07
Models and prices
What each generate model and tool costs, and how long a run can take
What you can pick from
In a generate panel you pick a mode first, and the model picker then offers the models that serve it. Credits are what one run costs at the model's default settings: ≥ marks a floor that rises with longer sources or more text, and / 1K chars and / s are rates. Times are each model's ceiling, not an average.
Image
12| Model | What it does | Credits | Up to |
|---|---|---|---|
| GPT Image 2.5 Sunburst | Text to Image Top-ranked fidelity and prompt adherence, up to 4K | 2.4 | 120s |
| Grok Imagine Image 2.0 | Text to Image A second top engine with its own house look | 5 | 120s |
| Nano Banana 2 | Text to Image Grounds images in live web search | 6.3 | 25s |
| Recraft V4.1 Pro Vector | Text to Image Editable SVG vectors, crisp at any size | 30 | 120s |
| Riverflow 2.0 Pro | Text to Image Transparent backgrounds at 4K | 15 | 120s |
| Krea V2 Large | Text to Image Follows up to 3 style images | 6 | 120s |
| Ideogram V3 Quality | Text to Image Strong lettering in images, takes 3 style images | 9 | 120s |
| Recraft V4 Style | Text to Image Draws in the look of your style images | ≥ 4.4 | 120s |
| GPT Image 2.5 Sunburst | Image to Image Highest-fidelity editing, 13 references plus 3 style images | ≥ 3.9 | 120s |
| Muse Image | Image to Image A different house look, 7 references plus 3 style images | 1.1 | 120s |
| Nano Banana Pro Ultra | Image to Image The only image model past 4K, up to 8K | 15 | 120s |
| Qwen Image Multiple Angles | Image to Image Re-shoots the subject from an exact camera angle | 2.5 | 120s |
Video
24| Model | What it does | Credits | Up to |
|---|---|---|---|
| Gemini Omni 1.1 Flash (Text-to-Video) | Text to Video · Multi-Shot Top-ranked, with synced audio, up to 4K | 80 | 600s |
| Wan 3.0 (Text-to-Video) | Text to Video · Multi-Shot Up to 30 seconds in one take, with native audio | 47.5 | 600s |
| Seedance 2.5 (Text-to-Video) | Text to Video · Multi-Shot 4K, steered by up to 50 reference assets | 162 | 600s |
| MiniMax H3 (Text-to-Video) | Text to Video · Multi-Shot Silent 2K, from 21:9 to 9:16 | 50 | 600s |
| Kling 3.0 4K | Text to Video · Multi-Shot 4K, with a negative prompt | 210 | 600s |
| Gemini Omni 1.1 Flash (Image-to-Video) | Image to Video · First-Last Frame · Multi-Shot Native audio, up to 4K, optional end frame | 80 | 600s |
| MiniMax H3 (Image-to-Video) | Image to Video · First-Last Frame · Multi-Shot Most-voted in its class, native 2K, 15 seconds | 50 | 600s |
| Wan 3.0 (Image-to-Video) | Image to Video · First-Last Frame · Multi-Shot 2 to 30 seconds, audio you can switch off | 47.5 | 600s |
| Seedance 2.5 (Image-to-Video) | Image to Video · First-Last Frame · Multi-Shot 4K from one frame, up to 30 seconds | 162 | 600s |
| Kling Video O3 4K | Image to Video · Multi-Shot Keeps the images you mention consistent through the clip, in 4K | 210 | 600s |
| Wan 2.2 Animate 2 | Image Animation Moves a character to a driving video, identity holds for 120 seconds | ≥ 12 | 600s |
| ByteDance DreamActor V2 | Image Animation Handles non-human and multi-character casts | 5 / s | 600s |
| MiniMax H3 (Reference) | Reference to Video · Multi-Shot Top-rated, 2K from mixed references | 50 | 600s |
| Gemini Omni 1.1 Flash (Reference) | Reference to Video · Multi-Shot 10 references, 4K, lip-synced audio | 80 | 600s |
| Wan 3.0 (Reference) | Reference to Video · Multi-Shot 30-second takes from up to 20 mixed assets | 47.5 | 600s |
| HappyHorse 1.1 | Reference to Video · Multi-Shot Holds the visual style, not just faces | 70 | 600s |
| FLUX 3 | First-Last Frame Lands exactly on the end frame | 85 | 600s |
| OmniHuman 1.5 | Talking Head Acts out the audio with the whole body, not just the mouth | 16 / s | 180s |
| Seedance 2.5 Talking Avatar | Talking Head A portrait that speaks your audio, up to two minutes | ≥ 176 | 180s |
| Sync Lipsync 3 | Talking Head Redubs real footage, up to 5 minutes | 13.5 / s | 600s |
| InfiniteTalk (Video-to-Video) | Talking Head Re-performs footage, head and body | ≥ 9 | 600s |
| InfiniteTalk (Multi) | Talking Head Two speakers in one shot, up to 10 minutes | ≥ 9 | 600s |
| Sync React 1 | Talking Head Pick the emotion: happy, angry, sad | 16.7 / s | 600s |
| LTX 2.3 | Talking Head 1080p, and it can invent the speaker | ≥ 9 | 600s |
Audio
15| Model | What it does | Credits | Up to |
|---|---|---|---|
| Mureka V9.5 (Song) | Text to Music · Reference to Music Top-ranked, sings the lyrics you write | 22.5 | 300s |
| Mureka V9.5 (BGM) | Text to Music Instrumental beds, no vocals | 22.5 | 300s |
| Lyria 3 Pro | Text to Music Mood set by a reference picture | 8 | 300s |
| MiniMax Music Cover | Reference to Music Keeps the melody, swaps the style | 15 | 300s |
| Sonilo Video to Music | Reference to Music Music timed to a clip's cuts | ≥ 9 | 300s |
| Mirelo SFX 1.6 | Sound Effects Foley-grade effects, seamless loops | 10 | 300s |
| Sonilo Text to SFX | Sound Effects One unbroken bed, up to 3 minutes | 1 | 10s |
| HunyuanVideo-Foley | Sound Effects 48 kHz foley locked to footage | 5 | 300s |
| Inworld Realtime TTS-2 | Text to Speech The most natural single-voice read here | 5 / 1K chars | 120s |
| Gemini 3.1 Flash TTS | Text to Speech Two named speakers in one dialogue | ≥ 1 | 120s |
| MiniMax Speech 2.8 HD | Text to Speech Custom pronunciation and named emotions | 10 / 1K chars | 120s |
| ElevenLabs Eleven v3 | Text to Speech Set steady or expressive for each take | 20 / 1K chars | 30s |
| MiniMax Voice Clone | Voice Cloning One cloned voice reused across shots | 10 / 1K chars | 120s |
| OmniVoice | Voice Cloning Clones a voice into 600+ languages | ≥ 0.5 | 120s |
| ElevenLabs Voice Changer | Voice Cloning Keeps your timing, swaps the voice | 0.4 / s | 120s |
Models behind the Tools menu
In a node's menu, under Tools, each of these entries runs one model the product picks for it, priced the same way as above. The rest of that menu runs no model: picture tools such as crop and rotate are free, and video tools such as cut and speed are charged for the compute they use.
| Tool | Model | Credits | Up to |
|---|---|---|---|
| Image · Remove background | Bria Remove Background | 1.8 | 120s |
| Image · Upscale | Pro Upscaler | 12 | 120s |
| Image · Talking head | Seedance 2.5 Talking Avatar | ≥ 176 | 180s |
| Image · Inpaint | Bria GenFill | 4 | 30s |
| Image · Erase | Bria Eraser | 4 | 15s |
| Image · Sketch edit | Seedream 4.5 Edit | 4 | 30s |
| Video · Upscale | SeedVR2 | 12 | 600s |
| Video · Interpolate | RIFE | 1 | 30s |
| Video · Extend | Seedance 2.5 | 138.6 | 600s |
| Video · Edit | Wan 3.0 | 50 | 600s |
| Video · Motion transfer | Kling 3.0 Pro | 50.4 | 180s |
| Video · Animate | Wan 2.2 Animate 2 | ≥ 12 | 600s |
| Audio · Separate vocals | Vocal Remover | 0.1 / s | 30s |
| Audio · Extend | Mirelo SFX 1.6 | 10 | 300s |
A deployment only offers the models it holds an API key for, so what you see in the product can be a subset of this list.
Chosen for you
Three places we pick the model
In the generate panels the model is yours to choose, and the Tools menu runs the models listed above. These three run on one we set, and your text reaches it the same way it reaches any other provider.
- Nano Banana 2 rewrites your description into a structured prompt before it draws
deepseek/deepseek-chat - Understand, in a node's menu, which reads the picture, video or sound on that node
google/gemini-3.8-flash - The assistant's replies, and the summaries it writes into memory
deepseek/deepseek-v4-pro