_STEP; + } else { + roundedHeight += SIZE_STEP; + } + pixels = roundedWidth * roundedHeight; + } + + return { width: roundedWidth, height: roundedHeight }; +} + export function getSizeFromAspectRatio(ar: string | null, quality: CliArgs["quality"]): string { - const is2k = quality === "2k"; - const defaultSize = is2k ? "1536*1536" : "1024*1024"; + const normalizedQuality = normalizeQuality(quality); + const sizes = normalizedQuality === "2k" ? LEGACY_STANDARD_SIZES_2K : LEGACY_STANDARD_SIZES; + const defaultSize = normalizedQuality === "2k" ? "1536*1536" : "1024*1024"; if (!ar) return defaultSize; @@ -55,86 +231,157 @@ export function getSizeFromAspectRatio(ar: string | null, quality: CliArgs["qual if (!parsed) return defaultSize; const targetRatio = parsed.width / parsed.height; - const sizes = is2k ? STANDARD_SIZES_2K : STANDARD_SIZES; - let best = defaultSize; let bestDiff = Infinity; - for (const [w, h] of sizes) { - const diff = Math.abs(w / h - targetRatio); + for (const [width, height] of sizes) { + const diff = Math.abs(width / height - targetRatio); if (diff < bestDiff) { bestDiff = diff; - best = `${w}*${h}`; + best = formatSize(width, height); } } return best; } -export function normalizeSize(size: string): string { - return size.replace("x", "*"); +export function getQwen2SizeFromAspectRatio(ar: string | null, quality: CliArgs["quality"]): string { + const normalizedQuality = normalizeQuality(quality); + + if (!ar) { + return QWEN_2_RECOMMENDED["1:1"][normalizedQuality]; + } + + const recommendedRatio = findKnownRatioKey(ar, Object.keys(QWEN_2_RECOMMENDED)); + if (recommendedRatio) { + return QWEN_2_RECOMMENDED[recommendedRatio][normalizedQuality]; + } + + const parsed = parseAspectRatio(ar); + if (!parsed) { + return QWEN_2_RECOMMENDED["1:1"][normalizedQuality]; + } + + const targetRatio = parsed.width / parsed.height; + const targetPixels = QWEN_2_TARGET_PIXELS[normalizedQuality]; + const rawWidth = Math.sqrt(targetPixels * targetRatio); + const rawHeight = Math.sqrt(targetPixels / targetRatio); + const fitted = fitToPixelBudget( + rawWidth, + rawHeight, + MIN_QWEN_2_TOTAL_PIXELS, + MAX_QWEN_2_TOTAL_PIXELS, + ); + + return formatSize(fitted.width, fitted.height); } -export async function generateImage( - prompt: string, - model: string, - args: CliArgs -): Promise { - const apiKey = getApiKey(); - if (!apiKey) throw new Error("DASHSCOPE_API_KEY is required"); - - if (args.referenceImages.length > 0) { - throw new Error( - "Reference images are not supported with DashScope provider in baoyu-image-gen. Use --provider google with a Gemini multimodal model." +function getQwenFixedSizeFromAspectRatio(ar: string | null, quality: CliArgs["quality"]): string { + if (quality === "normal") { + console.warn( + "DashScope qwen-image-max/plus/image models use fixed output sizes; --quality normal does not change the generated resolution." ); } - const size = args.size ? normalizeSize(args.size) : getSizeFromAspectRatio(args.aspectRatio, args.quality); - const url = `${getBaseUrl()}/api/v1/services/aigc/multimodal-generation/generation`; + if (!ar) return QWEN_FIXED_SPEC.defaultSize; - const body = { - model, - input: { - messages: [ - { - role: "user", - content: [{ text: prompt }], - }, - ], - }, - parameters: { - prompt_extend: false, - size, - }, - }; - - console.log(`Generating image with DashScope (${model})...`, { size }); - - const res = await fetch(url, { - method: "POST", - headers: { - "Content-Type": "application/json", - Authorization: `Bearer ${apiKey}`, - }, - body: JSON.stringify(body), - }); - - if (!res.ok) { - const err = await res.text(); - throw new Error(`DashScope API error (${res.status}): ${err}`); + const ratioKey = findKnownRatioKey(ar, Object.keys(QWEN_FIXED_SIZES_BY_RATIO)); + if (!ratioKey) { + throw new Error( + `DashScope model supports only fixed ratios ${Object.keys(QWEN_FIXED_SIZES_BY_RATIO).join(", ")}. ` + + `For custom ratios like "${ar}", use --model qwen-image-2.0-pro.` + ); } - const result = await res.json() as { - output?: { - result_image?: string; - choices?: Array<{ - message?: { - content?: Array<{ image?: string }>; - }; - }>; - }; + return QWEN_FIXED_SIZES_BY_RATIO[ratioKey]!; +} + +function validateSizeFormat(size: string): { width: number; height: number } { + const parsed = parseSize(size); + if (!parsed) { + throw new Error(`Invalid DashScope size "${size}". Expected x or *.`); + } + return parsed; +} + +function validateQwen2Size(size: string): string { + const normalized = normalizeSize(size); + const parsed = validateSizeFormat(normalized); + const totalPixels = parsed.width * parsed.height; + if (totalPixels < MIN_QWEN_2_TOTAL_PIXELS || totalPixels > MAX_QWEN_2_TOTAL_PIXELS) { + throw new Error( + `DashScope qwen-image-2.0* models require total pixels between ${MIN_QWEN_2_TOTAL_PIXELS} ` + + `and ${MAX_QWEN_2_TOTAL_PIXELS}. Received ${normalized} (${totalPixels} pixels).` + ); + } + return normalized; +} + +function validateQwenFixedSize(size: string): string { + const normalized = normalizeSize(size); + validateSizeFormat(normalized); + if (!QWEN_FIXED_SIZES.includes(normalized)) { + throw new Error( + `DashScope qwen-image-max/plus/image models support only these sizes: ${QWEN_FIXED_SIZES.join(", ")}. ` + + `Received ${normalized}.` + ); + } + return normalized; +} + +export function resolveSizeForModel( + model: string, + args: Pick, +): string { + const spec = getModelSpec(model); + + if (args.size) { + if (spec.family === "qwen2") return validateQwen2Size(args.size); + if (spec.family === "qwenFixed") return validateQwenFixedSize(args.size); + validateSizeFormat(args.size); + return normalizeSize(args.size); + } + + if (spec.family === "qwen2") { + return getQwen2SizeFromAspectRatio(args.aspectRatio, args.quality); + } + + if (spec.family === "qwenFixed") { + return getQwenFixedSizeFromAspectRatio(args.aspectRatio, args.quality); + } + + return getSizeFromAspectRatio(args.aspectRatio, args.quality); +} + +function buildParameters( + family: DashScopeModelFamily, + size: string, +): Record { + const parameters: Record = { + prompt_extend: false, + size, }; + if (family === "qwen2" || family === "qwenFixed") { + parameters.watermark = false; + parameters.negative_prompt = QWEN_NEGATIVE_PROMPT; + } + + return parameters; +} + +type DashScopeResponse = { + output?: { + result_image?: string; + choices?: Array<{ + message?: { + content?: Array<{ image?: string }>; + }; + }>; + }; +}; + +async function extractImageFromResponse(result: DashScopeResponse): Promise { let imageData: string | null = null; if (result.output?.result_image) { @@ -163,3 +410,54 @@ export async function generateImage( return Uint8Array.from(Buffer.from(imageData, "base64")); } + +export async function generateImage( + prompt: string, + model: string, + args: CliArgs +): Promise { + const apiKey = getApiKey(); + if (!apiKey) throw new Error("DASHSCOPE_API_KEY is required"); + + if (args.referenceImages.length > 0) { + throw new Error( + "Reference images are not supported with DashScope provider in baoyu-image-gen. Use --provider google with a Gemini multimodal model." + ); + } + + const spec = getModelSpec(model); + const size = resolveSizeForModel(model, args); + const url = `${getBaseUrl()}/api/v1/services/aigc/multimodal-generation/generation`; + + const body = { + model, + input: { + messages: [ + { + role: "user", + content: [{ text: prompt }], + }, + ], + }, + parameters: buildParameters(spec.family, size), + }; + + console.log(`Generating image with DashScope (${model})...`, { family: spec.family, size }); + + const res = await fetch(url, { + method: "POST", + headers: { + "Content-Type": "application/json", + Authorization: `Bearer ${apiKey}`, + }, + body: JSON.stringify(body), + }); + + if (!res.ok) { + const err = await res.text(); + throw new Error(`DashScope API error (${res.status}): ${err}`); + } + + const result = await res.json() as DashScopeResponse; + return extractImageFromResponse(result); +}