Seedance 2.5

Text-to-video & image-to-video qua Dreamina Seedance 2.5 (ByteDance) — ban MOI NHAT, chat luong cao nhat (Dola: "For best video quality"). Truyen prompt (t2v) hoac them anh tham chieu (i2v, toi da 10 anh). Duration 5/10s, 6 ti le khung. Async: 5s ~1-3 phut, 10s ~5 phut.

Seedance 2.0

Text-to-video & image-to-video qua Dreamina Seedance 2.0 (ByteDance) — browser automation. Truyen prompt (t2v) hoac them anh tham chieu (i2v). Duration 5/10s, nhieu ti le khung. Async 1-3 phut.

Veo 3.1

Veo 3.1 (Google Vertex, veo-3.1-generate-001 GA) - video generation chat luong cao nhat cua Google. Text-to-video + image-to-video (anh <=20MB, first+last frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, generate audio. Async (~60-90s).

Veo 3.1 Fast

Veo 3.1 Fast (Google Vertex, veo-3.1-fast-generate-001 GA) - video generation nhanh & re. Text-to-video + image-to-video (anh <=20MB, first+last frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, generate audio. Async (~40-70s).

Gemini Omni Flash

Tạo video qua Google Vertex (Gemini Omni Flash) — text→video kèm audio + thinking. Hỗ trợ ảnh→video (Start Image), tỉ lệ 16:9 / 9:16, tối đa 10 giây.

Dreamina-Seedance-2.0-mini

Model tạo VIDEO đa phương thức (Multimodality-to-video) của ByteDance/Doubao: nhận text + ảnh + video + audio tham chiếu → video (có thể kèm audio). 480P/720P, 4-15s, 24fps.

Dreamina-Seedance-2.0-fast

Seedance 2.0 Fast của ByteDance/Doubao — tạo VIDEO đa phương thức nhanh hơn: text + ảnh (1-9) + video (0-3) + audio (0-3) tham chiếu → 1 video (có thể kèm audio). 480P/720P, 4-15s, 24fps; các chế độ Text→Video, Ảnh→Video, Frame đầu/cuối, Tham chiếu đa phương thức.

Dreamina-Seedance-2.0

Seedance 2.0 (bản full) của ByteDance/Doubao — tạo VIDEO đa phương thức: text + ảnh (1-9) + video (0-3) + audio (0-3) tham chiếu → 1 video (có thể kèm audio). Hỗ trợ 480P/720P/1080P/4K, 4-15s, 24fps; các chế độ Text→Video, Ảnh→Video, Frame đầu/cuối, Tham chiếu đa phương thức.

Grok Video 1.5

Tạo video qua grok.com/imagine (browser, cần SuperGrok). Text-to-video (chỉ prompt) hoặc image-to-video (kèm 1 ảnh). Aspect Ratio áp dụng cho text-to-video; có ảnh input (i2v) thì giữ tỉ lệ ảnh gốc.

Veo 3.1 Lite

Veo 3.1 Lite (Google Vertex, veo-3.1-lite-generate-001 - Preview) - video generation re nhat dong Veo. Text-to-video + image-to-video (first frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, sound generation. Async (~40-70s).

HappyHorse 1.1 (Text→Video)

HappyHorse 1.1 text-to-video của Alibaba (Tongyi) — tạo video từ PROMPT. 720P/1080P, tỉ lệ 16:9/9:16/1:1/4:3/3:4, tối đa ~10s. DashScope async.

HappyHorse 1.1 (Image→Video)

HappyHorse 1.1 image-to-video của Alibaba (Tongyi) — tạo video từ ẢNH ĐẦU (first frame) + prompt. 720P/1080P, tối đa ~10s. DashScope async.

HappyHorse 1.0 (Video Editing)

HappyHorse 1.0 video-editing của Alibaba (Tongyi) — sửa VIDEO có sẵn theo instruction (prompt) + tối đa 5 ảnh tham chiếu. 720P/1080P. DashScope async (1-5 phút). ⚠️ Tính giá theo GIÂY video billable = độ dài NGUỒN + độ dài KẾT QUẢ (kết quả giữ nguyên độ dài video nguồn); 720P 42cr/giây · 1080P 72cr/giây.

HappyHorse 1.1 (Reference→Video)

HappyHorse 1.1 reference-to-video của Alibaba (Tongyi) — tạo video từ 1-4 ẢNH THAM CHIẾU + prompt (giữ nhân vật/phong cách). 720P/1080P, tối đa ~10s. DashScope async.

Midjourney Video

Midjourney Video (t2v) — tạo ảnh nền từ prompt rồi ANIMATE thành tối đa 4 video mp4 (h264 ~832×464, ~5s). Qua gateway browser (subscription Midjourney). Async ~1–3 phút. Hỗ trợ aspect_ratio + chọn ảnh nền (imageNum).

Wan2.7 Video Editing

Wan2.7 video-editing của Alibaba — sửa video có sẵn theo instruction (đổi phong cách, trang phục, cảnh...) + tối đa 4 ảnh tham chiếu. Output giữ độ dài video nguồn. ⚠️ Tính giá theo GIÂY billable (usage.duration = input+output); 720P 30cr/giây · 1080P 45cr/giây. DashScope async.

Wan2.7 Reference-to-Video

Wan2.7 reference-to-video của Alibaba — sinh video từ nhiều ảnh tham chiếu (1-4). Trong prompt gọi ảnh bằng "Image 1", "Image 2"... theo thứ tự. 720P/1080P, 2-15s. DashScope async (1-5 phút).

Wan2.7 Image-to-Video

Wan2.7 image-to-video của Alibaba — sinh video từ ảnh đầu (first frame); tuỳ chọn ảnh cuối (first+last frame) + audio lip-sync (nhép miệng theo giọng). 720P/1080P, 2-15s. DashScope async (1-5 phút).

Wan2.7 Text-to-Video

Wan2.7 text-to-video của Alibaba (Tongyi Wanxiang) — sinh video từ prompt, hỗ trợ tự sự đa cảnh (multi-shot). 720P/1080P, 2-15s. DashScope async (1-5 phút).

Gemini 3.6 Flash

Gemini 3.6 Flash (Google Vertex) - tri tue gan-Pro o gia/toc do Flash. Context 1 trieu token (1.048.576 vao / 65.535 ra). Input DA PHUONG THUC: text + anh (<=3000) + audio (~8.4h) + video (<=10, ~1h) + tai lieu pdf/txt. Thinking (suy luan sau - hien trace qua streaming khi bat thinking level), Grounding Google Search/Maps/Partners, function calling, structured output. TEXT-only output.

Claude Opus 5 is Anthropic's newest flagship reasoning model — a major step up in complex reasoning, agentic coding, and long-horizon autonomous task execution over the Opus 4.x line, with stronger instruction-following and professional knowledge work.

Grok 4.5

Claude Opus 4.8 delivers a major upgrade in complex reasoning, agentic coding, and professional knowledge work. With improved honesty, stronger long-horizon task execution, and a 1-million-token context window on supported platforms, it is built as Anthropic’s most capable generally available model for high-autonomy workflows and demanding enterprise use cases.

Claude Fable 5

Claude Opus 4.8 delivers a major upgrade in complex reasoning, agentic coding, and professional knowledge work. With improved honesty, stronger long-horizon task execution, and a 1-million-token context window on supported platforms, it is built as Anthropic’s most capable generally available model for high-autonomy workflows and demanding enterprise use cases.

GPT-5.6 Sol

GPT-5.6 Sol — flagship coding model mới nhất của OpenAI (frontier, GA 7/2026): agentic coding, nghiên cứu, task đa bước. Hỗ trợ vision (đọc ảnh) + reasoning effort low→ultra.

Claude Opus 4.8

Claude Opus 4.8 delivers a major upgrade in complex reasoning, agentic coding, and professional knowledge work. With improved honesty, stronger long-horizon task execution, and a 1-million-token context window on supported platforms, it is built as Anthropic’s most capable generally available model for high-autonomy workflows and demanding enterprise use cases.

Claude Sonnet 5

Claude Sonnet 5 — thế hệ Sonnet mới nhất của Anthropic: cân bằng tốc độ/chất lượng, coding + nội dung mạnh, hỗ trợ suy luận sâu (thinking) và web search.

GPT-5.6 Terra

GPT-5.6 Terra — biến thể cân bằng của series GPT-5.6. Vision + reasoning effort low→ultra.

GPT-5.6 Luna

GPT-5.6 Luna — biến thể nhẹ/nhanh của series GPT-5.6. Vision + reasoning effort low→ultra.

Gemini 3.5 Flash

Gemini 3.5 Flash (Google Vertex) - tri tue gan-Pro o gia/toc do Flash. Context 1 trieu token (1.048.576 vao / 65.535 ra). Input DA PHUONG THUC: text + anh (<=3000) + audio (~8.4h) + video (<=10, ~1h) + tai lieu pdf/txt. Thinking (suy luan sau - hien trace qua streaming khi bat thinking level), Grounding Google Search/Maps/Partners, function calling, structured output. TEXT-only output.

Gemini 3.1 Flash-Lite

Gemini 3.1 Flash-Lite (Google Vertex) - re va nhanh nhat nhom Gemini 3.x. Context 1 trieu token (1.048.576 vao / 65.535 ra). Input DA PHUONG THUC: text + anh (<=3000) + audio + video + tai lieu pdf/txt. Thinking (co hien trace), Grounding Google Search/Maps/Partners, function calling, structured output. TEXT-only output.

Gemini 3.1 Pro

Gemini 3.1 Pro (Google Vertex) - LLM reasoning/chat context 1 trieu token (1.048.576 vao / 65.536 ra). Input DA PHUONG THUC: text + anh + audio + video + tai lieu. Thinking + Grounding Google Search/Maps/Partners + function calling. TEXT-only output.

Grok Composer 2.5 Fast

Claude Opus 4.8 delivers a major upgrade in complex reasoning, agentic coding, and professional knowledge work. With improved honesty, stronger long-horizon task execution, and a 1-million-token context window on supported platforms, it is built as Anthropic’s most capable generally available model for high-autonomy workflows and demanding enterprise use cases.

Claude Opus 4.7

Claude Opus 4.7 delivers a quantum leap in logical reasoning and creative synthesis. With its next-gen architecture and massive 2-million-token context, it serves as the ultimate high-performance partner for those pushing the boundaries of innovation and discovery.

Claude Opus 4.6

Claude Sonnet 4.6 delivers strong reasoning, advanced coding capabilities, and native computer-use functionality. With adaptive thinking and a context window of up to 1 million tokens, it supports building reliable AI agents and large-scale automation workflows.

Claude Opus 4.5

Claude Opus 4.5 is Anthropic’s most intelligent model, combining maximum capability with practical performance across reasoning, coding, and complex problem solving.

Claude Sonnet 4.6

Claude Sonnet 4.6 delivers strong reasoning, advanced coding capabilities, and native computer-use functionality. With adaptive thinking and a context window of up to 1 million tokens, it supports building reliable AI agents and large-scale automation workflows.

Claude Sonnet 4.5

Developed by Anthropic, Claude Sonnet 4.5 is a premier large language model that excels in multi-step reasoning, complex coding, computer use, and specialized domains such as finance.

Claude Haiku-4-5

Claude Haiku 4.5 is Anthropic’s lightweight AI model designed for fast, cost-efficient AI applications. It delivers near-frontier performance for coding, chat, and automation tasks while providing significantly faster responses and lower cost than larger Claude models.

GPT-5-Codex

OpenAI Codex API delivers advanced coding capabilities built on GPT-5 Codex models. It supports code generation, debugging, and complex engineering workflows with scalable reasoning depth.

GPT-5.5

GPT-5.5 is OpenAI’s advanced reasoning model for agentic coding, knowledge work, scientific research, and complex multi-step task execution.

GPT-5.4

OpenAI's GPT-5.4 delivers powerful capabilities in complex reasoning, coding, and professional knowledge work. It enables faster development cycles, more reliable task execution, and higher quality results in applications.

DeepSeek V4 Pro

DeepSeek-V4 is a brand-new flagship hybrid reasoning model with breakthrough performance in complex tasks, advanced reasoning, code generation, and mathematical problem solving.

DeepSeek-V4 Flash

DeepSeek-V4-Flash is a high-efficiency hybrid reasoning model balancing speed and intelligence. Ideal for high-concurrency scenarios requiring fast responses.

DeepSeek-V3.2

DeepSeek-V3.2 Official Version: The latest iteration of V3 series. Enhanced performance on coding, math, and multilingual tasks with improved efficiency.

GLM-4.7

GLM-4.7 is Z.AI's flagship model in the GLM-4 series. Advanced capabilities in mathematical reasoning, code generation, and instruction following with native multimodal support.

Seed 2.0 Mini

Model hiểu đa phương thức (omni VLM) của ByteDance: nhận text + ảnh + video + audio, trả lời bằng văn bản. Bản mini nhẹ, nhanh, giá rẻ — hợp phân tích ảnh/audio/video nhanh.

Seed 2.0 Lite

Model hiểu đa phương thức (omni VLM) của ByteDance: nhận text + ảnh + video + audio, trả lời bằng văn bản. Bản lite mạnh hơn mini, cân bằng chất lượng/giá — phân tích đa phương thức sâu hơn.

Multimodal Embedding 001

Multimodal Embedding 001 (Google Vertex) - nhung TEXT va/hoac ANH vao CUNG khong gian vector 1408 chieu (tim anh bang text, so khop anh-text). Input text va/hoac anh.

Gemini Embedding 001

Gemini Embedding 001 (Google Vertex) - model embedding text MOI NHAT, vector 3072 chieu. Bill theo token (= gia Google). Input text; batch nhieu text.

Skylark-embedding-vision

Skylark Embedding Vision của ByteDance/Doubao — model EMBEDDING đa phương thức: vector hoá Text + Ảnh (multi-vector output), hỗ trợ song ngữ Trung/Anh. Dùng cho semantic search, RAG, so khớp ảnh↔text.

Xoa nen + depth map (Depth-Anything) — du lieu khac pha le/laser 3D Photo. Tra ve depth heightmap + anh cutout.

Chuyen anh 2D thanh model 3D mesh (.glb) qua WaveSpeed Hunyuan3D — cho in 3D, figurine, AR.

Midjourney Image-to-Image

Midjourney Image-to-Image (i2i) — dùng ≥1 ảnh tham chiếu (image_urls) định hướng phong cách/cấu trúc + prompt mô tả → 4 ảnh PNG. Mỗi URL = 1 image prompt (MJ tự re-host, khuyến nghị ≤5). Async ~30-60s. Nâng cao: --sref/--cref trong prompt.

Image-to-image với Seedream 5.0 Pro. Gửi 1–10 ảnh tham chiếu + prompt mô tả thay đổi mong muốn → trả về 1 ảnh đã biến đổi, không watermark. Khác bản text-to-image: t2i sinh 4 ảnh/lượt, i2i chỉ 1 ảnh.

GPT Image 2(img2img)

GPT Image 2 is OpenAI’s next-gen image model built for stronger photorealism, cleaner image editing, sharper text rendering, and more polished product photography. Designed for more advanced visual workflows, it pushes image generation beyond basic text-to-image output and into higher-quality creative, commercial, and design-ready use cases.

Qwen-Image-Edit Plus

Qwen-Image-Edit Plus của Alibaba — sửa ảnh theo câu lệnh, cân bằng chất lượng/chi phí (tier Plus, rẻ hơn Max). Nhận 1–3 ảnh tham chiếu + instruction.

Qwen-Image-Edit Max

Qwen-Image-Edit Max của Alibaba — sửa ảnh theo câu lệnh, chất lượng cao nhất dòng edit (tier Max). Nhận 1–3 ảnh tham chiếu + instruction.

Wan 2.6 Image (Edit)

Multi-image editing with wan2.6-image by Alibaba (DashScope): 1-4 reference images -> 1-4 edited images. Style transfer, subject consistency. Direct connection, no gateway.

Wan 2.7 Interactive Edit

Interactive editing with wan2.7-image-pro by Alibaba: place/blend objects using bounding boxes (bbox_list). Up to 2K. Direct connection, no gateway.

Wan 2.7 Image Edit

Image editing with wan2.7-image-pro by Alibaba (Wan 2.7 / DashScope). Attach 1-3 images + an instruction. Up to 2K. Direct connection, no gateway.

Qwen Image Edit

Sửa ảnh theo chỉ dẫn ngôn ngữ bằng Qwen-Image-Edit của Alibaba (DashScope). 1 ảnh (instruction-only) hoặc nhiều ảnh (reference). Kết nối thẳng, không qua gateway.

HappyHorse 1.1 (Image→Video)

HappyHorse 1.1 image-to-video của Alibaba (Tongyi) — tạo video từ ẢNH ĐẦU (first frame) + prompt. 720P/1080P, tối đa ~10s. DashScope async.

HappyHorse 1.1 (Reference→Video)

HappyHorse 1.1 reference-to-video của Alibaba (Tongyi) — tạo video từ 1-4 ẢNH THAM CHIẾU + prompt (giữ nhân vật/phong cách). 720P/1080P, tối đa ~10s. DashScope async.

Wan2.7 Reference-to-Video

Wan2.7 reference-to-video của Alibaba — sinh video từ nhiều ảnh tham chiếu (1-4). Trong prompt gọi ảnh bằng "Image 1", "Image 2"... theo thứ tự. 720P/1080P, 2-15s. DashScope async (1-5 phút).

Wan2.7 Image-to-Video

Wan2.7 image-to-video của Alibaba — sinh video từ ảnh đầu (first frame); tuỳ chọn ảnh cuối (first+last frame) + audio lip-sync (nhép miệng theo giọng). 720P/1080P, 2-15s. DashScope async (1-5 phút).

Fun-ASR (Speech to Text)

Fun-ASR (Alibaba DashScope) - chuyen giong noi thanh van ban, da ngon ngu (Trung/Anh/Viet...). Nhan audio URL cong khai (wav/mp3), tra transcript. Async qua transcription API.

Midjourney Image

Midjourney Image (t2i) — tạo 4 ảnh PNG chất lượng cao từ prompt qua gateway browser (subscription Midjourney Standard). Async ~30-60s. Hỗ trợ aspect_ratio, version, stylize, weird, chaos, raw, speed, negative.

Text-to-image & image-to-image với Seedream 5.0 Pro (ByteDance). t2i: chỉ cần prompt → 4 ảnh 2431×1365. i2i: thêm tối đa 10 ảnh tham chiếu → 1 ảnh. Ảnh KHÔNG watermark, host vĩnh viễn trên media.halcion.vn.

Nano Banana Pro

Nano Banana Pro (Gemini 3 Pro Image) qua Google Vertex — reasoning image: text rendering + world knowledge, tao/sua anh (≤14 anh ref) + Grounding Google Search + System instructions. Resolution 1K/2K/4K, 10 ti le.

Nano Banana 2

Google DeepMind’s Nano Banana 2 delivers sharper 2K imagery, intelligent 4K scaling, improved text rendering, and enhanced character consistency—offering a major leap in visual quality for creative and API-driven workflows.

Nano Banana 2 Lite

Nano Banana 2 Lite của Google (qua Vertex Gemini API) — bản nhẹ/nhanh nhất dòng Nano Banana (~4s/ảnh), tạo ảnh từ prompt + tối đa 14 ảnh tham chiếu (i2i). Resolution 1K (~1MP, Google chỉ hỗ trợ 1K), output png/jpeg/webp.

Grok Image

Tạo/sửa ảnh qua grok.com/imagine. Text-to-image (chỉ prompt → 4 ảnh) hoặc image-to-image (kèm 1–3 ảnh input → 1 ảnh sửa). Aspect Ratio áp dụng cho text-to-image; khi có ảnh input, output GIỮ tỉ lệ ảnh gốc (aspect bị bỏ qua). Cần gói SuperGrok.

Imagen 4 Fast

Imagen 4 Fast (imagen-4.0-fast-generate-001) - nhanh & re, tao 1-4 anh/lan. 5 ti le, resolution CHI 1K. Text to image.

Imagen 3

Imagen 3 (imagen-3.0-generate-002) - doi cu, ~1024px co dinh (khong resolution), tao 1-4 anh/lan, 5 ti le. Text to image.

Nano Banana

Nano Banana (Gemini 2.5 Flash Image) - model tao/sua anh goc cua Google DeepMind. Input toi da 3 anh tham chieu, 10 ti le, ~1024px co dinh. Manh sua/blend anh, re. Retire 2/10/2026.

GPT Image 2

GPT Image 2 is OpenAI’s next-gen image model built for stronger photorealism, cleaner image editing, sharper text rendering, and more polished product photography. Designed for more advanced visual workflows, it pushes image generation beyond basic text-to-image output and into higher-quality creative, commercial, and design-ready use cases.

Imagen 4

Imagen 4 (imagen-4.0-generate-001) - text to image chat luong cao cua Google. 5 ti le, resolution 1K/2K, tao 1-4 anh/lan. KHONG sua/tham chieu anh (text-only).

Imagen 4 Ultra

Imagen 4 Ultra (imagen-4.0-ultra-generate-001) - chat luong cao nhat, 1 anh/lan. 5 ti le, resolution 1K/2K. Text to image, KHONG sua anh.

Z-Image Turbo

Z-Image Turbo của Alibaba (Tongyi) — model text-to-image siêu nhẹ, tốc độ cao, chi phí thấp; render chữ Trung/Anh, độ phân giải linh hoạt 512–2048px. Cố định 1 ảnh/lần. prompt tối đa 800 ký tự.

ByteDance-Seedream-4.5

Seedream 4.5 delivers better editing consistency, improved multi-image fusion, finer detail control, natural small text and faces, and harmonious, aesthetic visuals.

The latest image generation model released by BytePlus. Reasoning and comprehension capabilities are further upgraded, allowing it to accurately interpret complex prompts and visual inputs.

ByteDance-Seedream-4.0

Seedream 4.0 is a SOTA multimodal image generation model, enabling text-to-image creation, image editing and multi-image generation within a single model.

Wan 2.7 Image Set

Coherent image-set generation with wan2.7-image-pro by Alibaba: one prompt -> N consistent images (enable_sequential). Up to 2K. Direct connection, no gateway.

Wan 2.7 Image

Text-to-image with wan2.7-image by Alibaba (Wan 2.7 / DashScope). 2K resolution. Direct connection, no gateway.

Wan 2.7 Image Pro

Text-to-image with wan2.7-image-pro by Alibaba (Wan 2.7 / DashScope). 4K resolution. Direct connection, no gateway.

Qwen Image Plus

Text-to-image with qwen-image-plus by Alibaba (Qwen/DashScope). Optional reference images. Direct connection, no gateway.

Qwen Image Max

Text-to-image with qwen-image-max by Alibaba (Qwen/DashScope). Optional reference images. Direct connection, no gateway.

Qwen Image 2.0 Pro

Text-to-image with qwen-image-2.0-pro by Alibaba (Qwen/DashScope). Optional reference images. Direct connection, no gateway.

Qwen Image 2.0

Text-to-image with qwen-image-2.0 by Alibaba (Qwen/DashScope). Optional reference images. Direct connection, no gateway.

Wan 2.6 Image (Interleaved)

Text-to-image with interleaved text+image output via wan2.6-image by Alibaba (DashScope, SSE streaming reassembled server-side). Optional 1 reference image. Direct connection, no gateway.

Lyria 2

Lyria 2 (Google Vertex) — tao NHAC instrumental tu prompt tieng Anh. Clip ~30s, WAV. Chi nhan prompt tieng Anh.

Lyria 3 Clip

Lyria 3 Clip (Google Vertex, Preview) — bien the clip NGAN ~30s cua Lyria 3. text->music & image->music, vocal/instrumental, BPM/Intensity/Lyrics NHUNG trong prompt (KHONG co Duration). Output MP3.

Lyria 3 Pro

Lyria 3 Pro (Google Vertex, Preview) — tao NHAC the he 3: full song (nhac+loi+giong hat), text->music & image->music, toi da 184s. Dieu khien BPM/Intensity/Duration/Instrumental/Lyrics/Structure NHUNG trong prompt. Output MP3 + loi + mo ta.

Gemini 3.1 Flash TTS

Gemini 3.1 Flash TTS (Google Vertex, Preview) - TTS the he 3.1, chat luong cao. 30 giong + 78 ngon ngu BCP-47. WAV 24kHz.

Gemini 2.5 Flash Lite TTS

Gemini 2.5 Flash Lite TTS (Google Vertex, Preview) - TTS re nhat, high-volume. 30 giong + 78 ngon ngu. WAV 24kHz.

Gemini 2.5 Flash TTS

Gemini 2.5 Flash TTS (Google Vertex) - text-to-speech nhanh & re. 30 giong prebuilt (Achernar...Zubenelgenubi, nam/nu) + 78 ngon ngu BCP-47. Output WAV 24kHz. Async qua generateContent + speechConfig.

Gemini-2.5-pro-tts

text-to-speech

Qwen3-TTS Instruct Flash

Qwen3-TTS Instruct Flash của Alibaba — text→giọng nói + ĐIỀU KHIỂN phong cách qua instructions (tốc độ, ngữ điệu, cảm xúc). 19 giọng, đa ngôn ngữ. Trả WAV.

Qwen3-TTS Flash

Qwen3-TTS Flash của Alibaba (Tongyi) — chuyển văn bản thành giọng nói tự nhiên, đa ngôn ngữ (Trung/Anh/Nhật/Hàn/Pháp/Đức/Ý/Nga/Bồ/Tây Ban Nha…). 19 giọng dựng sẵn. Trả file WAV. Tối đa ~600 ký tự/lần.

Seedance 2.5

Text-to-video & image-to-video qua Dreamina Seedance 2.5 (ByteDance) — ban MOI NHAT, chat luong cao nhat (Dola: "For best video quality"). Truyen prompt (t2v) hoac them anh tham chieu (i2v, toi da 10 anh). Duration 5/10s, 6 ti le khung. Async: 5s ~1-3 phut, 10s ~5 phut.

Seedance 2.0

Text-to-video & image-to-video qua Dreamina Seedance 2.0 (ByteDance) — browser automation. Truyen prompt (t2v) hoac them anh tham chieu (i2v). Duration 5/10s, nhieu ti le khung. Async 1-3 phut.

Veo 3.1

Veo 3.1 (Google Vertex, veo-3.1-generate-001 GA) - video generation chat luong cao nhat cua Google. Text-to-video + image-to-video (anh <=20MB, first+last frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, generate audio. Async (~60-90s).

Veo 3.1 Fast

Veo 3.1 Fast (Google Vertex, veo-3.1-fast-generate-001 GA) - video generation nhanh & re. Text-to-video + image-to-video (anh <=20MB, first+last frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, generate audio. Async (~40-70s).

Gemini Omni Flash

Tạo video qua Google Vertex (Gemini Omni Flash) — text→video kèm audio + thinking. Hỗ trợ ảnh→video (Start Image), tỉ lệ 16:9 / 9:16, tối đa 10 giây.

Dreamina-Seedance-2.0-mini

Model tạo VIDEO đa phương thức (Multimodality-to-video) của ByteDance/Doubao: nhận text + ảnh + video + audio tham chiếu → video (có thể kèm audio). 480P/720P, 4-15s, 24fps.

Dreamina-Seedance-2.0-fast

Seedance 2.0 Fast của ByteDance/Doubao — tạo VIDEO đa phương thức nhanh hơn: text + ảnh (1-9) + video (0-3) + audio (0-3) tham chiếu → 1 video (có thể kèm audio). 480P/720P, 4-15s, 24fps; các chế độ Text→Video, Ảnh→Video, Frame đầu/cuối, Tham chiếu đa phương thức.

Dreamina-Seedance-2.0

Seedance 2.0 (bản full) của ByteDance/Doubao — tạo VIDEO đa phương thức: text + ảnh (1-9) + video (0-3) + audio (0-3) tham chiếu → 1 video (có thể kèm audio). Hỗ trợ 480P/720P/1080P/4K, 4-15s, 24fps; các chế độ Text→Video, Ảnh→Video, Frame đầu/cuối, Tham chiếu đa phương thức.

Grok Video 1.5

Tạo video qua grok.com/imagine (browser, cần SuperGrok). Text-to-video (chỉ prompt) hoặc image-to-video (kèm 1 ảnh). Aspect Ratio áp dụng cho text-to-video; có ảnh input (i2v) thì giữ tỉ lệ ảnh gốc.

Veo 3.1 Lite

Veo 3.1 Lite (Google Vertex, veo-3.1-lite-generate-001 - Preview) - video generation re nhat dong Veo. Text-to-video + image-to-video (first frame). Duration 4/6/8 giay, 720p/1080p, 16:9/9:16, 24 FPS, sound generation. Async (~40-70s).

HappyHorse 1.1 (Text→Video)

HappyHorse 1.1 text-to-video của Alibaba (Tongyi) — tạo video từ PROMPT. 720P/1080P, tỉ lệ 16:9/9:16/1:1/4:3/3:4, tối đa ~10s. DashScope async.

Midjourney Video

Midjourney Video (t2v) — tạo ảnh nền từ prompt rồi ANIMATE thành tối đa 4 video mp4 (h264 ~832×464, ~5s). Qua gateway browser (subscription Midjourney). Async ~1–3 phút. Hỗ trợ aspect_ratio + chọn ảnh nền (imageNum).

Wan2.7 Text-to-Video

Wan2.7 text-to-video của Alibaba (Tongyi Wanxiang) — sinh video từ prompt, hỗ trợ tự sự đa cảnh (multi-shot). 720P/1080P, 2-15s. DashScope async (1-5 phút).

HappyHorse 1.0 (Video Editing)

HappyHorse 1.0 video-editing của Alibaba (Tongyi) — sửa VIDEO có sẵn theo instruction (prompt) + tối đa 5 ảnh tham chiếu. 720P/1080P. DashScope async (1-5 phút). ⚠️ Tính giá theo GIÂY video billable = độ dài NGUỒN + độ dài KẾT QUẢ (kết quả giữ nguyên độ dài video nguồn); 720P 42cr/giây · 1080P 72cr/giây.

Wan2.7 Video Editing

Wan2.7 video-editing của Alibaba — sửa video có sẵn theo instruction (đổi phong cách, trang phục, cảnh...) + tối đa 4 ảnh tham chiếu. Output giữ độ dài video nguồn. ⚠️ Tính giá theo GIÂY billable (usage.duration = input+output); 720P 30cr/giây · 1080P 45cr/giây. DashScope async.