Chat, image, video, audio. Prices in Rupiah, one API key for all, pay for what you use.
gem-3.8-flash · GoogleLight, fast model for agents, coding and long-running tasks. Stable for large-scale production, a good fit for customer service chatbots.
tt-6-luna · OpenAIFlagship model with stronger reasoning, coding and multimodal understanding, plus very long context. Built for professional development and high-quality chat.
tt-6.1-terra · OpenAIThe cost-efficient member of the TT-6.1 series: strong reasoning, coding and multimodal input with very long context. Good for everyday coding and high-volume chat.
hk-5-5 · AnthropicFast, low-cost lightweight model with improved coding and tool calling over the last generation. Suited to high-volume daily tasks and agent workflows.
gem-4-argon · GoogleGEM's fourth-generation flagship reasoning model, with much stronger coding and agent skills, native multimodal input and very long context. Built for hard, complex tasks.
tt-6.1-sol · OpenAIFlagship model with stronger reasoning, coding and multimodal understanding, plus very long context. Built for professional development and high-quality chat.
tt-6-astra · OpenAIFlagship model with stronger reasoning, coding and multimodal understanding, plus very long context. Built for professional development and high-quality chat.
gem-3.5-flash · GoogleLight, fast model for agents, coding and long-running tasks. Stable for large-scale production, a good fit for customer service chatbots.
tt-5.6-sol · OpenAIFlagship model with smoother conversation, multimodal understanding and better coding. Suited to high-quality multi-turn chat and professional work.
gem-3.1-pro · GoogleAdvanced reasoning model with broad world knowledge and cross-modal understanding. Suited to complex analysis, research and tasks that need deep thinking.
op-5-5 · AnthropicThe latest OP flagship, built for deep reasoning, complex coding and long-horizon tasks. Careful, reliable output for demanding professional work.
kimi-k3.1-pro · MoonshotMoonshot AI's flagship reasoning model with thinking always on, image understanding, and stronger coding, reasoning and agent skills. Standard and fast speed options.
kimi-k3 · MoonshotMoonshot AI's flagship reasoning model with 1M-token context, thinking always on, and native image and video understanding. Strong at coding, reasoning and agents.
gem-3.7-flash · GoogleLight, fast model for agents, coding and long-running tasks. Stable for large-scale production, a good fit for customer service chatbots.
tt-5.5 · OpenAIFlagship model with smoother conversation, multimodal understanding and better coding. Suited to high-quality multi-turn chat and professional work.
deepseek-v4.1-flash · DeepSeekLightweight DS V4.1: cost-efficient, high throughput, image input, up to a million tokens of context and tool calling. Good for everyday chat and light text tasks.
MiniMax-M3.1-Flash · MiniMaxThe fast, lightweight MiniMax M3.1, keeping M3's coding and agent skills at a lower cost, with 1M context and image understanding.
fb-5 · AnthropicNew-generation FB flagship with adaptive thinking that sets its depth per task. Stronger at complex reasoning, long-horizon tasks and creative writing.
tt-5.6-luna · OpenAIFlagship model with smoother conversation, multimodal understanding and better coding. Suited to high-quality multi-turn chat and professional work.
tt-5.4 · OpenAIModel for complex professional work, with deep reasoning, multimodal understanding and tool calling. Suited to analysis, coding and creative writing.
gem-3.6-flash · GoogleLight, fast model for agents, coding and long-running tasks. Stable for large-scale production, a good fit for customer service chatbots.
deepseek-v4-pro · DeepSeekFlagship open-source DS V4 model with top-tier reasoning and agent skills and very long context. A cost-efficient pick for coding and agents.
tt-6-sol · OpenAINew-generation flagship with stronger reasoning, coding and multimodal understanding. Very long context, suited to professional development and high-quality multi-turn chat.
doubao-seed-2-1-pro-260628 · ByteDanceByteDance flagship with very long context, deep thinking and web search. Reads images, video and PDFs together, suited to complex reasoning, long documents and expert Q&A.
fb-5-1 · AnthropicUpgrade to FB-5 with always-on adaptive thinking that scales depth to the task. Stronger at complex reasoning, long-running tasks, coding and creative writing.
op-5 · AnthropicLatest flagship of the series, with the strongest deep reasoning, complex coding and long-running task ability in its line. Careful, reliable output for demanding professional work.
qwen3.8-max · AlibabaLatest Qwen flagship from Alibaba, upgraded for complex reasoning, knowledge Q&A and professional writing. Very long context, web search and document understanding.
tt-5.6-terra · OpenAIFlagship model with smoother conversation, better multimodal understanding and stronger coding. Suited to high-quality multi-turn chat and professional use.
op-4-8 · AnthropicLatest-generation flagship of the series, sharper at logical reasoning, mathematical proofs and creative writing with a natural, polished style.
glm-5.3 · ZhipuZhipu flagship for software engineering and agent work, with much stronger coding than GLM-5.2. 1M context, up to 128K output, always-on thinking and structured tool calling.
gem-3-pro · GoogleModel with very long context for reading and analyzing large documents, whole books or long videos, then answering detailed questions. Multimodal input and web search.
op-4-6 · AnthropicA model in this series strong at logical reasoning, mathematical proofs and detailed creative writing, with a natural, polished style.
sn-5-5 · AnthropicBalanced new-generation flagship with stronger coding and agent skills. Rigorous logic and reliable output, built for coding, agents and business workflows.
doubao-seed-2-0-pro-260215 · ByteDanceByteDance flagship with 256K context, deep thinking and web search. Understands images, video and PDFs, and performs well at reasoning, coding and multimodal analysis.
deepseek-v4-flash · DeepSeekLightweight DeepSeek V4 model focused on value and throughput, suited to general chat and basic text tasks. Supports up to 1M-token context.
glm-5.3-flash · ZhipuFast GLM-5.3 variant keeping 1M context and solid coding and reasoning, with quicker responses and lower cost. Accepts images, video and documents plus web search; good for daily tasks.
gk-4.6 · xAINew-generation flagship improved at long-chain reasoning, agent tool calling and long context, with low hallucination and strict instruction following. For coding, strategy analysis and real-time chat.
sn-5 · AnthropicBalanced flagship with strong coding and agent skills. Rigorous logic and reliable output, built for coding, agents and business workflows.
deepseek-v4-flash-vision-exp · DeepSeekExperimental DeepSeek V4 Flash with image understanding at Flash-level cost. 1M-token context and tool calling; good for image chat, long documents and multimodal agents.
qwen3.8-max-0902 · AlibabaSnapshot of Qwen 3.8 Max with stronger image and video understanding and deep thinking. For complex reasoning, knowledge Q&A and professional writing, with very long context.
op-4-7 · AnthropicNew-generation flagship of the series, improved at logical reasoning, mathematical proofs and creative writing with a natural, polished style.
deepseek-v4-pro-0813 · DeepSeekThe 0813 snapshot of DeepSeek V4 Pro, with top-tier reasoning and agent ability. Very long context, suited to complex tasks and in-depth analysis.
gk-4.7 · xAINew-generation flagship with further gains in long-chain reasoning, agent tool calling and long context, low hallucination and strict instruction following. For coding, strategy and real-time chat.
tt-5.4-mini · OpenAIA faster, more efficient take on TT-5.4, built for high-volume work like customer service bots, summaries, and everyday automation.
doubao-seed-2-1-turbo-260628 · ByteDanceByteDance's fast model balancing speed and quality. Long context, deep thinking, web search, and understands images, video, and PDFs.
qwen3.8-flash · AlibabaThe fast, cost-efficient member of the Qwen 3.8 series. Reads images and documents, searches the web, very long context. Suited to high-frequency daily tasks.
kimi-k2.6 · MoonshotMoonshot AI's general reasoning model with steadier long-horizon coding, better instruction following, and self-correction. Good for chat and agent tasks.
glm-5.3-flashx · ZhipuA faster GLM-5.3 family model that keeps solid coding and reasoning. Takes images, video, and documents, plus web search. Fits daily tasks and light agent workflows.
tt-4o · OpenAIAn all-round multimodal model that understands text and images, fast and cost-effective. Good for everyday chat, image analysis, and coding help.
glm-5.2 · ZhipuFlagship open-source GLM-5 model (744B MoE) with 1M context and up to 128K output. Focused on agentic coding, long-horizon planning, and structured tool calls.
tt-5.2 · OpenAIA coding-tuned model that can read an entire codebase at once, helping with system-level architecture improvements and bug hunting.
hk-4.5 · AnthropicA balanced, safety-focused all-rounder with computer use: it can click and operate software to handle repetitive work for you.
sn-4-6 · AnthropicA balanced, capable model for coding, agents, and business workflows, with a focus on safety and natural responses.
MiniMax-M3 · MiniMaxMiniMax's flagship agentic model for coding, tool calling, and agent workflows. 1M context with image and video understanding.
gk-4.5 · xAINext-gen GK with stronger complex reasoning, agent tool calling, and real-time information, low hallucination and strict instruction following. Good for coding, strategy, and live chat.
qwen3.7-max · AlibabaFlagship of the Qwen 3.7 series: natively multimodal with 1M-token context and web search. Understands images, video, and documents; strong at complex reasoning and Q&A.
gem-3.1-flash · GoogleThe most cost-efficient multimodal model in the GEM 3.1 series, fast for high-volume light work: bulk agent tasks, simple data extraction, and low-latency apps.
kimi/kimi-k2.7-code-highspeed · MoonshotThe high-speed version of Kimi K2.7 Code, with quicker responses and higher throughput for real-time coding and agent work.
deepseek-v4-flash-0731 · DeepSeekThe 0731 snapshot of DS V4 Flash: cost-efficient and high-throughput for general chat and basic text tasks, with up to 1M-token context.
op-4-5 · AnthropicThe smartest model in its series, strong at complex logical reasoning, math, and creative writing with polished, natural prose.
kimi/kimi-k2.7-code · MoonshotMoonshot AI's coding model for programming and agents: code generation, completion, refactoring, and multi-step tool calls with very long context.
doubao-seed-evolving · ByteDanceA Doubao model that automatically tracks the latest generation. Long context, deep thinking, web search, and image/video/PDF understanding. Good for long-document analysis.
gk-4-20 · xAIA fast GK model with strong tool calling, low hallucination, and strict instruction following. A thinking toggle: off for quick chat, on to reason before answering.
stepfun/step-3.7-flash · StepFunStepFun's responsive, cost-efficient multimodal model with visual understanding, reasoning, and long context. Fits daily chat, document reading, and high-frequency calls.
mimo-v2.6-pro · XiaomiXiaomi's in-house flagship for text chat and image understanding, with deep thinking and tool calling. Suited to complex reasoning and long context.
gem-3.5-flash-lite · GoogleThe lightest, lowest-cost model in the GEM 3.5 series, suited to bulk work like translation, classification, and customer service. Supports multimodal input.
qwen3.5-plus · AlibabaA natively multimodal model with 1M-token context, deep thinking, and web search. Understands images and video; strong at reasoning, coding, and structured output.
gem-3-flash · GoogleThe lightweight version of GEM 3, low-latency and cost-effective, well suited to large-scale data extraction or real-time voice conversations.
qwen3.5-flash · AlibabaA fast multimodal model with 1M-token context, deep thinking, and web search. Understands images and video, low cost, good for everyday use.
qwen3.6-plus · AlibabaA new-generation natively multimodal model with 1M-token context, deep thinking, and web search. Improved complex reasoning, coding, and multimodal understanding.
tt-5.4-nano · OpenAIThe lightest, fastest version of TT-5.4, built for tasks where speed and cost matter most, like classification or automated chat replies.
qwen3.8-flash-next · AlibabaA fast model on a new efficient architecture, with better inference efficiency and value. Reads images and documents, very long context, suited to daily tasks.
qwen3.7-plus · AlibabaA new-generation natively multimodal model with 1M-token context, deep thinking, and web search. Strong at complex reasoning, coding, and multimodal understanding.
MiniMax-M2.7 · MiniMaxMiniMax's flagship reasoning model for coding, tool calling, search, and office productivity. 204K context with deep thinking.
doubao-seed-character-260628 · ByteDanceDoubao's native roleplay model, good at keeping personas, playing out storylines, and immersive multi-turn chat. Takes image/video/audio input with very long context.
gk-4.3 · xAIA GK model tuned for low hallucination, agent tool calling, and strict instruction following. Good for complex reasoning, coding, strategy, and real-time chat.
qwen3.6-flash · AlibabaThe fast, cost-efficient member of the Qwen 3.6 series: natively multimodal with deep thinking and web search. Suited to high-frequency daily tasks.
gk-build-0.1 · xAIA GK coding model for agentic coding, tool calling, and multi-step task execution with adaptive reasoning. Good for code collaboration and automated builds.
seedance-2.0-anci · ByteDanceSeedance 2.0 reference mode: upload 0–9 images for a 720p video with sound, 4–15 s, five aspect ratios. Flat fee per video, not per second. Failed jobs are refunded automatically.
suno-v6 · SunoCreate full vocal songs or instrumentals from lyrics and a style prompt. Comes in V6, V6 Wild and V6 Mini; each run returns 2 tracks to choose from.
seedance-2.5-guanfang-anmiao · ByteDanceFull Seedance 2.5 (text, first-last frame, reference, edit and extend), 4–30 s videos with sound. Real-person photos and videos need no face verification. Billed per second.
tt-image-2.5 · OpenAIImage model for posters, e-commerce product shots and social content. 1K/2K/4K output (up to 3840×2160), transparent backgrounds, multi-image reference and local edits.
gk-video-3.5 · xAITurn one first-frame image into a 1–15 s short video with audio. Flexible aspect ratio and duration; fast and low-cost for TikTok or Reels content.
kling-v4-video · KuaishouKling 4th generation: text-to-video or reference with up to 10 images and 5 videos. 720p/1080p, 3–30 s, 21:9 widescreen, consistent subjects. Billed per second.
speech-2.8 · MiniMaxClone a voice from an audio sample; once activated it stays available for good. HD or Turbo quality, with control over speed, pitch and emotion. For voice-overs and podcasts.
seedance-2.5-guanfang · ByteDanceOne model for text, first-last frame and reference (up to 30 images, 10 videos, 10 audio). 4–30 s, 480p–1080p, with sound, no watermark; supports edit and extend.
tt-image-2.5-token · OpenAISame as TT Image 2.5, billed by actual token usage. Up to 4K, transparent backgrounds, multi-image reference, local edits, and more precise resolution and aspect ratio.
viduq4-preview · ViduReference mode: 1–15 images plus up to 3 mp3 audio clips, five aspect ratios. First-frame mode: one image. Videos with sound from 540P to 4K, 3–16 s, billed per second.
tt-image-2 · OpenAIFlagship image model strong at in-image text and complex layouts: infographics, comics, text-heavy posters. Faces, logos and product details stay consistent across edits.
kwvideo-v2-ref · ByteDanceSeedance 2.0 reference mode: upload 1–9 images and it blends style, elements and composition into a 4–15 s video with sound. Standard and fast versions. Billed by token.
suno-v4.5 · SunoCreate full vocal songs from lyrics and a style prompt, plus instrumentals, song extension and karaoke subtitles. Up to 4 minutes long; a song can be ready in about 80 s.
minimax-h3 · MiniMaxHailuo H3 in one model: text, first-last frame, reference and omni reference (image, video, audio). Outputs 2K video with sound, 4–15 s. Billed per second.
banana-2.1 · GoogleFast Banana-series image model, charged per image. 1K/2K/4K output, up to 14 reference images, many aspect ratios, plus web search. Good for bulk product and social images.
seedance-2.0-guanfang · ByteDanceOne model for text, first-last frame and image/video/audio reference. Real-person media needs no face verification. Standard, fast and Mini; 480p–4K, with sound, no watermark.
seedance-2.0-guanfang-anmiao · ByteDanceFull Seedance 2.0 (text, first-last frame, multimodal reference), up to 4K with sound. Pick 4–15 s; charged once at order. No face verification for real people.
banana-2.1-token · GoogleBanana 2.1 billed by actual token usage. 1K/2K/4K output, up to 14 reference images, many aspect ratios and web search. For bulk e-commerce and social content.
hailuo-h3-cankaosheng · MiniMaxUpload 1–9 reference images and it blends subject, style and composition into a 2K video with sound. At least one image required; text-only isn't supported. First 5 images free.
music-2.5+ · MiniMaxMake songs from lyrics and a style, or instrumentals from a style prompt alone. AI lyric writing, 14+ song-structure tags and genre blending. Handy for ad jingles.
omni_flash-10s · GoogleOne prompt becomes a 10 s 720P video with sound effects. Upload up to 7 reference images to lock people, products and scenes. Good for short ads and creative clips.
tt-image-2-token · OpenAISame as TT Image 2, billed by actual token usage. Strong at in-image text, complex layouts, multi-image composition and faithful edits; faces and logos stay consistent.
omni-flash · GoogleShort videos with sound effects from text, or from 1–3 reference images for people, objects or style. 16:9 or 9:16, 4/6/8/10 s. Fits TikTok and Reels.
kling-v3-omni-cankao · KuaishouKling V3 Omni reference mode: text only or 1–7 images, with AI-planned shots into a 5–15 s video with sound. Standard and high-quality modes.
banana-pro · GooglePro image model, very strong at in-image text: multilingual posters, logos, infographics. Up to 4K, 10 aspect ratios, 14 references (5 people, 6 objects), chat-based editing.
wan3.0-video-quannengcankao · AlibabaMix images, video, audio and text as references to reuse characters, products, scenes and voices. Outputs 1080P video with sound up to 30 s (including reference video).
doubao-tts-2.0 · ByteDanceText-to-speech with hundreds of voices, multiple emotions and languages, and long text up to 100,000 characters. Natural sound for voice-overs and audiobooks.
seedance-2.5-anmiao-quannengcankao · ByteDanceSeedance 2.5 multimodal reference: up to 50 assets (30 images, 10 videos, 10 audio; at least 1) for a seamless 30 s video in one pass. Billed per second. Failed jobs are refunded automatically.
banana-2 · GoogleThe efficient take on Banana Pro: near-flagship quality, faster and cheaper. Up to 4K, 14 references, 512px previews, ultra-wide ratios up to 8:1, plus web search.
doubao-seedance-2-5-cankaosheng · ByteDanceSeedance 2.5 reference mode: upload 1–30 images and it blends characters, style and composition into a cinematic 4–30 s video with sound. Billed by token.
gk-video-3 · xAIQuickly turn a still image into a 720P video of 6 or 10 seconds. Handy for creators testing ideas or gathering short-form footage.
doubao-seedream-5-0-pro-260628 · ByteDanceSeedream 5.0 Pro flagship image model: sharper detail, up to 10 reference images, local edits and style transfer. 1K/2K output for sales posters, brand visuals and portraits.
gem-3.1-tts · GoogleText-to-speech with 30 voices and 24 languages, two-speaker dialogue and tone set by plain description. Good for audiobooks, audio fiction and narration.
wan3.0 · AlibabaOne model for text, first-last frame, reference (images, PPT/PDF/Word docs, web links) and omni reference. Up to 30 s, 1080P, with sound. Billed per second.
banana-pro-token · GoogleBanana Pro billed by token. Very strong at in-image text, up to 4K, 14 references (5 people, 6 objects), chat-based edits. For posters, packaging and logos.
hailuo-h3-shouweizhen · MiniMaxUpload a first frame, or first and last frames, with a prompt for a 2K video with sound. Ratio follows the image, 4–15 s, billed per second. At least one image required.
hailuo-h3-quannengcankao · MiniMaxCombine text, images, video and audio as references for a 2K video with sound. Needs at least one image or video. Audio is free, and the first 5 images are free.
gk-image-2.0 · xAIImage model with realistic texture, consistent characters and better lighting. Text-to-image or edits with up to 3 references, 1K/2K output, 5 ratios. For product and social shots.
kwvideo-v2-quannengcankao · ByteDanceSeedance 2.0 multimodal reference: combine text, images, video and audio (at least one image or video) into a video with sound at 480p, 720p or 1080p. Billed by token.
music-2.5 · MiniMaxCreate full songs from lyrics and a style, with AI lyrics, 14+ song-structure tags and genre blending (pop, rock, electronic, classical, jazz). Clear vocals for jingles and content.
kling-avatar-image2video · KuaishouTurn a portrait into a talking avatar: upload a photo and audio, and the lips sync to the voice. Standard and pro modes, with control over gestures, emotion and camera moves.
banana-2-token · GoogleFast image model with near-flagship quality. Up to 4K output, 14 reference images, wide ratios like 1:4 and 8:1, and web search grounding. Great for bulk product and social content.
hailuo-h3 · MiniMaxText to 2K video with sound and natural, realistic motion. 4–15 seconds, six aspect ratios. Draft at 768P first, then upgrade the clip you like to 2K.
seedance-2.0-anmiao-quannengcankao · ByteDanceVideo from any mix of text, images, video and audio (up to 9 images + 3 videos + 3 audio; at least 1 image or video). 480p–1080p, billed per second. Failed jobs are refunded automatically.
mj_imagine · MidjourneyImage generator known for cinematic aesthetics. Native 2K output, any aspect ratio up to 14:1, style reference (sref) and character reference (oref). Ideal for posters, illustration and concept art.
seedance-2.5-anmiao · ByteDanceUpload 1–30 reference images and the model blends characters, style and composition into a 4–30 second video with sound. 480p–1080p, billed per second. Failed jobs are refunded automatically.
gem-2.5-tts · GoogleText-to-speech with 30 preset voices and 24 languages. Supports two-speaker dialogue and tone control in plain language. Good for ad voice-overs, audiobooks and long-form narration.
seedance-2.0-anmiao-cankaosheng · ByteDanceUpload 1–9 reference images for a 4–15 second video with sound. Mini, Fast and Standard tiers; 480p up to 4K (Mini and Fast up to 720p). Billed per second; failed jobs are refunded automatically.
vidu-image-2 · ViduImage model built for subject consistency: up to 7 references keep people, outfits and props the same across images. Generate, reference or edit at 1K/2K/3K. Great for product shots and brand characters.
happyhorse-r2v · AlibabaBlend characters or objects from several reference images into one short clip. 720P/1080P, about 3–15 seconds, multiple aspect ratios including 16:9 and 9:16.
video-enhance · 其他Upscale any video (up to 2K input) to 720P–8K, with optional frame interpolation to 60/120fps plus sharpening and color correction. For AI videos, short dramas, UGC and old footage.
doubao-seedream-5-0-260128 · ByteDanceByteDance image model with deep thinking and web search, turning live information into data posters and infographics. Up to 3K, multi-image fusion, and precise local edits that leave the rest untouched.
doubao-seedance-2-5-260628 · ByteDanceText to 4–30 second video with sound, including a single continuous 30-second shot. Prompts and voice-over in 11 languages, with near-live-action quality. Billed by tokens.
omni-1.1 · GoogleVideo from text or a first-frame image, with built-in sound effects. Adds 1080P and 4K output, any length from 3 to 10 seconds, faithful first frames and stable motion.
wan2.7-image · AlibabaGenerate and edit in one model: up to 9 references, sets of up to 12 images, Pro up to 4K. Clean in-image text in about 12 languages, including tables and formulas. For posters, e-commerce and storyboards.
doubao-seedance-2-5-quannengcankao · ByteDanceVideo from up to 50 references at once (30 images + 10 videos + 10 audio), even audio alone. Edit or extend videos by instruction, with continuous 30-second output. Billed by tokens.
kling-v3-video · KuaishouKling's third-generation video model: text or image to video, with first and last frame control. Standard and pro modes, 5–15 second videos with sound and improved motion.
doubao-seedream-4-5-251128 · ByteDanceByteDance image model up to 4K, strong at dense text and layout: posters, key visuals, logos and small print. Single or multi-image editing keeps details intact; good for product shots and ads.
kwvideo-v2 · ByteDanceVideo from text, a first frame, or first and last frames, with sound generated automatically. 4–15 seconds, standard or fast version. Billed by tokens.
viduq3 · ViduVideo from text, a first frame, or a first-to-last frame transition. Dialogue and sound effects are generated with the video. Turbo and pro versions, up to 16 seconds.
kling-v3-omni · KuaishouKling's Omni image model, up to 4K, with series mode: 2–9 images at once with consistent characters, scenes and style. Up to 10 references; for storyboards, posters and brand visuals.
viduq3-turbo-cankaosheng · ViduUpload 1–7 reference images for a video with sound and consistent subjects, or use text only. 540P–1080P, 1–16 seconds, fast and economical. Good for animated comics and brand characters.
wan3.0-video · AlibabaText to 1080P video with sound, up to 30 seconds. Pick any length from 2 to 30 seconds or let the model set it from the story. Standard and fast versions.
seedance-2.0-anmiao · ByteDanceVideo from text, a first frame, or first and last frames, with sound, 4–15 seconds. Mini, Fast and Standard; 480p to 4K (Mini and Fast up to 720p). Billed per second; failed jobs are refunded.
viduq3-drama · ViduVideo model for short dramas and animated comics. Upload 1–7 reference images (required) for scenes with sound, consistent characters, accurate blocking and dialogue, and strong camera storytelling.
wan2.6-image · AlibabaImage model for editing and blending references, up to 4 images (e.g. style from one, background from another). Text replacement, region editing, edits up to 2K. Good for product photo retouching.
wan3.0-video-cankaosheng · AlibabaVideo from text or up to 10 reference images for characters and scenes. Also turns PPT/PDF/Word files or web links into narrated 1080P explainer videos up to 30 seconds.
kling-v3-omni-videoref · KuaishouUpload a reference video plus 0–4 optional images. Create a new video using its camera moves or style, or edit the original by instruction. Billed per second.
kling-v3 · KuaishouKling's third-generation image model from text or one reference image, 1K or 2K, up to 9 images per run. Fast and economical for social posts, simple posters and product shots.
happyhorse-video-edit · AlibabaEdit an existing video by instruction: restyle, change outfits or rewrite parts, with optional reference images. 720P/1080P. Billed on input plus output video length combined.
hailuo-h3-max-shouweizhen · MiniMaxUpload a first frame, or first and last frames, plus a prompt for a quick 480P/768P video with sound. Ratio follows the image, 5–15 seconds. At least one image is required.
hailuo-h3-max · MiniMaxFaster version of Hailuo H3: text to 480P/768P video with sound and natural motion. 5–15 seconds, six aspect ratios. Good for quick drafts and batch production.
kling-motion-control-v3 · KuaishouUpload a photo and a motion video, and the person in the photo moves like the video. Built on the V3 engine, standard and pro modes, optional original audio, orientation from photo or video.
seedance-2.5-anmiao-shouweizhen · ByteDanceUpload a first frame, or first and last frames, plus a prompt for a 4–30 second video with sound and near-live-action quality. 480p–1080p, billed per second. Failed jobs are refunded automatically.
pixverse-v6-shouweizhen · PixVerseVideo from text (no image), a first frame (1 image) or first and last frames (2 images), with sound generated automatically. 360P–1080P, 3–15 seconds.
minimax-h3-max · MiniMaxFast Hailuo H3 variant covering both text-to-video and first-last frame in one model. 480P/768P video with sound, 5–15 seconds. Good for quick drafts and batch production.
vidu-mv · ViduUpload 1–7 character or style images plus one audio track (1–180 seconds) to create a music video with a storyline. 540p–1080p, multiple ratios, optional lip sync and auto subtitles.
vo3.1 · GoogleHighly controllable video model with first and last frame control and precise camera instructions. Videos come with background music; good for ads and professional content.
kling-motion-control · KuaishouUpload a photo and a motion video, and the person in the photo moves like the video. Standard and pro modes, optional original audio, orientation from photo or video.
doubao-seedance-2-5-shouweizhen · ByteDanceUpload a first frame, or first and last frames, plus a prompt for a 4–30 second video with sound. Ratio follows the first frame, with near-live-action quality. Billed by tokens.
kling-v3-omni-shouweizhen · KuaishouKling V3 Omni from 1 image (first frame) or 2 images (first + last), with automatic shot planning and sound. Standard or high-quality mode, 5–15 s.
happyhorse-1.1-r2v · AlibabaBlend characters or objects from several reference images into one short clip. 720P/1080P, about 3–15 s, aspect ratios like 16:9 and 9:16.
happyhorse-1.1-t2v · AlibabaTurn a text prompt into a short video with smooth motion and natural lighting. 720P/1080P, 3–15 s, 5 aspect ratios. Good for TikTok and Reels.
omni_flash-10s-fl · GoogleUpload a first and last frame and the model fills in the camera motion and transition. 10-second 720P output with sound effects, great for product transitions.
wan2.2-animate-mix · AlibabaUpload one person photo and one reference video, and the AI replaces the person in the video with the one in the photo. No prompt needed; standard and pro modes.
wan3.0-video-shouweizhen · AlibabaImage-to-video from 1 image (first frame) or 2 (first-to-last transition): 1080P video with sound, up to 30 s. Auto or custom 2–30 s duration, standard or fast version.
wan2.6-cankaosheng · AlibabaPulls characters from reference images to create single- or multi-character interaction videos. Fast or high-quality mode, 720P/1080P, up to 10 s.
seedance-2.5-anmiao-wensheng · ByteDanceSeedance 2.5 text mode: a prompt becomes a 4–30 s video with sound, including 30 s single takes. Billed per second, 480p/720p/1080p. Failed jobs are refunded automatically.
viduq3-turbo · ViduFast video with sound: 1 image for a first frame, 2 for a first-to-last transition, with dialogue and sound effects built in. 540P–1080P, 1–16 s. Good for batch ad videos.
happyhorse-t2v · AlibabaTurn a text prompt into a short video with smooth motion and natural lighting. 720P/1080P, 3–15 s, 5 aspect ratios. Fits product content and Reels.
happyhorse-i2v · AlibabaUses one image as the first frame plus a prompt to make a smooth, natural short video. Image required. 720P/1080P, 3–15 s, aspect ratio follows the image.
wan2.7-shouweizhen · AlibabaWan 2.7 image-to-video: upload 1 image for a first frame, or 2 images for first and last frames. 720P/1080P, up to 15 s.
pixverse-c1-cankaosheng · PixVerseTuned for dynamic scenes like fights, spell effects, and fast motion. No image means text-to-video; with images (up to 7) it uses reference mode. Auto sound, 360P–1080P, 1–15 s.
pixverse-c1-shouweizhen · PixVerseTuned for fights, spell effects, and fast motion. First-frame or first-and-last-frame mode, auto sound, 360P–1080P, 1–15 s.
viduq3-cankaosheng · ViduUpload 1–7 reference images and get a video with sound that keeps the subject consistent. Optimized for comic-style dramas, up to 1080P, 4–16 s.
wan2.6-shouzheng · AlibabaCreate video from a first-frame image or from text alone, with sound included. Fast or high-quality mode, 720P/1080P, up to 15 s.
wan2.7-cankaosheng · AlibabaTwo modes in one: text alone runs text-to-video, adding reference images or video switches to reference mode. 720P/1080P, up to 15 s.
vidu-jieshuoman · ViduScript-driven narrated comic videos with 3–10 character/scene/prop references, per-character voices, Chinese/English TTS, and lip-sync. 720P/1080P. Processing can take up to about 10 hours.
happyhorse-1.1-i2v · AlibabaOne image as the first frame plus a prompt gives a smooth, natural short video. Image required. 720P/1080P, 3–15 s, aspect ratio follows the image. Good for product catalogs.
pixverse-v5.6-shouweizhen · PixVerseThree modes: no image runs text-to-video, 1 image sets the first frame, 2 images set first and last frames. Auto sound, 360P–1080P, 5–10 s.
pixverse-v5.6-r2v · PixVerseUpload 1–7 reference images and the AI blends their characters, style, and scenes into a video with sound. 360P–1080P, 5–10 s.
viduq2-cankaosheng · ViduUpload 1–7 reference images for style-consistent video. Standard mode (rich detail) or high-quality mode (supports video reference and editing). Up to 1080P, 5–10 s.
wan2.7-xuxie · AlibabaExtends an existing video clip, optionally with a last-frame image to guide where it goes. 720P/1080P, up to 15 s.
hailuo-h3-zaishengcheng · MiniMaxUpscales a 768P Hailuo H3 video to 2K, billed per output second. Only works on Hailuo H3 768P outputs; if you need 2K from the start, use Hailuo H3 Reference directly.
No models match.
Prices in Rupiah. Chat is billed per 1M input / output tokens, images and video per call or per second. The console shows the price that applies.