Zyka AI Media Generation#

Generate AI videos, images, voice, and use AI-powered apps directly from the terminal using the zyka CLI. One command, 40+ AI models. No code writing required.

Setup#

The zyka CLI is published on npm: https://www.npmjs.com/package/zyka Source: https://github.com/kshitijdixit/zyka-sdk

Required: install the CLI once before using this skill. All examples below assume zyka is on your PATH.

bash

npm install -g zyka            # one-time install (latest)
zyka --help                    # verify install

Set your API key (required — get one at https://zyka.ai/settings/api-keys):

bash

export ZYKA_API_KEY=zk_live_...

Security & Privacy#

Before using this skill, understand what runs and where data flows:

• Local install required: the skill calls the locally installed zyka binary. No remote code is fetched at runtime. Install pulls the latest published version from npm ( https://www.npmjs.com/package/zyka); inspect the package source at https://github.com/kshitijdixit/zyka-sdk before installing if you want stronger assurance.
• Outbound data flow: any local files you reference (images, audio, video) are uploaded to Zyka's API. Zyka proxies generation requests to third-party model providers (OpenAI, Google, ByteDance, ElevenLabs, Kling, etc.). Do not pass private personal data, proprietary content, credentials, or files you don't want third parties to see. Review https://zyka.ai privacy/terms for retention and downstream-provider policies.
• API key handling: treat ZYKA_API_KEY as a service secret. Use a least-privilege key, never commit it, and rotate or delete it when no longer needed. Monitor usage in your Zyka dashboard.
• Autonomous use: this skill is designed for agent-driven invocation. The agent can call Zyka endpoints (and spend credits) without an extra prompt once the skill is enabled. If you want explicit user confirmation per call, run commands interactively rather than relying on autonomous invocation.

CRITICAL: Always use CLI commands, never write code scripts#

When a user asks to generate media, run the zyka generate command directly — do NOT write JavaScript files. The CLI handles file uploads, waiting, and downloading.

---

Generate Videos#

bash

zyka generate video -m MODEL -p "prompt" [options]

Video Models#

Model-m valueKey optionsOpenAI Sorasora``-s sora-2, -d 4/8/12, --size 1280x720Google Veoveo``-s veo-3.1-generate-001, -d 4-8, -a 16:9/9:16, --size 720p/1080p/4kKling AIkling``-s kling-v2-master, -d 5/10, -a 16:9, --mode std/proKling V3kling``-s kling-v3 or kling-v3-pro, -d 3-15, --first-frame ./img.jpgKling O3kling``-s kling-o3 or kling-o3-pro, -d 3-15Kling Omnikling``-s kling-video-o1, -d 3-10Kling Multi-Imagekling``-s multi-image-to-video (pass image_list via SDK)ByteDance Seedance V1.5 Probytedance``-s "Seedance V1.5 Pro", -d 4-12, --resolution 720pByteDance Seedance 2.0bytedance``-s "Seedance 2.0", -d 4-12, --resolution 720pByteDance Seedance 2.0 Fastbytedance``-s "Seedance 2.0 Fast", -d 4-12, --resolution 720pByteDance OmniHumanbytedance``-s "OmniHuman", --image ./face.jpg --audio ./speech.mp3ByteDance OmniHuman v1.5bytedance``-s "OmniHuman v1.5", --image ./face.jpg --audio ./speech.mp3Alibaba WAN T2Vwan``-s wan-2-6-t2v, -d 5/10/15, --size 1280*720Alibaba WAN 2.7wan``-s wan-2-7, -d 5/10/15, --size 1280*720Alibaba WAN I2Vwan``-s wan-2-6-i2v, --image ./img.jpg, -d 5/10/15Alibaba WAN I2V 2.5wan``-s wan-2-5-i2v, --image ./img.jpg, -d 5/10/15WAN Animate Replacewan``-s wan-v2-2-animate-replace, --video ./vid.mp4 --image ./char.pngWAN Animate Movewan``-s wan-v2-2-animate-move, --video ./vid.mp4 --image ./char.pngTalking Headinfinite_talk``--image ./face.jpg --audio ./speech.mp3Aurora (Lip Sync)aurora``--video ./face.mp4 --audio ./speech.mp3, --audio-guidance-scale 7LTX Video T2Vltx``-s ltx-2.3-text-to-videoLTX Video I2Vltx``-s ltx-2.3-image-to-video, --image ./img.jpgGrok Videogrok``-s grok-imagine-video, -d 1-15, --resolution 720p

Video Examples#

Text to video:

bash

zyka generate video -m wan -p "A cinematic sunset over mountains" -d 5 -o ./sunset.mp4

Image to video (animate a photo):

bash

zyka generate video -m kling -s kling-v2-master -p "gentle zoom in with wind" --image ./photo.jpg -d 5 -a 16:9 --mode pro -o ./animated.mp4

Talking head (lip sync):

bash

zyka generate video -m infinite_talk -p "lip sync" --image ./face.jpg --audio ./speech.mp3 -o ./talking.mp4

Aurora lip sync (video + audio):

bash

zyka generate video -m aurora --video ./face.mp4 --audio ./speech.mp3 -o ./lipsync.mp4

First/last frame interpolation:

bash

zyka generate video -m veo -s veo-3.1-generate-001 -p "smooth transition" --first-frame ./start.jpg --last-frame ./end.jpg -d 8 -a 16:9 -o ./transition.mp4

LTX video:

bash

zyka generate video -m ltx -s ltx-2.3-text-to-video -p "A flowing river through autumn forest" -o ./ltx.mp4

Grok video:

bash

zyka generate video -m grok -p "Medieval knight in mystical forest" -d 6 --resolution 720p -o ./knight.mp4

WAN animate replace (swap character in video):

bash

zyka generate video -m wan -s wan-v2-2-animate-replace --video ./original.mp4 --image ./new-character.png -o ./swapped.mp4

---

Generate Images#

bash

zyka generate image -m MODEL -p "prompt" [options]

Image Models#

Model-m valueNotesNano Banananano_banana``-s nano-banana-1 (default), nano-banana-pro (4K), nano-banana-2 (fast 4K)DALL·E 2dall_e_2``--size 256x256/512x512/1024x1024DALL·E 3dall_e_3``--quality standard/hd, --style vivid/naturalGPT Image 1gpt_image_1``--background transparent, --quality auto/low/medium/highGPT Image 1 Minigpt_image_1_miniCheaper variantGPT Image 1.5gpt_image_1_5Latest OpenAIFlux Schnellflux_1_schnellFastFlux 2 Devflux_2_devHigh qualityFlux 2 Klein 9Bflux_2_klein_9bCompact high qualityKling Imagekling``-s kling-v1, kling-v2, kling-image-v3, kling-image-v3-text-to-image, omni-image, kling-image-o1, multi-image-to-imageSD XLstable_diffusion_xl_base_1_0``--negative-prompt "blurry"SD img2imgstable_diffusion_v1_5_img2imgNeeds --imageLucid Originlucid_originLeonardo AIPhoenix 1.0phoenix_1_0Leonardo AIZyka Helionzyka_helionFast Zyka-nativeGrok Imaginegrok_imaginexAI GrokQwen Image 2 Proqwen_image_2_proChinese/English supportZ Image Turboz_image_turboFast

Image Examples#

Generate from text:

bash

zyka generate image -m gpt_image_1 -p "A neon cyberpunk cityscape" -o ./city.png

Edit an existing image:

bash

zyka generate image -m nano_banana -s nano-banana-pro -p "make the hair straight" --image ./me.png -o ./result.png

4K high-res image:

bash

zyka generate image -m nano_banana -s nano-banana-2 -p "cinematic portrait" --resolution 4K --size 5504x3072 -o ./portrait.png

Transparent background:

bash

zyka generate image -m gpt_image_1 -p "product photo of sneakers" --background transparent -o ./sneakers.png

Grok image:

bash

zyka generate image -m grok_imagine -p "Abstract golden particles, data visualization style" -o ./abstract.png

Qwen image:

bash

zyka generate image -m qwen_image_2_pro -p "A serene mountain landscape at sunset" -o ./landscape.png

Zyka Helion (fast):

bash

zyka generate image -m zyka_helion -p "cyberpunk cat in neon city" -o ./cat.png

---

Generate Text-to-Speech#

bash

zyka generate tts --script "text" [options]

TTS Providers#

Provider--provider valueNotesElevenLabselevenlabsNeeds --voice-idQwen3qwen3Voice design/clone/customChatterboxchatterboxClone + emotion tags [happy], [sad], [angry], [calm]VoxCPMvoxcpmVoice cloning only (requires reference audio)MiniMaxminimax17 preset voices (Wise_Woman, Friendly_Person, Deep_Voice_Man, etc.)MOSS-TTSmoss-ttsRunPod-basedFish Audiofish-audioInstant voice cloning

TTS Examples#

Generate speech:

bash

zyka generate tts --provider elevenlabs --voice-id VOICE_ID --script "Welcome to Zyka" -o ./speech.mp3

Clone a voice:

bash

zyka generate tts --provider chatterbox --voice ./my-voice.mp3 --script "[happy] This sounds like me!" -o ./cloned.mp3

Fish Audio:

bash

zyka generate tts --provider fish-audio --voice ./sample.wav --script "Hello world" -o ./fish.mp3

MiniMax preset voice:

bash

zyka generate tts --provider minimax --script "Hello! Welcome to MiniMax TTS." -o ./minimax.mp3

MiniMax with emotion and stereo:

bash

zyka generate tts --provider minimax --script "Hello!" --emotion happy --channel 2 -o ./minimax.mp3

---

AI Apps (via SDK)#

These apps are available through the ZykaClient SDK:

js

const { ZykaClient } = require('zyka-sdk');
const client = new ZykaClient();

Image Apps#

AppMethodParamsUpscalecreateUpscale()``{ image, resolution: '1k'/'2k'/'4k' }Face SwapcreateFaceSwap()``{ type: 'image'/'video', url, face_image }Virtual Try-OncreateVirtualTryOn()``{ human_image, cloth_image }Outfit SwapcreateOutfitSwap()``{ user_image, character_image }Skin EnhancercreateSkinEnhancer()``{ image, type: 'perfect_skin'/'realistic_skin'/'imperfect_skin' }Behind the ScenecreateBehindTheScene()``{ image, type: 'image'/'video' }Camera AnglescreateAngles()``{ image, angle: { azimuth, elevation } }9 ShortscreateNineShorts()``{ image }ZoomscreateZooms()``{ image }Story GeneratorcreateStoryGenerator()``{ image }Holi SpecialcreateHoliSpecial()``{ image }Simple AppcreateSimpleApp()``{ image, app_id?, prompt? }

Video Apps#

AppMethodParamsCaption GeneratorcreateCaptionGenerator()``{ url, language?, caption_style? }Video to ScriptcreateVideoToScript()``{ url, script_style?: 'general'/'screenplay'/'blog_post'/'social_media'/'documentary' }Video CleanercreateVideoCleaner()``{ url, language? }Video UpscalercreateVideoUpscaler()``{ video_url, target_resolution: '1080p'/'2k'/'4k', target_fps: '30fps'/'60fps' }Video Dubbing (HeyGen)createVideoDubbing()``{ video_url, model: 'heygen', output_language: 'Hindi (India)', mode?, enable_caption?, enable_speech_enhancement?, translate_audio_only? }Video Dubbing (ElevenLabs)createVideoDubbing()``{ video_url, model: 'elevenlabs', output_language: 'hi', source_lang?, num_speakers?, highest_resolution?, drop_background_audio?, use_profanity_filter? }Video Dubbing (Sarvam)createVideoDubbing()``{ video_url, model: 'sarvam', output_language: 'Hindi' (or comma-separated), source_lang?, num_speakers?, genre?: 'general'/'news'/'entertainment'/'education'/'sports'/'religious' }Dubbing LanguagesgetVideoDubbingLanguages(model)model: 'heygen'/ 'elevenlabs'/ 'sarvam' — returns supported languagesShort Video CreatorcreateShortVideoCreator()``{ url, clip_duration_sec: 'auto'/5/15/30/45 }B-rollcreateBroll()``{ url, broll_duration_sec?: 'auto'/2-10 }YouTube DownloadercreateYouTubeDownloader()``{ url, quality?: '720p', format?: 'mp4' }Voice ChangercreateVoiceChanger()``{ source_audio_url, target_voice_url?, voice_strength? }Image to SVGcreateImageToSvg()``{ image_url }

CLI App Commands#

bash

zyka generate upscale --image ./photo.jpg --resolution 4k -o ./upscaled.jpg
zyka generate face-swap --type image --url ./target.jpg --face ./face.jpg -o ./result.jpg
zyka generate skin-enhancer --image ./photo.jpg --type perfect_skin -o ./enhanced.jpg
zyka generate virtual-try-on --human ./me.jpg --cloth ./dress.jpg -o ./tryon.jpg
zyka generate outfit-swap --user-image ./me.jpg --character-image ./celeb.jpg -o ./outfit.jpg
zyka generate behind-the-scene --image ./photo.jpg --type image -o ./scene.jpg
zyka generate nine-shorts --image ./photo.jpg -o ./angles.jpg
zyka generate zooms --image ./photo.jpg -o ./zooms.jpg
zyka generate story-generator --image ./photo.jpg -o ./story.jpg
zyka generate holi-special --image ./photo.jpg -o ./holi.jpg
zyka generate simple-app --image ./photo.jpg --app-id my-app -o ./result.jpg
zyka generate caption --url ./video.mp4 --language en -o ./captioned.mp4
zyka generate video-to-script --url ./video.mp4 --script-style screenplay -o ./script.txt
zyka generate video-cleaner --url ./video.mp4 -o ./cleaned.mp4
zyka generate video-upscaler --url ./video.mp4 --resolution 4k -o ./upscaled.mp4
zyka generate video-dubbing --url ./video.mp4 --model heygen --language "Hindi (India)" --mode precision -o ./dubbed.mp4
zyka generate video-dubbing --url ./video.mp4 --model elevenlabs --language hi --source-lang en --num-speakers 2 --highest-resolution -o ./dubbed.mp4
zyka generate video-dubbing --url ./video.mp4 --model sarvam --language "Hindi,Tamil" --genre education -o ./dubbed.mp4
zyka generate short-video --url ./long.mp4 --duration auto -o ./clips/
zyka generate broll --url ./video.mp4 -o ./with-broll.mp4
zyka generate youtube-download --url "https://youtube.com/watch?v=..." --quality 720p -o ./video.mp4
zyka generate voice-changer --audio ./input.mp3 --voice ./reference.mp3 -o ./output.mp3
zyka generate image-to-svg --image ./photo.png -o ./result.svg

---

Guidelines#

• ALWAYS use zyka generate CLI commands — never write JavaScript files
• Set ZYKA_API_KEY env var before running commands
• Use -o ./filename to save results to disk (auto-downloads)
• Use --image ./path to pass local image files (auto-uploads)
• Use --audio ./path to pass local audio files (auto-uploads)
• Use --video ./path to pass local video files (auto-uploads)
• All commands auto-wait for completion — no polling needed
• When editing images, default to -m nano_banana -s nano-banana-pro
• When generating videos from text, default to -m wan
• When animating images to video, default to -m kling -s kling-v2-master
• When generating 4K images, use -m nano_banana -s nano-banana-2 --resolution 4K
• For fast image generation, use -m zyka_helion or -m flux_1_schnell
• For transparent backgrounds, use -m gpt_image_1 --background transparent

CLI Options Reference#

FlagDescription-m, --modelModel name (required for video/image)-p, --promptText prompt (required)-s, --sub-modelModel variant-d, --durationVideo duration in seconds-a, --aspect-ratioAspect ratio (16:9, 9:16, 1:1)--sizeOutput size (e.g. 1024x1024, 720p)--imageInput image path (for editing/animating)--audioInput audio path (for talking heads)--videoInput video path (for V2V)-o, --outputSave result to this file path--no-waitDon't wait for completion--negative-promptWhat to avoid in generation--modeKling mode: std or pro--resolutionResolution: 480p, 720p, 1080p, 1K, 2K, 4K--first-frameFirst frame image (Kling, Veo 3.1, Bytedance)--last-frameLast frame image (Kling, Veo 3.1, Bytedance)--qualityQuality: standard, hd, auto, low, medium, high--backgroundGPT Image background: transparent, opaque, auto--styleDALL-E 3 style: vivid, natural-n, --countNumber of images to generate--titleTitle for the generation job--audio-guidance-scaleAurora lip sync guidance (0-10)--emotionMiniMax TTS emotion (neutral/happy/sad/angry/fearful/disgusted/surprised)--channelMiniMax TTS audio channel (1=mono, 2=stereo)