# Model Catalog 1 credit = 1 cent. $1 = 100 credits. ## Video Models Use with `varg.videoModel("id")` or `fal.videoModel("id")`. | Model ID | Credits | Duration | Notes | |----------|---------|----------|-------| | `kling-v3` | 150 | 3-15s (integer) | **Best quality (default)**. O3 Pro tier. | | `kling-v3-standard` | 100 | 3-15s (integer) | O3 Standard tier. Good quality, cheaper. | | `kling-v2.6` | 150 | 3-15s (integer) | Native audio support: `providerOptions: { varg: { generate_audio: true } }` | | `kling-v2.5` | 100 | **5 or 10 ONLY** | Legacy. Any other duration causes 422 error. | | `kling-v2.1` | 100 | 5 or 10 | Legacy. | | `kling-v2` | 100 | 5 or 10 | Legacy. | | `wan-2.5` | 80 | varies | Fast and affordable. | | `wan-2.5-preview` | 60 | varies | Preview version. | | `minimax` | 80 | varies | Alternative provider. | | `ltx-2-19b-distilled` | 50 | varies | **Cheapest**. Uses `num_frames` not `duration`, `video_size` not `aspect_ratio`. Native audio support. | | `grok-imagine` | 100 | varies | xAI model. Native audio support. | ### Video Editing / Motion Control | Model ID | Credits | Notes | |----------|---------|-------| | `grok-imagine-edit` | 100 | Video editing via xAI. | | `kling-v2.6-motion` | 150 | Motion control (camera trajectories). | | `kling-v2.6-motion-standard` | 100 | Motion control, standard tier. | ### Lipsync Models | Model ID | Credits | Notes | |----------|---------|-------| | `sync-v2` | 50 | Standard lipsync. | | `sync-v2-pro` | 80 | Higher quality lipsync. | | `lipsync` | 50 | Basic lipsync. | | `omnihuman-v1.5` | varies | Advanced human motion. | | `veed-fabric-1.0` | varies | VEED fabric lipsync. | ### Lipsync Model Selection Guide | Model | Pipeline | Input | Speed | Quality | Best For | |-------|----------|-------|-------|---------|----------| | `veed-fabric-1.0` | Image + audio -> video | Still image + audio | Fast (~30-50s) | Good | Speech-first workflows, narrator clips | | `sync-v2-pro` | Video + audio -> video | Pre-animated video + audio | Medium (~60-90s) | Best | High-quality talking heads, facial detail | | `sync-v2` | Video + audio -> video | Pre-animated video + audio | Medium | Good | Budget alternative to sync-v2-pro | | `omnihuman-v1.5` | Image + audio -> video | Still image + audio | Slow | Variable | Full-body motion, experimental | **Decision matrix:** - **"I have a speech audio and a character image"** -> `veed-fabric-1.0` (simplest, fastest) - **"I have an animated video and want to add lip movement"** -> `sync-v2-pro` (best quality) - **"I need full-body gestures matching speech"** -> `omnihuman-v1.5` (experimental) **VEED Fabric workflow** (speech-first): ```tsx const portrait = Image({ model: varg.imageModel("nano-banana-pro"), prompt: "..." }); const talking = Video({ model: varg.videoModel("veed-fabric-1.0"), keepAudio: true, prompt: { images: [portrait], audio: speechSegment }, providerOptions: { varg: { resolution: "720p" } }, // 480p or 720p only }); ``` **sync-v2-pro workflow** (animate-then-lipsync): ```tsx const portrait = Image({ ... }); const animated = Video({ model: varg.videoModel("kling-v3"), prompt: { images: [portrait], text: "person talking" }, duration: 5 }); const talking = Video({ model: varg.videoModel("sync-v2-pro"), keepAudio: true, prompt: { images: [animated], audio: speechSegment }, }); ``` ### Video Prompt Format **Text-to-video** (string prompt): ```tsx Video({ model: varg.videoModel("kling-v3"), prompt: "a cat playing piano", duration: 5 }) ``` **Image-to-video** (object prompt with ONE image): ```tsx Video({ model: varg.videoModel("kling-v3"), prompt: { text: "cat starts playing keys", images: [catImage] }, duration: 5 }) ``` **Lipsync** (video + audio): ```tsx Video({ model: varg.videoModel("sync-v2-pro"), prompt: { video: animatedCharacter, audio: voiceover } }) ``` ### Common Provider Options (video) ```tsx Video({ model: varg.videoModel("kling-v2.6"), prompt: "...", duration: 5, aspectRatio: "9:16", providerOptions: { varg: { generate_audio: true, // Native audio (kling-v2.6, ltx, grok) resolution: "2K", // Higher resolution } } }) ``` --- ## Image Models Use with `varg.imageModel("id")` or `fal.imageModel("id")`. | Model ID | Credits | Prompt Format | Notes | |----------|---------|---------------|-------| | `nano-banana-pro` | 5 | `string` | **Best default**. Text-to-image. | | `nano-banana-pro/edit` | 5 | `{ text, images }` | Reference-based editing. Always pass reference via `images: [ref]`. | | `nano-banana-2` | 5 | `{ text, images? }` | Newer model (slower). | | `nano-banana-2/edit` | 5 | `{ text, images }` | Explicit edit mode. | | `flux-schnell` | 5 | `string` | Fast text-to-image. | | `flux-dev` | 8 | `string` | Better quality Flux. | | `flux-pro` | 10 | `string` | Best Flux quality. | | `recraft-v3` | 10 | `string` | Stylized / design images. | | `recraft-v4-pro` | 10 | `string` | Latest Recraft. | | `seedream-v4.5/edit` | 5 | `{ text, images }` | ByteDance image editing. | | `qwen-angles` | 8 | `{ text, images }` | Multi-angle generation from reference. | | `qwen-image-2` | varies | `string` | Qwen image generation. | | `qwen-image-2-pro` | varies | `string` | Qwen pro tier. | | `soul` | 15 | `string` | Higgsfield character generation. 80+ style presets. | ### Image Prompt Examples **Text-to-image** (nano-banana-pro, flux): ```tsx Image({ model: varg.imageModel("nano-banana-pro"), prompt: "a sunset over the ocean", aspectRatio: "16:9" }) ``` **Reference editing** (nano-banana-pro/edit): ```tsx Image({ model: varg.imageModel("nano-banana-pro/edit"), prompt: { text: "same person in a tropical beach setting", images: [referenceImage] }, aspectRatio: "9:16" }) ``` ### Aspect Ratios (images and videos) | Ratio | Pixels | Use Case | |-------|--------|----------| | `16:9` | 1920 x 1080 | YouTube, landscape video | | `9:16` | 1080 x 1920 | TikTok, Reels, Shorts | | `1:1` | 1080 x 1080 | Instagram feed, square | | `4:3` | 1440 x 1080 | Classic TV, presentations | | `3:4` | 1080 x 1440 | Portrait photos | | `4:5` | 1080 x 1350 | Instagram portrait (recommended for feed) | Most models support all standard ratios. --- ## Speech Models (ElevenLabs) Use with `varg.speechModel("id")` or `elevenlabs.speechModel("id")`. | Model ID | Credits | Notes | |----------|---------|-------| | `turbo` | 20 | Alias for `eleven_turbo_v2`. Fast, recommended. | | `eleven_turbo_v2` | 20 | Fast English TTS. | | `eleven_turbo_v2_5` | 20 | Updated turbo. | | `eleven_flash_v2` | 20 | Ultra-fast. | | `eleven_flash_v2_5` | 20 | Updated flash. | | `eleven_multilingual_v2` | 25 | Multi-language support. | | `eleven_v3` | 25 | Latest, highest quality. | ### Available Voices `adam`, `alice`, `aria`, `bill`, `brian`, `callum`, `charlie`, `charlotte`, `chris`, `daniel`, `eric`, `george`, `jessica`, `laura`, `liam`, `lily`, `matilda`, `rachel`, `river`, `roger`, `sarah`, `will` Classic voices: `rachel`, `domi`, `sarah`, `antoni`, `elli`, `josh`, `arnold`, `adam`, `sam` ### Speech Usage ```tsx const voice = Speech({ model: varg.speechModel("eleven_v3"), voice: "rachel", children: "Welcome to our product showcase." }) ``` --- ## Music Model (ElevenLabs) Use with `varg.musicModel("music_v1")` or `elevenlabs.musicModel()`. | Model ID | Credits | Notes | |----------|---------|-------| | `music_v1` | 30 | AI music generation. Always set `duration`. | ### Music Usage ```tsx ``` **Important**: Always set `duration` on Music. Without it, ElevenLabs generates ~60s which extends the video. --- ## Transcription Models | Model ID | Credits | Notes | |----------|---------|-------| | `whisper` | 10 | OpenAI Whisper via fal. | | `whisper-large-v3` | 10 | Whisper large model. | --- ## Quick Reference: Recommended Defaults | Task | Model | Credits | |------|-------|---------| | Image (default) | `nano-banana-pro` | 5 | | Image editing | `nano-banana-pro/edit` | 5 | | Image (fast) | `flux-schnell` | 5 | | Video (default) | `kling-v3` | 150 | | Video (budget) | `kling-v3-standard` | 100 | | Video (cheapest) | `ltx-2-19b-distilled` | 50 | | Speech | `eleven_v3` or `turbo` | 20-25 | | Music | `music_v1` | 30 | | Lipsync | `sync-v2-pro` | 80 |