---
name: zyka-ai
version: 0.4.14
description: Generate AI videos, images, voice, and use AI apps using the Zyka CLI. Use when users want to create AI-generated media — videos (Sora, Veo, Kling, WAN, Seedance, Grok, LTX, Aurora, MiniMax), images (DALL·E, GPT Image, Flux, Nano Banana, Kling, Grok, Zyka Helion, Qwen, Seedream, MAI), TTS (ElevenLabs, Chatterbox, Qwen3, MiniMax, Fish Audio), talking heads, or AI apps (upscale, face swap, captions, dubbing, etc.).
homepage: https://zyka.ai
repository: https://github.com/kshitijdixit/zyka-sdk
license: MIT
metadata:
  openclaw:
    emoji: "🎬"
    requires:
      env: [ZYKA_API_KEY]
      bins: [zyka]
    primaryEnv: ZYKA_API_KEY
    install:
      - id: npm
        kind: node
        package: zyka
        bins: [zyka]
        label: "Install Zyka CLI (npm)"
---

<!-- zyka-skill-version: 0.4.14 -->

# 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` value | Key options |
|---|---|---|
| OpenAI Sora | `sora` | `-s sora-2`, `-d 4/8/12`, `--size 1280x720` |
| Google Veo | `veo` | `-s veo-3.1-generate-001`, `-d 4-8`, `-a 16:9/9:16`, `--size 720p/1080p/4k` |
| Kling AI | `kling` | `-s kling-v2-master`, `-d 5/10`, `-a 16:9`, `--mode std/pro` |
| Kling V3 | `kling` | `-s kling-v3` or `kling-v3-pro`, `-d 3-15`, `--first-frame ./img.jpg` |
| Kling O3 | `kling` | `-s kling-o3` or `kling-o3-pro`, `-d 3-15` |
| Kling Omni | `kling` | `-s kling-video-o1`, `-d 3-10` |
| Kling Multi-Image | `kling` | `-s multi-image-to-video` (pass image_list via SDK) |
| ByteDance Seedance V1.5 Pro | `bytedance` | `-s "Seedance V1.5 Pro"`, `-d 4-12`, `--resolution 720p` |
| ByteDance Seedance 2.0 | `bytedance` | `-s "Seedance 2.0"`, `-d 4-12`, `--resolution 720p` |
| ByteDance Seedance 2.0 Fast | `bytedance` | `-s "Seedance 2.0 Fast"`, `-d 4-12`, `--resolution 720p` |
| ByteDance OmniHuman | `bytedance` | `-s "OmniHuman"`, `--image ./face.jpg --audio ./speech.mp3` |
| ByteDance OmniHuman v1.5 | `bytedance` | `-s "OmniHuman v1.5"`, `--image ./face.jpg --audio ./speech.mp3` |
| Alibaba WAN T2V | `wan` | `-s wan-2-6-t2v`, `-d 5/10/15`, `--size 1280*720` |
| Alibaba WAN 3.0 | `wan` | `-s wan-3-0`, `-d 2-30`, `--resolution 480p/720p/1080p`, `--image ./start.jpg --end-image ./end.jpg` (I2V) or `--reference-images/--reference-videos/--reference-audios` (R2V); native audio on by default (`--no-generate-audio` to disable) |
| Alibaba WAN 2.7 | `wan` | `-s wan-2-7`, `-d 5/10/15`, `--size 1280*720` |
| Alibaba WAN I2V | `wan` | `-s wan-2-6-i2v`, `--image ./img.jpg`, `-d 5/10/15` |
| Alibaba WAN I2V 2.5 | `wan` | `-s wan-2-5-i2v`, `--image ./img.jpg`, `-d 5/10/15` |
| WAN Animate Replace | `wan` | `-s wan-v2-2-animate-replace`, `--video ./vid.mp4 --image ./char.png` |
| WAN Animate Move | `wan` | `-s wan-v2-2-animate-move`, `--video ./vid.mp4 --image ./char.png` |
| Talking Head | `infinite_talk` | `--image ./face.jpg --audio ./speech.mp3` |
| Aurora (Lip Sync) | `aurora` | `--video ./face.mp4 --audio ./speech.mp3`, `--audio-guidance-scale 7` |
| LTX Video T2V | `ltx` | `-s ltx-2.3-text-to-video` |
| LTX Video I2V | `ltx` | `-s ltx-2.3-image-to-video`, `--image ./img.jpg` |
| Grok Video | `grok` | `-s grok-imagine-video`, `-d 1-15`, `--resolution 720p` |
| MiniMax H3 | `minimax` | `-s minimax-h3` (default), `-d 5-15`, `--resolution 480P/768P/2K/4K`, `--image ./start.jpg --end-image ./end.jpg`, `--prompt-expansion-mode fast/balanced/quality` |
| MiniMax H3 Max | `minimax` | `-s minimax-h3-max`, `-d 5-15`, `--resolution 480P/768P`, `-a adaptive`, `--reference-images ./a.png ./b.png --reference-videos ./v.mp4 --reference-audios ./a.mp3` (R2V) |

### 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
```

**MiniMax H3 (first + last frame, 2K):**
```bash
zyka generate video -m minimax -s minimax-h3 -p "Slow dolly-in on the lighthouse as waves crash, golden hour" --image ./lighthouse.jpg --end-image ./lighthouse-dusk.jpg -d 8 --resolution 2K -a 16:9 -o ./lighthouse.mp4
```

**MiniMax H3 Max reference-to-video:**
```bash
zyka generate video -m minimax -s minimax-h3-max -p "The character from the reference walks through a busy Tokyo crossing at night" --reference-images ./char-front.png ./char-side.png --reference-videos ./walk-cycle.mp4 -d 10 --resolution 768P -a adaptive -o ./tokyo.mp4
```

**WAN 3.0 image-to-video with native audio:**
```bash
zyka generate video -m wan -s wan-3-0 -p "Camera orbits the sports car as it idles, engine rumble" --image ./car.jpg -d 12 --resolution 1080p -a 16:9 -o ./car.mp4
```

**WAN 3.0 reference-to-video:**
```bash
zyka generate video -m wan -s wan-3-0 -p "Recreate the dance in the reference video with the person from the image" --reference-images ./person.png --reference-videos ./dance.mp4 -d 6 --resolution 720p -o ./dance.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` value | Notes |
|---|---|---|
| Nano Banana | `nano_banana` | `-s nano-banana-1` (default), `nano-banana-pro` (4K), `nano-banana-2` (fast 4K) |
| DALL·E 2 | `dall_e_2` | `--size 256x256/512x512/1024x1024` |
| DALL·E 3 | `dall_e_3` | `--quality standard/hd`, `--style vivid/natural` |
| GPT Image 1 | `gpt_image_1` | `--background transparent`, `--quality auto/low/medium/high` |
| GPT Image 1 Mini | `gpt_image_1_mini` | Cheaper variant |
| GPT Image 1.5 | `gpt_image_1_5` | Latest OpenAI |
| Flux Schnell | `flux_1_schnell` | Fast |
| Flux 2 Dev | `flux_2_dev` | High quality |
| Flux 2 Klein 9B | `flux_2_klein_9b` | Compact high quality |
| Kling Image | `kling` | `-s kling-v1`, `kling-v2`, `kling-image-v3`, `kling-image-v3-text-to-image`, `omni-image`, `kling-image-o1`, `multi-image-to-image` |
| SD XL | `stable_diffusion_xl_base_1_0` | `--negative-prompt "blurry"` |
| SD img2img | `stable_diffusion_v1_5_img2img` | Needs `--image` |
| Lucid Origin | `lucid_origin` | Leonardo AI |
| Phoenix 1.0 | `phoenix_1_0` | Leonardo AI |
| Zyka Helion | `zyka_helion` | Fast Zyka-native |
| Grok Imagine | `grok_imagine` | xAI Grok |
| Qwen Image 2 Pro | `qwen_image_2_pro` | Chinese/English support |
| Grok Imagine 2.0 | `grok_imagine` | `-s grok-imagine-image-v2`, `--resolution 1k/2k`, `--quality low/medium`, `--num-images 1-4`, `--aspect-ratio 16:9/auto`; edit via `--image-list` (1-3) |
| Qwen Image 3 | `qwen_image_3` | `--size 1024*1024/2048*2048/landscape_16_9`, `--num-images 1-6`, `--negative-prompt`; edit via `--image-list` (1-3) |
| Seedream V5 Pro | `seedream_v5_pro` | ByteDance. `--size auto_2K/2048*1536/square_hd`; edit via `--image-list` (1-10). Always 1 image |
| MAI-Image-2.5-Pro | `mai_image_2_5_pro` | Microsoft. `--aspect-ratio auto/16:9/9:16/3:2`, `--output-format png/jpeg/webp`; edit via `--image-list` (1-10). Always 1 image |
| Z Image Turbo | `z_image_turbo` | Fast |

### 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
```

**Grok Imagine 2.0 (2K, multiple images):**
```bash
zyka generate image -m grok_imagine -s grok-imagine-image-v2 -p "A neon-lit ramen shop in the rain, cinematic" --aspect-ratio 16:9 --resolution 2k --num-images 2 --output-format png -o ./ramen.png
```

**Qwen Image 3:**
```bash
zyka generate image -m qwen_image_3 -p "Product shot of a matte black wireless earbud case on marble" --negative-prompt "blurry, text, watermark" --size 2048*2048 -o ./earbuds.png
```

**Seedream V5 Pro edit (multi-reference):**
```bash
zyka generate image -m seedream_v5_pro -p "Put the subject in a red trench coat, keep the pose" --image-list ./ref1.jpg ./ref2.jpg --size auto_2K -o ./coat.png
```

**MAI-Image-2.5-Pro:**
```bash
zyka generate image -m mai_image_2_5_pro -p "Isometric illustration of a cozy coffee shop interior, pastel palette" --aspect-ratio 16:9 --output-format webp -o ./coffee.webp
```

**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` value | Notes |
|---|---|---|
| ElevenLabs | `elevenlabs` | Needs `--voice-id` |
| Qwen3 | `qwen3` | Voice design/clone/custom |
| Chatterbox | `chatterbox` | Clone + emotion tags `[happy]`, `[sad]`, `[angry]`, `[calm]` |
| VoxCPM | `voxcpm` | Voice cloning only (requires reference audio) |
| MiniMax | `minimax` | 17 preset voices (Wise_Woman, Friendly_Person, Deep_Voice_Man, etc.) |
| MOSS-TTS | `moss-tts` | RunPod-based |
| Fish Audio | `fish-audio` | Instant 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
| App | Method | Params |
|---|---|---|
| Upscale | `createUpscale()` | `{ image, resolution: '1k'/'2k'/'4k' }` |
| Face Swap | `createFaceSwap()` | `{ type: 'image'/'video', url, face_image }` |
| Virtual Try-On | `createVirtualTryOn()` | `{ human_image, cloth_image }` |
| Outfit Swap | `createOutfitSwap()` | `{ user_image, character_image }` |
| Skin Enhancer | `createSkinEnhancer()` | `{ image, type: 'perfect_skin'/'realistic_skin'/'imperfect_skin' }` |
| Behind the Scene | `createBehindTheScene()` | `{ image, type: 'image'/'video' }` |
| Camera Angles | `createAngles()` | `{ image, angle: { azimuth, elevation } }` |
| 9 Shorts | `createNineShorts()` | `{ image }` |
| Zooms | `createZooms()` | `{ image }` |
| Story Generator | `createStoryGenerator()` | `{ image }` |
| Holi Special | `createHoliSpecial()` | `{ image }` |
| Simple App | `createSimpleApp()` | `{ image, app_id?, prompt? }` |

### Video Apps
| App | Method | Params |
|---|---|---|
| Caption Generator | `createCaptionGenerator()` | `{ url, language?, caption_style? }` |
| Video to Script | `createVideoToScript()` | `{ url, script_style?: 'general'/'screenplay'/'blog_post'/'social_media'/'documentary' }` |
| Video Cleaner | `createVideoCleaner()` | `{ url, language? }` |
| Video Upscaler | `createVideoUpscaler()` | `{ 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 Languages | `getVideoDubbingLanguages(model)` | model: `'heygen'`/`'elevenlabs'`/`'sarvam'` — returns supported languages |
| Short Video Creator | `createShortVideoCreator()` | `{ url, clip_duration_sec: 'auto'/5/15/30/45 }` |
| B-roll | `createBroll()` | `{ url, broll_duration_sec?: 'auto'/2-10 }` |
| YouTube Downloader | `createYouTubeDownloader()` | `{ url, quality?: '720p', format?: 'mp4' }` |
| Voice Changer | `createVoiceChanger()` | `{ source_audio_url, target_voice_url?, voice_strength? }` |
| Image to SVG | `createImageToSvg()` | `{ image_url }` |
| Transcription | `createTranscription()` | `{ audio_url, language? }` — returns `{ transcript, confidence, duration, detected_language }` directly (sync, Deepgram) |
| List Transcriptions | `listTranscriptions()` | `{ page?, limit?, search?, from_date?, to_date? }` |
| Delete Transcription | `deleteTranscription()` | `id` |

### 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
zyka generate transcription --audio ./meeting.mp3 --language en-US -o ./transcript.txt
```

---

## 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`

### Model limit warnings

The SDK validates video params against per-model configs (supported durations, resolutions, aspect ratios, prompt limits, file-size caps, audio-duration windows) and prints `[zyka-sdk]`-prefixed warnings when something looks off — e.g. unknown sub_model, duration not in the supported list, or audio file too large. The request still goes to the server (warnings are advisory). Examples of caught issues:

- `OmniHuman v1.5` audio duration limits: 60s @ 720p, 30s @ 1080p
- `WAN` audio must be 2–30s
- `wan-2-7` only accepts duration `5`, `10`, or `15`
- `wan-3-0` duration must be 2–30s; `minimax` 5–15s
- Too many reference inputs for `minimax-h3-max` (≤9 images / ≤3 videos / ≤3 audio) or `wan-3-0` (≤10 / ≤5 / ≤5)
- File size caps (per model, e.g. Kling: 10 MB image, OmniHuman v1.5: 50 MB audio for Seedance V1.5 Pro)

For programmatic introspection from SDK code: `client.getModelConfig('bytedance', 'OmniHuman v1.5')`.

## CLI Options Reference

| Flag | Description |
|---|---|
| `-m, --model` | Model name (required for video/image) |
| `-p, --prompt` | Text prompt (required) |
| `-s, --sub-model` | Model variant |
| `-d, --duration` | Video duration in seconds |
| `-a, --aspect-ratio` | Aspect ratio (16:9, 9:16, 1:1) |
| `--size` | Output size (e.g. 1024x1024, 720p) |
| `--image` | Input image path (for editing/animating) |
| `--audio` | Input audio path (for talking heads) |
| `--video` | Input video path (for V2V) |
| `-o, --output` | Save result to this file path |
| `--no-wait` | Don't wait for completion |
| `--negative-prompt` | What to avoid in generation |
| `--mode` | Kling mode: std or pro |
| `--resolution` | Resolution: 480p, 720p, 1080p, 1K, 2K, 4K (minimax: 480P/768P/2K/4K; grok v2: 1k/2k) |
| `--end-image` | End/last frame image (minimax, wan-3-0) |
| `--reference-images/--reference-videos/--reference-audios` | Reference inputs for reference-to-video (minimax-h3-max, wan-3-0) |
| `--prompt-expansion-mode` | MiniMax prompt expansion: fast, balanced, quality |
| `--no-generate-audio` | Disable WAN 3.0 native audio |
| `--num-images` | Number of images (grok v2 1-4, qwen_image_3 1-6) |
| `--image-list` | Multiple edit references (gpt_image_2, grok v2, qwen_image_3, seedream_v5_pro, mai_image_2_5_pro) |
| `--first-frame` | First frame image (Kling, Veo 3.1, Bytedance) |
| `--last-frame` | Last frame image (Kling, Veo 3.1, Bytedance) |
| `--quality` | Quality: standard, hd, auto, low, medium, high |
| `--background` | GPT Image background: transparent, opaque, auto |
| `--style` | DALL-E 3 style: vivid, natural |
| `-n, --count` | Number of images to generate |
| `--title` | Title for the generation job |
| `--audio-guidance-scale` | Aurora lip sync guidance (0-10) |
| `--emotion` | MiniMax TTS emotion (neutral/happy/sad/angry/fearful/disgusted/surprised) |
| `--channel` | MiniMax TTS audio channel (1=mono, 2=stereo) |