What a generation costs.
Studio is the media side of a scene: images, video, speech and meshes, with no account to connect. Every model below is one an agent can pick by name, and the rate beside it is what leaves your balance when it runs.
Text to a still image.
| Model | Maker | Price |
|---|---|---|
| FLUX.1 schnell # Black Forest Labs fastest, ~1s | Black Forest Labs | $0.003 per megapixel |
| FLUX.1 devdefault # Black Forest Labs balanced default | Black Forest Labs | $0.025 per megapixel |
| FLUX.1 pro v1.1 # Black Forest Labs production quality | Black Forest Labs | $0.04 per megapixel |
| Recraft v3 # Recraft vector and brand-consistent styles | Recraft | $0.04 per image |
| Ideogram v3 # Ideogram best at text inside the image | Ideogram | $0.05 per image |
| FLUX.1 pro ultra # Black Forest Labs highest resolution | Black Forest Labs | $0.06 per image |
| Nano Banana Pro # Google strong prompt adherence | $0.15 per image |
An image in, a changed image out.
| Model | Maker | Price |
|---|---|---|
| FLUX.1 Kontext pro # Black Forest Labs character consistency across edits | Black Forest Labs | $0.04 per image |
| Nano Banana 2 # Google cheaper, older | $0.08 per image | |
| Nano Banana Prodefault # Google best all-round editor | $0.15 per image |
Some models animate a still, others work from text alone.
| Model | Maker | Works | Price |
|---|---|---|---|
| Kling 2.1 standarddefault # Kuaishou · from an image cheapest animator | Kuaishou | from an image | $0.05 per second |
| Kling 2.1 pro # Kuaishou · from an image | Kuaishou | from an image | $0.09 per second |
| Kling 3 standard # Kuaishou · from an image | Kuaishou | from an image | $0.15 per second |
| Kling 3 pro # Kuaishou · from an image cinematic, generates its own audio | Kuaishou | from an image | $0.168 per second |
| Seedance 2.0default # ByteDance · from text real physics, generates its own audio | ByteDance | from text | $0.3034 per second |
| Seedance 2.0 # ByteDance · from an image camera control | ByteDance | from an image | $0.3034 per second |
Text to spoken audio.
| Model | Maker | Price |
|---|---|---|
| Dia # Nari Labs dialogue and nonverbals, voice cloning | Nari Labs | $0.05 per 1k characters |
| Gemini 3.1 Flash TTS # Google fast, steerable tone | $0.05 per 1k characters | |
| Index TTS 2.0 # Bilibili cheaper, good at Mandarin | Bilibili | $0.05 per 1k characters |
| ElevenLabs Turbo v2.5default # ElevenLabs most natural, 70+ languages | ElevenLabs | $0.10 per 1k characters |
Recorded audio to text.
| Model | Maker | Price |
|---|---|---|
| Nemotron ASR # NVIDIA multilingual, strong on accents | NVIDIA | $0.006 per minute |
| Wizperdefault # OpenAI (Whisper v3) general-purpose, 90+ languages | OpenAI (Whisper v3) | $0.006 per minute |
| Scribe v2 # ElevenLabs best accuracy, speaker labels | ElevenLabs | $0.01 per minute |
An image to a textured mesh.
| Model | Maker | Price |
|---|---|---|
| Tripo v2.5default # Tripo3D bake-off winner on hard-surface objects | Tripo3D | $0.30 per call |
How to read this. Media is metered several incompatible ways, so every rate carries the unit it is charged against. A cost is computed from the request rather than looked up: a five second clip at $0.05 per second is $0.25, and the same model asked for fifteen seconds is $0.75. The default in each group is what runs when nothing is named.
We take no markup. Generation is passed through at cost, the same as model usage.
Studio spends credit, not your plan allowance. Generation draws on your top-up balance rather than the monthly usage included in a plan, so it is the one part of a scene that needs credit on the account before it will run.