Model directory
Every model,
one credit balance
Prices are pinned to real inference cost — 1 credit = $0.002 of compute, for every model below. Pick one deliberately, or leave it on Auto and the assistant routes each prompt to the model whose strengths fit it.
Image
16 modelsFrom 2-credit drafts to 4K finals. Editing models accept reference images — drag any asset from a board into the prompt.
Flux Schnell
Black Forest Labs
Fastest and cheapest — quick drafts and rapid iteration at lower fidelity.
Flux 2 Turbo
Black Forest Labs
Fast Flux 2 — the best quality-per-credit for everyday work.
Flux Dev
Black Forest Labs
Balanced quality vs speed; dependable general-purpose photorealism.
Flux Krea
Black Forest Labs
Opinionated, aesthetic look — stylized and artistic imagery.
Flux 2 Pro
Black Forest Labs
Premium Flux 2 — high-fidelity final images with strong prompt coherence.
Flux Pro 1.1
Black Forest Labs
Premium photorealism with high detail.
Flux Pro 1.1 Ultra
Black Forest Labs
2K high-resolution photoreal for large, detailed finals; takes a style reference.
Flux 2 Max
Black Forest Labs
State-of-the-art image quality — the best-looking (and priciest) Flux.
Nano Banana
Excellent instruction-following and image editing; blends up to 10 references.
Nano Banana 2
Latest Google image model — top instruction-following, text, and edits.
Nano Banana Pro
High quality up to 4K with strong text rendering plus editing.
GPT Image Mini
OpenAI
Fast, cheap GPT Image — drafts that still handle text well.
GPT Image 1.5
OpenAI
Strong text rendering and instruction-following.
GPT Image 2
OpenAI
Best-in-class text and typography — logos, infographics, precise prompt adherence.
Recraft V3
Recraft
Vector art, logos, icons, and brand assets with precise style control.
Ideogram Character
Ideogram
Consistent characters across a set of images from a single reference.
Prices are per image at standard (~1MP) resolution; larger renders scale.
Video
6 modelsText-to-video and image-to-video. The strongest workflow: generate a still you love, then animate it — art direction on the cheap medium, video credits only on motion.
Kling 2.5 Turbo
Kling
Cinematic, smooth realistic motion — the workhorse for exploring shots.
Hailuo 02 Pro
MiniMax
1080p with strong motion realism, particularly human movement.
Kling O3
Kling
Precise control — accepts a start and end frame, ideal for storyboard transitions.
Seedance 2.0
ByteDance
Native audio with realistic physics — dynamic action, sports, crowds.
Happy Horse
Alibaba
Native audio and flexible clip lengths.
Veo 3.1
Highest quality with native synchronized audio, up to 4K — the premium final.
Video bills per second of output; a typical clip is 5–8 seconds.
Audio
9 modelsSpeech, music, sound effects — and models that watch your video and score it with synced foley.
ElevenLabs Multilingual v2
ElevenLabs
Highest-quality, natural speech in 29 languages — voiceovers and narration.
ElevenLabs Turbo v2.5
ElevenLabs
Fast, low-latency speech; slightly less fidelity than Multilingual.
ElevenLabs Music
ElevenLabs
Full tracks with structure from a text brief — beds, themes, and jingles.
ElevenLabs SFX
ElevenLabs
One-shot sound effects from a description.
Stable Audio 2.5
Stability AI
High-quality music and ambience up to several minutes.
Stable Audio 3 SFX
Stability AI
Cheap, fast sound effects.
ControlFoley
fal
Watches your video and generates synced foley for it.
ThinkSound
fal
Video-to-audio scoring on a budget.
Kling Video-to-Audio
Kling
A quick soundtrack for any silent clip.
Speech bills per 1,000 characters; music per minute; sound design per second.
Utilities
6 modelsThe finishing pass: upscale drafts into deliverables and cut subjects out cleanly.
Topaz Upscale
Topaz
Crisp production upscales at the best price per megapixel.
Clarity Upscaler
fal
Creative upscaling that re-imagines detail as it enlarges.
Recraft Crisp
Recraft
Cheap, clean 2–4× upscales for flat design and vector-ish art.
ESRGAN
open source
Classic fast upscaler; great for batches.
Bria Background Remove
Bria
Licensed-data background removal, safe for commercial work.
BiRefNet v2
open source
High-accuracy masks for tricky edges — hair, fur, glass.
Upscalers bill on output size — a 4× upscale is 16× the pixels.
Not sure which to pick?
Don’t. Describe what you want and Auto routes it — drafts to fast models, finals to premium ones. Every job shows its price first.