Create one valid .creavax-template.json object for Creavax Template Builder.
تمپلت قوانین +قابلیتهای مدلهای منتخب
کانتکست مرجع و قابلخواندن توسط ماشین برای Template Builder کریواکس، شامل اسکیمای فعلی، قوانین گراف و قابلیتهای واقعی مدلهای منتخب و قابلمشاهدهای که در Template Builder استفاده میشوند.
How an AI should build a Creavax Template
این کانتکست مستقل از Provider است. مرجع مدلها فقط از مجموعه فعلی و قابلمشاهده Admin Curated Models میآید؛ با این حال Model slotهای قابلانتقال تمپلت باید بر پایه قابلیت/نیازمندی باقی بمانند تا کریواکس بتواند یک مدل فعال و سازگار را Resolve کند.
Use the live schema in this context as the package authority. Do not invent unsupported fields or input types.
Creavax Model Source-of-Truth Rule: use only models and capabilities explicitly declared by the Creavax Live Models Catalog exposed through this context and GET /ai-models/available. Do not analyze or infer model capabilities yourself.
Never infer, add or expand a capability from a model name, model family, provider name, provider marketing page, external documentation, general AI knowledge, similarity to another model, provider reputation or prior knowledge about that model.
If a capability, operation, input type, output type, feature or runtime control is absent or unknown in the Creavax live model contract, treat it as unsupported for Template compatibility. Unknown is not Supported.
Reference-image support is not equivalent to primary image input, image_to_image or image_edit. Treat those as supported only when Creavax explicitly declares them in the live model capability contract.
If your own knowledge about a model conflicts with Creavax, always follow Creavax. Do not choose, recommend or assume a real model that is not present in the current live_models catalog.
Before creating a Model slot, derive the Template requirements first, then inspect live_models only to understand what Creavax currently declares as available. Keep the slot provider-neutral and requirement-based; never hard-code the selected real model_key.
For numeric buyer controls, use the canonical Template input type reported by the schema. The current canonical numeric type is number; preserve numeric runtime values instead of routing them through select-option fallbacks.
Use live_models only to understand capabilities from the exact visible admin-curated model set exposed in Template Builder. The curated category membership controls discovery/availability; it does not redefine a model's real capability contract.
Keep template model slots portable and requirement-based. Do not hard-code provider IDs or model_key values inside template.models.
Connect user inputs to Agents, Agents to their Models, optional Agents to downstream Agents, and producing Agents to final Outputs.
Each Model slot/node belongs to exactly one Agent. If multiple Agents need the same capability, create separate Model slots; Creavax may still resolve them to the same real model.
Every Agent MUST include a non-empty system_prompt. Missing, null, empty or whitespace-only system_prompt values are invalid. The prompt must describe the Agent role, responsibilities, input handling and how it should use its connected Models.
If a Model slot defines prompt_composition (final Prompt text), it must be non-empty. The literal {{prompt}} placeholder is optional: use it only when you want to choose where structured connected text inputs are inserted; without it, Creavax appends those structured inputs automatically.
Independent Agents may execute in parallel. Agent -> Agent creates both a data edge and an execution dependency, so downstream Agents wait for all connected upstream Agents; cycles are forbidden.
Model slots owned by the same Agent execute sequentially in target_port order: chatModel1 -> chatModel2 -> chatModel3. Parallel execution applies between independent Agents, not between Model slots inside one Agent.
Never connect a Model directly to an Output.
When a user-facing control changes a real model setting such as video duration, aspect ratio, resolution, FPS, quality or output count, use model.setting_bindings when supported by the schema.
Every requirements.inputs, requirements.outputs, requirements.operations and requirements.features entry is mandatory; Creavax intersects all constraints when resolving a real model.
Every model.setting_bindings[].setting is also a mandatory runtime capability/control before model selection. A model that lacks any bound control (for example aspect_ratio, duration_seconds, resolution, fps, quality or output_count) must not be selected or called.
Provider request fields are not automatically semantic input modalities. Prompt/instruction fields, JSON/config values, reference URLs, task IDs, masks, bounding boxes and provider-specific controls must not create a new generation operation unless the endpoint independently supports that operation.
For every setting_binding, the referenced input must exist, its type must be compatible with the setting, the setting must be a canonical supported runtime control, and the binding must not conflict with the Model slot requirements.
If a preferred or referenced model is unavailable or unregistered, Creavax may replace it only with an active model that satisfies the complete inputs + outputs + operations + features + required runtime-controls intersection.
Return valid JSON only when the user asks for the final Creavax template package.
Use the live schema, not assumptions
Schema status: live. Supported input types: text · select · number · image · video · audio · three_d · file. SEO keywords: maximum 10.
{
"schema_version": "1.0",
"type": "creavax_template",
"template": "{...}"
}90 مدل منتخب Template Builder
مدلها دقیقاً از عضویتهای قابلمشاهده دستههای Admin Curated Models که Template Builder استفاده میکند بارگذاری میشوند. عضویت Curated فقط نمایش و کشف مدل را کنترل میکند؛ برای تعریف نیازمندیها از ورودیها، خروجیها، Capability Bucketها و قابلیتهای واقعی هر مدل استفاده کنید و تمپلت قابلانتقال را به دسته، Provider یا model key خاص متصل نکنید.
Creavax 3D Auto
creavax-3d-autotext · image → three_dtext_to_3d · image_to_3dHi3D Image to 3D
hitem3d/hi3d/v3.0/image-to-3dimage → three_dimage_to_3dHi3D Multiview to 3D
hitem3d/hi3d/v3.0/multi-view-to-3dimage → three_dimage_to_3dHunyuan 3D Pro Image to 3D
fal-ai/hunyuan-3d/v3.1/pro/image-to-3dimage → three_dimage_to_3dHunyuan 3D Pro Text to 3D
fal-ai/hunyuan-3d/v3.1/pro/text-to-3dtext → three_dtext_to_3dHyper3D - Rodin V2.5 - Image to 3D
fal-ai/hyper3d/rodin/v2.5image → three_dimage_to_3dHyper3D - Rodin V2.5 - Text to 3D
fal-ai/hyper3d/rodin/v2.5/text-to-3dtext · three_d · json → three_dtext_to_3dMeshy 7.1 Image to 3D
meshy/v7.1/image-to-3dimage → three_dimage_to_3dMeshy 7.1 Multi Image to 3D
meshy/v7.1/multi-image-to-3dimage → three_dimage_to_3dMeshy 7.1 Text to 3D
meshy/v7.1/text-to-3dtext · file · image → three_dtext_to_3dTrellis 2
fal-ai/trellis-2image → three_dimage_to_3dTripo P2 Image to 3D
tripo3d/p2/image-to-3dimage → three_dimage_to_3dTripo P2 Text to 3D
tripo3d/p2/text-to-3dtext → three_dtext_to_3dCreavax Audio Auto
creavax-audio-autotext → audiotext_to_audioDeepgram: Aura-2
deepgram/aura-2text → audiotext_to_audioDeepgram: Nova-3
deepgram/nova-3audio → textaudio_to_textElevenlabs Music v2.5
elevenlabs/music/v2.5text · audio → audiotext_to_audioElevenlabs Sound Effects V2
fal-ai/elevenlabs/sound-effects/v2text → audiotext_to_audioElevenlabs Tts Eleven V3
fal-ai/elevenlabs/tts/eleven-v3text → audiotext_to_audioFish Audio: S2 Pro
fish-audio/s2-protext → audiotext_to_audioGoogle: Lyria 3 Pro Preview
google/lyria-3-pro-previewtext · image → audio · texttext_to_audio · image_to_audioMicrosoft AI: MAI-Transcribe 2
microsoft/mai-transcribe-2audio → textaudio_to_textMicrosoft AI: MAI-Voice-2
microsoft/mai-voice-2text → audiotext_to_audioMiniMax: Speech 2.8 HD
minimax/speech-2.8-hdtext → audiotext_to_audioMiniMax: Speech 2.8 Turbo
minimax/speech-2.8-turbotext → audiotext_to_audioOpenAI: GPT Transcribe
openai/gpt-transcribeaudio → textaudio_to_textOpenAI: Whisper Large V3
openai/whisper-large-v3audio → textaudio_to_textOpenAI: Whisper Large V3 Turbo
openai/whisper-large-v3-turboaudio → textaudio_to_textQwen: Qwen-Audio-3.0-TTS Flash
qwen/qwen-audio-3.0-tts-flashtext → audiotext_to_audioQwen: Qwen-Audio-3.0-TTS Plus
qwen/qwen-audio-3.0-tts-plustext → audiotext_to_audioBlack Forest Labs: FLUX.2 Flex
black-forest-labs/flux.2-flextext · image → imagetext_to_image · image_to_imageBlack Forest Labs: FLUX.2 Max
black-forest-labs/flux.2-maxtext · image → imagetext_to_image · image_to_imageByteDance Seed: Seedream 5.0 Pro
bytedance-seed/seedream-5-0-protext · image → imagetext_to_image · image_to_imageChrono Edit
fal-ai/chrono-edittext · image → imageimage_to_imageCreavax Image Auto
creavax-image-autotext · image → imagetext_to_image · image_to_imageGoogle: Nano Banana 2 (Gemini 3.1 Flash Image)
google/gemini-3.1-flash-imagetext · image → image · texttext_to_image · image_to_imageGoogle: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)
google/gemini-3.1-flash-lite-imagetext · image → image · texttext_to_image · image_to_imageGoogle: Nano Banana Pro (Gemini 3 Pro Image)
google/gemini-3-pro-imagetext · image → image · texttext_to_image · image_to_imageIdeogram V4.0 Text to Image
ideogram/v4text → imagetext_to_imageIdeogram V4.0q Image to Image
ideogram/v4/image-to-imagetext · image → imageimage_to_imageKrea: Krea 2 Large
krea/krea-2-largetext → imagetext_to_imageMeta: Muse Image
meta/muse-imagetext · image → imagetext_to_image · image_to_imageMicrosoft AI: MAI-Image-2.6
microsoft/mai-image-2.6text · image → imagetext_to_image · image_to_imageMicrosoft AI: MAI-Image-2.6 Flash
microsoft/mai-image-2.6-flashtext · image → imagetext_to_image · image_to_imageNano Banana 2
fal-ai/nano-banana-2/edittext · image · video · audio · file → imageimage_to_imageOpenAI: GPT Image 2.5 Flare
openai/gpt-image-2.5-flaretext · image → imagetext_to_image · image_to_imageOpenAI: GPT Image 2.5 Sunburst
openai/gpt-image-2.5-sunbursttext · image → imagetext_to_image · image_to_imageOpenAI: GPT-5.4 Image 2
openai/gpt-5.4-image-2text · image · file → image · texttext_to_image · image_to_image · document_to_imageQwen: Qwen Image 3 Pro
qwen/qwen-image-3-protext · image → imagetext_to_image · image_to_imageRecraft: Recraft V4.1 Pro
recraft/recraft-v4.1-protext · image → imagetext_to_image · image_to_imageSeedream 5.0 Pro Image Editing
bytedance/seedream/v5/pro/edittext · image → imageimage_to_imageSourceful: Riverflow V2.5 Pro
sourceful/riverflow-v2.5-protext · image → imagetext_to_image · image_to_imageAnthropic: Claude Fable 5.1
anthropic/claude-fable-5.1text · image · file → texttext_to_text · image_to_text · document_to_textAnthropic: Claude Opus 5.5
anthropic/claude-opus-5.5text · image · file → texttext_to_text · image_to_text · document_to_textCreavax Text Auto
creavax-text-autotext → texttext_to_textDeepSeek: DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flashtext · image → texttext_to_text · image_to_textGoogle: Gemini 3.8 Flash
google/gemini-3.8-flashtext · image · video · audio · file → texttext_to_text · image_to_text · video_to_text · audio_to_text · document_to_textMoonshotAI: Kimi K2.6
moonshotai/kimi-k2.6text · image → texttext_to_text · image_to_textOpenAI: GPT-5.6 Luna
openai/gpt-5.6-lunatext · image · file → texttext_to_text · image_to_text · document_to_textOpenAI: GPT-5.6 Sol
openai/gpt-5.6-soltext · image · file → texttext_to_text · image_to_text · document_to_textOpenAI: GPT-6 Astra
openai/gpt-6-astratext · image · file → texttext_to_text · image_to_text · document_to_textOpenAI: GPT-6 Astra Pro
openai/gpt-6-astra-protext · image · file → texttext_to_text · image_to_text · document_to_textQwen: Qwen3.8 27B
qwen/qwen3.8-27btext · image · video → texttext_to_text · image_to_text · video_to_textQwen: Qwen3.8 Max (0902)
qwen/qwen3.8-max-0902text · image · video → texttext_to_text · image_to_text · video_to_textSpaceXAI: Grok 4.20
x-ai/grok-4.20text · image · file → texttext_to_text · image_to_text · document_to_textXiaomi: MiMo-V2.5-Pro
xiaomi/mimo-v2.5-protext → texttext_to_textZ.ai: GLM 5.3 Flash
z-ai/glm-5.3-flashtext · image · video → texttext_to_text · image_to_text · video_to_textBria Increase Resolution: Upscale Images up to 4x Without Losing Detail | fal
bria/increase-resolutionimage → imageimage_upscaleClarity Upscaler
fal-ai/clarity-upscalerimage → imageimage_upscaleCreavax Upscale Auto
creavax-upscale-autoimage → imageimage_upscaleCrystal Upscaler
clarityai/crystal-upscalerimage → imageimage_upscaleRecraft Creative Upscale
fal-ai/recraft/upscale/creativeimage → imageimage_upscaleRecraft Crisp Upscale
fal-ai/recraft/upscale/crispimage → imageimage_upscaleSeedVR2
fal-ai/seedvr/upscale/image/seamlessimage → imageimage_upscaleSeedVR2
fal-ai/seedvr/upscale/imageimage → imageimage_upscaleTopaz Restore Image
topaz/restore/imageimage → imageimage_to_imageTopaz Upscale Image Generative
topaz/upscale/image/generativeimage → imageimage_upscaleTopaz Upscale Image Precision
topaz/upscale/image/precisionimage → imageimage_upscaleAlibaba: Wan 3.0 Prime
alibaba/wan-3.0-primetext · image → videotext_to_video · image_to_videoBlack Forest Labs: FLUX.3 Video
black-forest-labs/flux-3-videotext · image · video → videotext_to_video · image_to_video · video_reference_to_videoByteDance: Seedance 2.5
bytedance/seedance-2.5text · image · video · audio → videotext_to_video · image_to_video · video_reference_to_video · audio_to_videoCreavax Video Auto
creavax-video-autotext · image · video → videotext_to_video · image_to_video · video_reference_to_videoGoogle: Veo 3.1
google/veo-3.1text · image → videotext_to_video · image_to_videoGoogle: Veo 3.1 Fast
google/veo-3.1-fasttext · image → videotext_to_video · image_to_videoKling: Video O1
kwaivgi/kling-video-o1text · image → videotext_to_video · image_to_videoKling: Video v3.0 Pro
kwaivgi/kling-v3.0-protext · image → videotext_to_video · image_to_videoMiniMax: H3 Max
minimax/hailuo-3-maxtext · image → videotext_to_video · image_to_videoRunway: Aleph 2.0
runway/aleph-2text · image · video → videovideo_reference_to_videoRunway: Gen-4.5
runway/gen-4.5text · image → videotext_to_video · image_to_videoSpaceXAI: Grok Imagine Video 1.5
x-ai/grok-imagine-video-1.5text · image → videotext_to_video · image_to_video