Models

Model Catalog

Complete catalog of all supported models on the Skytells platform with pricing, capabilities, and input schemas. Last updated March 17, 2026.

Every model on Skytells has its own namespace, pricing structure, capabilities, and input/output schema. This page is a complete reference for all currently supported models, grouped by type.

Use the namespace value when making API calls. Text models use the Inference API endpoints (/v1/chat/completions, /v1/responses). All other models use the Predictions API endpoint (/v1/predictions).

# Text models — Inference API
curl https://api.skytells.ai/v1/chat/completions \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepbrain-router", "messages": [{"role": "user", "content": "Hello!"}]}'

# Image / Video / Audio models — Predictions API
curl -X POST https://api.skytells.ai/v1/predictions \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "truefusion-pro", "input": {"prompt": "A sunset over mountains"}}'

Text Models

Text models are accessed via the Inference API, which is fully OpenAI-compatible. You can use the OpenAI SDK by pointing base_url to https://api.skytells.ai/v1.

GPT-5

PropertyValue
Namespacegpt-5
VendorOpenAI
Pricing$0.50 / 1M input tokens · $1.25 / 1M output tokens
Capabilitiestext-to-text, coding, writing, reasoning, chat, analysis, summarization, instruction-following, problem-solving
StatusOperational
OpenAI Compatible✅

OpenAI's flagship fifth-generation model, offering state-of-the-art reasoning, coding, and instruction-following capabilities.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be gpt-5
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "gpt-5",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}
Token TypePrice
Input tokens$0.50 / 1M tokens
Output tokens$1.25 / 1M tokens

Billed separately for input and output. See Pricing for details.


GPT-5.4

PropertyValue
Namespacegpt-5.4
VendorOpenAI
Pricing$0.50 / 1M input tokens · $1.25 / 1M output tokens
Capabilitiestext-to-text, coding, writing, reasoning, chat, analysis, summarization, instruction-following, problem-solving
StatusOperational
OpenAI Compatible✅

An incremental update to GPT-5 with improved accuracy, stronger instruction adherence, and refined reasoning in complex multi-turn conversations.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be gpt-5.4
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "gpt-5.4",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}
Token TypePrice
Input tokens$0.50 / 1M tokens
Output tokens$1.25 / 1M tokens

Billed separately for input and output. See Pricing for details.


GPT-5.4 Mini

PropertyValue
Namespacegpt-5.4-mini
VendorOpenAI
Pricing$0.75 / 1M input tokens · $0.075 / 1M cached input tokens · $4.50 / 1M output tokens
Capabilitiestext-to-text, language, writing, reasoning, chat, analysis, problem-solving, quality, coding, responses, fast, mini
StatusOperational
OpenAI Compatible✅
Edge Compatible✅

GPT-5.4 mini brings the strengths of GPT-5.4 to a faster, more efficient model designed for high-volume workloads.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be gpt-5.4-mini
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "gpt-5.4-mini",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}
Token TypePrice
Input tokens$0.75 / 1M tokens
Cached input tokens$0.075 / 1M tokens
Output tokens$4.50 / 1M tokens

Billed separately for input, cached input, and output. See Pricing for details.


GPT-5.3 Codex

PropertyValue
Namespacegpt-5.3-codex
VendorOpenAI
Pricing$1.75 / 1M input tokens · $0.175 / 1M cached input tokens · $14 / 1M output tokens
Capabilitiestext-to-text, coding, writing, reasoning, chat, analysis, problem-solving, quality
StatusOperational
OpenAI Compatible✅
Edge Compatible✅

GPT-5.3 Codex achieves state-of-the-art performance on SWE-Bench Pro, a rigorous evaluation of real-world software engineering. SWE-Bench Pro spans four languages and is more contamination-resistant, challenging, diverse, and industry-relevant than SWE-bench Verified (which only tests Python). It also far exceeds the previous state-of-the-art on Terminal-Bench 2.0, which measures terminal skills for coding agents. GPT-5.3 Codex does so with fewer tokens than prior models, so you can build more per dollar.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be gpt-5.3-codex
inputstring | array✅—User prompt as a string, or an array of chat-style message objects (same idea as /v1/chat/completions messages)
instructionsstring——System-level instructions (equivalent to a system message)
previous_response_idstring——Continue a prior turn without resending full history
streamboolean—falseStream the response as SSE
max_output_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling (0–1)

See the Responses API reference for the full schema.

{
  "id": "resp_...",
  "object": "response",
  "model": "gpt-5.3-codex",
  "status": "completed",
  "output_text": "..."
}

Shape follows the OpenAI Responses object; see Responses — Response object for all fields.

Token TypePrice
Input tokens$1.75 / 1M tokens
Cached input tokens$0.175 / 1M tokens
Output tokens$14 / 1M tokens

Billed separately for input, cached input, and output. See Pricing for details.


Llama 3.1 8B

PropertyValue
Namespacellama-3.1-8b
VendorMeta
Pricing$0.12 / 1M input tokens · $0.12 / 1M cached input tokens · $0.12 / 1M output tokens
Capabilitiestext-to-text, language, writing, chat, analysis, quality, coding, responses, fast
StatusOperational
OpenAI Compatible✅
Edge Compatible✅

Llama 3.1 8B brings powerful performance in a smaller, more efficient package. With improved multilingual support, tool use, and a 128K context length, it enables sophisticated use cases like interactive agents and compact coding assistants while remaining lightweight and accessible.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be llama-3.1-8b
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "llama-3.1-8b",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}
Token TypePrice
Input tokens$0.12 / 1M tokens
Cached input tokens$0.12 / 1M tokens
Output tokens$0.12 / 1M tokens

Billed separately for input, cached input, and output. See Pricing for details.


DeepBrain Router

PropertyValue
Namespacedeepbrain-router
VendorSkytells ✓
Pricing$0.50 / 1M input tokens · $1.25 / 1M output tokens
Capabilitiestext-to-text, coding, writing, reasoning, chat, analysis, summarization, instruction-following, problem-solving, quality, fast, router
StatusOperational
OpenAI Compatible✅
Edge Compatible✅
Cold BootNo

DeepBrain Router is Skytells' advanced model orchestration layer, built to intelligently choose the right model for the right task. Optimized for coding, writing, reasoning, and complex multi-domain workloads, it dynamically routes requests across a curated set of flagship models from leading providers. The result is stronger output quality, improved cost-performance balance, and a more reliable AI experience at scale.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be deepbrain-router
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2). Lower = more deterministic
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)

Unlike traditional LLMs, Reasoning models, and other flagship models, DeepBrain Router model works differently. It dynamically selects the best underlying model for each request, so you get consistently high quality without managing model selection yourself, which advances AI experiance and Agentic workflows with less overhead and more flexibility.

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "deepbrain-router",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}

DeepBrain Router dynamically selects from the following models based on task complexity, domain, and quality requirements:

ModelVersionProvider
DeepBrain Mini—Skytells
DeepBrain 2.0—Skytells
GPT-4o2024-11-20OpenAI
GPT-4o Mini2024-07-18OpenAI
GPT-4.1 Nano2025-04-14OpenAI
GPT-4.1 Mini2025-04-14OpenAI
GPT-4.12025-04-14OpenAI
o4-mini2025-04-16OpenAI
GPT-5 Nano2025-08-07OpenAI
GPT-5 Mini2025-08-07OpenAI
GPT-5 Chat2025-08-07OpenAI
GPT-52025-08-07OpenAI
GPT-5.2 Chat2025-12-11OpenAI
GPT-5.22025-12-11OpenAI
GPT-OSS 120B—OpenAI
Llama-4 Maverick 17B-128E (FP8)—Meta
DeepSeek-V3.1—DeepSeek
DeepSeek-V3.2—DeepSeek
Grok-4—xAI
Grok-4 Fast Reasoning—xAI
Claude Haiku 4.520251001Anthropic
Claude Sonnet 4.520250929Anthropic
Claude Opus 4.120250805Anthropic
Claude Opus 4.6—Anthropic

You cannot control which model is selected. The router optimises for quality and task fit. All routed requests are billed at the DeepBrain Router rate regardless of the underlying model chosen.

Token TypePrice
Input tokens$0.50 / 1M tokens
Output tokens$1.25 / 1M tokens

Billed separately for input and output. See Pricing for details.


Kimi K2.6

PropertyValue
Namespacekimi-k2.6
VendorMoonshot AI
Pricing$0.74 / 1M input tokens · $4.66 / 1M output tokens
Capabilitiestext-to-text, language, writing, reasoning, chat, analysis, problem-solving, coding, fast
StatusOperational
OpenAI Compatible✅
Edge Compatible✅

Moonshot AI's next-generation multimodal model designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration. Handles complex end-to-end tasks across Python, Rust, and Go. Its agent swarm architecture scales to hundreds of parallel sub-agents for autonomous task decomposition — delivering documents, websites, and spreadsheets in a single run without human oversight.

ParameterTypeRequiredDefaultDescription
modelstring✅—Must be kimi-k2.6
messagesarray✅—Array of {role, content} objects. Roles: system, user, assistant
streamboolean—falseEnable server-sent events streaming
max_tokensinteger—8192Maximum tokens to generate
temperaturenumber—0.7Sampling temperature (0–2)
top_pnumber—0.95Nucleus sampling probability (0–1)
frequency_penaltynumber—0.0Penalise token frequency (-2.0–2.0)
presence_penaltynumber—0.0Penalise new topics (-2.0–2.0)
stopstring | array——Stop sequences (up to 4)
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "kimi-k2.6",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 25,
    "total_tokens": 35
  }
}
Token TypePrice
Input tokens$0.74 / 1M tokens
Cached input tokens$0.07 / 1M tokens
Output tokens$4.66 / 1M tokens

Billed separately for input, cached input, and output. See Pricing for details.


Image Models

TrueFusion

PropertyValue
Namespacetruefusion
VendorSkytells
Pricing$0.03 / image
Capabilitiestext-to-image
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for generation
aspect_ratioenum—1:11:1, 16:9, 4:3, 3:2, 2:3, 3:4, 9:16, 21:9
number_of_imagesinteger—1Number of images (1–9)
prompt_optimizerboolean—trueUse prompt optimizer
{
  "type": "array",
  "items": { "type": "string", "format": "uri" }
}

Returns an array of image URLs.


TrueFusion Pro

PropertyValue
Namespacetruefusion-pro
VendorSkytells
Pricing$0.05 / image
Capabilitiestext-to-image, image-to-image
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Prompt for generated image
aspect_ratioenum—1:11:1, 16:9, 21:9, 3:2, 2:3, 4:5, 5:4, 3:4, 4:3, 9:16, 9:21
imagestring (URI)——Input image for img2img mode
prompt_strengthnumber—0.8Prompt strength for img2img (0–1)
num_outputsinteger—1Number of outputs (1–4)
num_inference_stepsinteger—28Denoising steps (1–50). Recommended: 28–50
guidancenumber—3Guidance scale (0–10)
seedinteger——Random seed for reproducibility
output_formatenum—webpwebp, jpg, png
output_qualityinteger—80Output quality (0–100)
go_fastboolean—trueUse fp8 quantized model for speed
megapixelsenum—11, 0.25
disable_safety_checkerboolean—falseDisable safety checker
{
  "type": "array",
  "items": { "type": "string", "format": "uri" }
}

Returns an array of image URLs.


TrueFusion Max

PropertyValue
Namespacetruefusion-max
VendorSkytells
Pricing$0.12 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for image generation
aspect_ratioenum—1:111 ratios including 21:9, 9:21
image_promptstring (URI)——Image reference to guide composition
image_prompt_strengthnumber—0.1Blend between prompt and image (0–1)
safety_toleranceinteger—2Safety level (1=strict, 6=permissive)
seedinteger——Random seed
rawboolean—falseGenerate less processed, more natural images
output_formatenum—jpgjpg, png
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


TrueFusion Ultra

PropertyValue
Namespacetruefusion-ultra
VendorSkytells
Pricing$0.15 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational

Flagship model with stunning photorealism, artistic creativity, and unmatched consistency across styles. Supports inpainting, style references, and magic prompt optimization.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for image generation
aspect_ratioenum—1:115 ratios including 1:3, 3:1
resolutionenum—NoneSpecific resolution override (e.g., 1024x1024, 1536x640)
magic_prompt_optionenum—AutoAuto, On, Off — optimizes prompt for quality
imagestring (URI)——Image for inpainting (requires mask)
maskstring (URI)——Black/white mask for inpainting
style_typeenum—NoneAuto, General, Realistic, Design
style_reference_imagesarray (URI)——Style reference images
seedinteger——Random seed (max 2,147,483,647)
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


TrueFusion 2.0

PropertyValue
Namespacetruefusion-2
VendorSkytells
Pricing$0.07–$0.10 / image (resolution-based)
Capabilitiestext-to-image, image-to-image, reference, quality
StatusOperational

Attach up to 3 reference images as ground truth and tag them in your prompt using @tag_name. Preserves identity, style, and materials while giving control over angle, composition, and lighting.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for image generation
aspect_ratioenum—16:916:9, 9:16, 4:3, 3:4, 1:1, 21:9
resolutionenum—1080p720p, 1080p
reference_imagesarray (URI)—[]Up to 3 reference images (0.5–2 aspect ratio)
reference_tagsarray (string)—[]Tags for references (use @tag_name in prompt)
seedinteger——Random seed
remix_idstring——Remix from another Skytells prediction
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.

ResolutionPrice
720p$0.07 / image
1080p$0.10 / image

TrueFusion Optima

PropertyValue
Namespacetruefusion-2-optima
VendorSkytells
Pricing$0.008 / computing second
Capabilitiestext-to-image, image-to-image, quality
StatusOperational
Cold BootYes (CPU-based deployment)

Next-generation MoE architecture delivering unmatched realism, lifelike lighting, and film-grade image precision. Billed by compute time rather than per-image.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for image generation
aspect_ratioenum—1:11:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:3, 3:1
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


TrueFusion X

PropertyValue
Namespacetruefusion-x
VendorSkytells
Pricing$0.10 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational

Ultra-fast, ultra-high-resolution with inpainting support and quality control.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
nnumber—1Number of images
imagearray (URI)——Reference images
maskstring (URI)——Inpainting mask
qualityenum—mediumlow, medium, high
sizeenum—1024x1024768x1152, 1152x768, 1792x1024, 1920x822, etc.
output_formatenum—jpgjpg, png
output_compressionnumber—100Compression ratio (1–100)
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


TrueFusion Edge

PropertyValue
Namespacetruefusion-edge
VendorSkytells
Pricing$0.01 / image
Capabilitiestext-to-image, image-to-image, fast
StatusOperational

Ultra-fast, lightweight model optimized for speed. Only 4 denoising steps needed.

ParameterTypeRequiredDefaultDescription
promptstring✅—Prompt for generated image
aspect_ratioenum—1:111 ratios
num_outputsinteger—1Number of outputs (1–4)
num_inference_stepsinteger—4Denoising steps (1–4)
seedinteger——Random seed
output_formatenum—webpwebp, jpg, png
output_qualityinteger—80Quality (0–100)
go_fastboolean—trueUse fp8 quantized model
megapixelsenum—11, 0.25
disable_safety_checkerboolean—falseDisable safety checker
{
  "type": "array",
  "items": { "type": "string", "format": "uri" }
}

Returns an array of image URLs.


TrueFusion Pano

PropertyValue
Namespacetruefusion-pano
VendorSkytells
Pricing$0.02 / GPU second
Capabilitiestext-to-image, image-to-image
StatusOperational

Multiple model variants available including multilingual (English + Chinese).

ParameterTypeRequiredDefaultDescription
promptstring✅—Input prompt
negative_promptstring—""Things to exclude
model_variantenum—1600M-1024px1600M-1024px, 1600M-1024px-multilang, 1600M-512px, 600M-1024px-multilang, 600M-512px-multilang
widthinteger—1024Output width
heightinteger—1024Output height
num_inference_stepsinteger—18Denoising steps
guidance_scalenumber—5CFG scale (1–20)
pag_guidance_scalenumber—2PAG guidance (1–20)
seedinteger——Random seed
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


TrueFusion Standard

PropertyValue
Namespacetruefusion-standard
VendorSkytells
Pricing$0.05 / image
Capabilitiestext-to-image
StatusOperational

Full-featured model with LoRA support. Load custom LoRA weights from Skytells, HuggingFace, or CivitAI.

ParameterTypeRequiredDefaultDescription
promptstring✅—Prompt for generated image
aspect_ratioenum—1:111 ratios
imagestring (URI)——Input image for img2img
prompt_strengthnumber—0.8Prompt strength for img2img (0–1)
num_outputsinteger—1Number of outputs (1–4)
num_inference_stepsinteger—28Denoising steps (1–50)
guidancenumber—3Guidance (0–10)
lora_weightsstring——LoRA weights URL or identifier (e.g., skytells/truefusion-base)
lora_scalenumber—1LoRA strength (-1 to 3)
seedinteger——Random seed
output_formatenum—webpwebp, jpg, png
output_qualityinteger—80Quality (0–100)
go_fastboolean—trueUse fp8 quantized model
megapixelsenum—11, 0.25
disable_safety_checkerboolean—falseDisable safety checker
{
  "type": "array",
  "items": { "type": "string", "format": "uri" }
}

Returns an array of image URLs.


TrueFusion Variant

PropertyValue
Namespacetruefusion-variant
VendorSkytells
Pricing$0.05 / image
Capabilitiestext-to-image, image-to-image
StatusOperational

Schema not yet published — contact support for input details.


Flux.1 Edge

PropertyValue
Namespaceflux-fast
VendorSkytells
Pricing$0.01 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational

Super-fast Flux model optimized by Skytells for instant generation, with configurable speed modes.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
speed_modeenum—Extra Juiced 🔥Speed optimization level
guidancenumber—3.5Guidance scale
image_sizeinteger—1024Base image size (longest side)
aspect_ratioenum—1:111 ratios
output_formatenum—jpgpng, jpg, webp
output_qualityinteger—80Quality (1–100)
num_inference_stepsinteger—28Inference steps
seedinteger—-1Random seed
{
  "type": "string",
  "format": "uri"
}

Returns a single image URL.


Imagen 3

PropertyValue
Namespacegoogle-imagen-3
VendorGoogle
Pricing$0.08 / image
Capabilitiestext-to-image, image-to-image
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
negative_promptstring——What to discourage
aspect_ratioenum—1:11:1, 9:16, 16:9, 3:4, 4:3
safety_filter_levelenum—block_medium_and_aboveblock_low_and_above, block_medium_and_above, block_only_high
{
  "type": "string",
  "format": "uri"
}

Imagen 4

PropertyValue
Namespacegoogle-imagen-4
VendorGoogle
Pricing$0.08 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational

Google's flagship text-to-image model.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
aspect_ratioenum—1:11:1, 9:16, 16:9, 3:4, 4:3
safety_filter_levelenum—block_medium_and_aboveblock_low_and_above, block_medium_and_above, block_only_high
{
  "type": "string",
  "format": "uri"
}

Nano Banana

PropertyValue
Namespacenano-banana
VendorGoogle
Pricing$0.06 / image
Capabilitiestext-to-image, image-to-image, fast
StatusOperational

Google's Gemini 2.5-based image editing model with multi-image input support.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
image_inputarray (URI)—[]Input images for transformation (multiple supported)
aspect_ratioenum—match_input_imagematch_input_image, 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9
output_formatenum—jpgjpg, png
{
  "type": "string",
  "format": "uri"
}

Sana

PropertyValue
Namespacenvidia-sana
VendorNvidia
Pricing$0.05 / image
Capabilitiestext-to-image
StatusOperational

Fast image model with wide artistic range and resolutions up to 4096×4096. Multiple model variants including multilingual.

ParameterTypeRequiredDefaultDescription
promptstring✅—Input prompt
negative_promptstring—""Things to exclude
model_variantenum—1600M-1024px1600M-1024px, 1600M-1024px-multilang, 1600M-512px, 600M-1024px-multilang, 600M-512px-multilang
widthinteger—1024Output width
heightinteger—1024Output height
num_inference_stepsinteger—18Denoising steps
guidance_scalenumber—5CFG scale (1–20)
pag_guidance_scalenumber—2PAG guidance (1–20)
seedinteger——Random seed
{
  "type": "string",
  "format": "uri"
}

GPT-Image-1

PropertyValue
Namespacegpt-image-1
VendorOpenAI
Pricing$0.002 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational
InferencePartner-served

Requires your own OpenAI API key.

ParameterTypeRequiredDefaultDescription
openai_api_keystring (password)✅—Your OpenAI API key
promptstring✅—Text description
aspect_ratioenum—1:11:1, 3:2, 2:3
input_imagesarray (URI)——Input images for editing
number_of_imagesinteger—1Number of images (1–10)
qualityenum—autolow, medium, high, auto
backgroundenum—autoauto, transparent, opaque
output_formatenum—webppng, jpeg, webp
output_compressioninteger—90Compression (0–100%)
moderationenum—autoauto, low
user_idstring——End-user identifier for abuse monitoring
{
  "type": "array",
  "items": { "type": "string", "format": "uri" }
}

Returns an array of image URLs.


GPT Image 2

PropertyValue
Namespacegpt-image-2
VendorOpenAI
PricingToken-based (see Pricing Details)
Capabilitiestext-to-image, image-to-image, quality
StatusOperational
InferencePartner-served

OpenAI's state-of-the-art image generation model for fast, high-quality image generation and editing. Supports flexible image sizes and high-fidelity image inputs. Requires your own OpenAI API key.

ParameterTypeRequiredDefaultDescription
openai_api_keystring (password)✅—Your OpenAI API key
promptstring✅—Text prompt for image generation
input_imagesarray (URI)—[]Input images for image-to-image editing (max 8, jpeg/png/gif/webp)
ninteger—1Number of images to generate (1–10)
sizeenum—1024x10241024x1024, 1024x1536, 1536x1024
output_formatenum—pngwebp, jpg, png
output_compressioninteger—50Output quality compression (0–100)
{
  "type": "string",
  "format": "uri"
}

Returns an image URL.

Priced per token consumed. Baseline rates (USD per 1M tokens):

Token TypeBaseline Rate
Image output$30 / 1M tokens
Image input$8 / 1M tokens
Cached image input$2 / 1M tokens
Text input$5 / 1M tokens
Cached text input$1.25 / 1M tokens

Size adjustments: Portrait (1024×1536) and Landscape (1536×1024) outputs are billed at $32/1M image output tokens vs $30 for square.

Image-to-image: When input_images is provided, text input rate increases to $5.25/1M and image input to $10/1M.


FLUX-1.1 Pro

PropertyValue
NamespaceFLUX-1.1-pro
VendorBlack Forest Labs
Pricing$0.04 / image
Capabilitiestext-to-image, image-to-image, quality
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
nnumber✅1Number of images
sizeenum✅1024x10241024x1024, 768x1152, 1152x768, 1792x1024, etc.
output_formatenum✅jpgjpg, png
{
  "type": "string",
  "format": "uri"
}

FLUX.2 Pro

PropertyValue
NamespaceFLUX.2-pro
VendorBlack Forest Labs
Pricing$0.02 / image megapixel
Capabilitiestext-to-image, image-to-image, quality
StatusOperational

Megapixel-based pricing. Supports up to 8 input reference images.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
input_imagesarray (URI)—[]Up to 8 reference images
aspect_ratioenum—1:1match_input_image, custom, plus standard ratios
resolutionenum—1 MPmatch_input_image, 0.5 MP, 1 MP, 2 MP, 4 MP
widthinteger——Custom width (256–2048, multiples of 32)
heightinteger——Custom height (256–2048, multiples of 32)
safety_toleranceinteger—2Safety level (1–5)
seedinteger——Random seed
output_formatenum—pngwebp, jpg, png
output_qualityinteger—100Quality (0–100)
{
  "type": "string",
  "format": "uri"
}
ResolutionPrice per image
0.5 MP$0.01
1 MP$0.02
2 MP$0.04
4 MP$0.08

Flux 2 Pro Legacy

PropertyValue
Namespaceflux-2-pro-legacy
VendorBlack Forest Labs
Pricing$0.02 / image megapixel
Capabilitiestext-to-image, image-to-image, editing, quality
StatusOperational

Supports up to 8 reference images. Same schema as FLUX.2 Pro with additional safety_tolerance control.


Flux 2 Flex

PropertyValue
Namespaceflux-2-flex
VendorBlack Forest Labs
Pricing$0.08 / image megapixel
Capabilitiestext-to-image, image-to-image, editing, quality
StatusOperational

Max-quality generation with up to 10 reference images, configurable inference steps, guidance, and prompt upsampling.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
input_imagesarray (URI)—[]Up to 10 reference images
aspect_ratioenum—1:1match_input_image, custom, plus standard ratios
resolutionenum—1 MP0.5 MP, 1 MP, 2 MP, 4 MP
width / heightinteger——Custom dimensions (256–2048)
stepsinteger—30Inference steps (1–50)
guidancenumber—4.5Guidance scale (1.5–10)
prompt_upsamplingboolean—trueAuto-modify prompt for creativity
safety_toleranceinteger—2Safety (1–5)
seedinteger——Random seed
output_formatenum—pngwebp, jpg, png
output_qualityinteger—100Quality (0–100)
{
  "type": "string",
  "format": "uri"
}

Video Models

TrueFusion Video Pro

PropertyValue
Namespacetruefusion-video-pro
VendorSkytells
Pricing$0.196 / second
Capabilitiestext-to-video, image-to-video
StatusOperational
ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for video generation
negative_promptstring—""Things to exclude
aspect_ratioenum—16:916:9, 9:16, 1:1
start_imagestring (URI)——First frame of the video
end_imagestring (URI)——Last frame of the video
cfg_scalenumber—0.5Flexibility scale (0–1). Higher = more constrained
durationenum—55, 10 seconds
{
  "type": "string",
  "format": "uri"
}

Returns a video URL.


TrueFusion Video

PropertyValue
Namespacetruefusion-video
VendorSkytells
Pricing$0.112 / second
Capabilitiestext-to-video, image-to-video
StatusOperational

Same schema as TrueFusion Video Pro but without end_image support. Lower cost option.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
negative_promptstring—""Things to exclude
aspect_ratioenum—16:916:9, 9:16, 1:1
start_imagestring (URI)——First frame
cfg_scalenumber—0.5Flexibility scale (0–1)
durationenum—55, 10 seconds
{
  "type": "string",
  "format": "uri"
}

Mera

PropertyValue
Namespacemera
VendorSkytells
Pricing$3.42 / prediction
Capabilitiestext-to-video, image-to-video, audio, quality
StatusOperational

Skytells's latest video generation model — physically accurate, super realistic, and controllable. Supports reference images for subject consistency.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
secondsenum✅84, 6, 8, 12 seconds
sizeenum✅720x1280720x1280, 1280x720
input_referencearray (URI)—[]1–3 reference images for R2V
{
  "type": "string",
  "format": "uri"
}

Mera 1.1 Fast

PropertyValue
Namespacemera-1.1-fast
VendorSkytells
Pricing$0.03–$0.05 / second (resolution + draft tier)
Capabilitiesimage-to-video, text-to-video, audio-to-video, fast
StatusOperational

Mera's fast variant combining precision with rapid draft previews, robust lipsync, built-in audio generation, and up to 1080p at 48 FPS with multi-aspect ratio support and prompt upsampling.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt for video generation
imagestring (URI)——Input image for image-to-video (jpg, jpeg, png, webp)
last_frame_imagestring (URI)——Reference image for the last frame
audiostring (URI)——Input audio to condition video generation (flac, mp3, wav)
durationinteger—5Duration in seconds (1–20). Ignored when audio is provided
aspect_ratioenum—16:916:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:1. Ignored when input image is provided
resolutionenum—720p720p, 1080p
fpsenum—2424, 48
draftboolean—falseDraft mode — generates a lower-quality preview
prompt_upsamplingboolean—trueEnhance the prompt automatically
save_audioboolean—trueSave the video with audio
disable_safety_filterboolean—trueDisable safety filter for prompts and input image
seedinteger——Random seed for reproducible generation
{
  "type": "string",
  "format": "uri"
}
TierResolutionDraftPrice per second
720p_standard720pOff$0.03
720p_draft720pOn$0.007
1080p_standard1080pOff$0.05
1080p_draft1080pOn$0.03

Mera Avatar

PropertyValue
Namespacemera-avatar
VendorSkytells
Pricing$0.036 / second (720p) · $0.06 / second (1080p)
Capabilitiesimage-to-video, text-to-video, audio-to-video, fast, avatar, lipsync
StatusOperational

Flagship avatar video generation model with robust lip synchronization, native audio generation, and highly controllable generation up to 1080p at 48 FPS. Supports voice scripting with 30+ voice presets or custom audio upload.

ParameterTypeRequiredDefaultDescription
imagestring (URI)✅—First frame image (jpg, jpeg, png, webp)
resolutionenum—720p720p, 1080p
audiostring (URI)——Optional uploaded audio to drive avatar speech. Overrides voice_script
voice_scriptstring—""Exact words the avatar should say. Required when no audio is uploaded
voiceenum—Zephyr (Female)Voice preset (30+ options: Zephyr, Puck, Charon, Kore, Fenrir, Leda, Orus, Aoede, and more)
voice_languageenum—English (US)English (US), English (UK), Spanish, French, German, Italian, Portuguese (Brazil), Japanese, Korean, Hindi
voice_promptstring—"Say the following."Style instructions for speech (tone, pacing, accent, emotion) — not spoken aloud
video_promptstring—"The person is talking."Visual prompt describing appearance or behavior while speaking
negative_promptstring—""What to exclude from the video
strength_negative_promptnumber—0.5Strength of negative prompt (0–4)
hyper_motionboolean—falseAdaptive inference for extreme motion, multi-character, and cinematic scenarios
disable_prompt_upsamplingboolean—falseSkip automatic visual prompt enhancement
disable_safety_filterboolean—trueDisable safety filter
seedinteger——Random seed for reproducible generation
{
  "type": "string",
  "format": "uri"
}
ResolutionPrice per second
720p$0.036
1080p$0.06

Lumo

PropertyValue
Namespacelumo
VendorSkytells
Pricing$1.12 / prediction
Capabilitiesimage-to-video
StatusOperational

Specialized for motion, animations, and general use cases. Schema not yet published.


LipFusion

PropertyValue
Namespacelipfusion
VendorSkytells
Pricing$0.04 / second
Capabilitiesvideo-to-video, audio-to-video
StatusOperational

Ultra-realistic lip-syncing for videos, animations, avatars, and live streams. Supports audio file input or text-to-speech with 45+ voice presets.

ParameterTypeRequiredDefaultDescription
video_urlstring (URI)✅—Target video (.mp4/.mov, under 100MB, 2–10s, 720p–1080p)
audio_filestring (URI)✅—Audio for lip sync (.mp3/.wav/.m4a/.aac, under 5MB)
textstring——Text for TTS lip sync (Enterprise only)
voice_idenum—en_AOT45+ voice presets (English + Chinese)
motion_awareboolean—trueAdjust sync based on subject movements
voice_speednumber—1Speech rate for TTS (0.8–2)
{
  "type": "string",
  "format": "uri"
}

Veo 3.1

PropertyValue
Namespaceveo-3.1
VendorGoogle
Pricing$0.43 / second
Capabilitiestext-to-video, image-to-video, quality
StatusOperational

Google's next-gen video model with context-aware audio, reference images, and last-frame interpolation.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
aspect_ratioenum—16:916:9, 9:16
durationenum—84, 6, 8 seconds
imagestring (URI)——Start image
last_framestring (URI)——End image for interpolation
reference_imagesarray (URI)—[]1–3 reference images for R2V (16:9, 8s only)
negative_promptstring——What to exclude
resolutionenum—1080p720p, 1080p
generate_audioboolean—trueGenerate audio with video
seedinteger——Random seed
{
  "type": "string",
  "format": "uri"
}

Veo 3.1 Fast

PropertyValue
Namespaceveo-3.1-fast
VendorGoogle
Pricing$0.13–$0.17 / second (audio-dependent)
Capabilitiestext-to-video, image-to-video, fast
StatusOperational

Faster variant of Veo 3.1 with dynamic pricing based on audio generation.

ParameterTypeRequiredDefaultDescription
promptstring✅—Text prompt
aspect_ratioenum—16:916:9, 9:16
durationenum—84, 6, 8 seconds
imagestring (URI)——Start image
last_framestring (URI)——End image for interpolation
negative_promptstring——What to exclude
resolutionenum—1080p720p, 1080p
generate_audioboolean—trueGenerate audio with video
seedinteger——Random seed
{
  "type": "string",
  "format": "uri"
}
Audio GenerationPrice per second
Enabled$0.17
Disabled$0.13

Veo 3.1 (Preview)

PropertyValue
Namespaceveo-3.1-preview
VendorGoogle
Pricing$0.43 / second
Capabilitiestext-to-video, image-to-video, quality, sound
StatusOperational

Preview version with person_generation safety control. Same schema as Veo 3.1 with additional person generation options (allow_adult, dont_allow).


Sora 2

PropertyValue
Namespacesora-2
VendorOpenAI
Pricing$0.002 / video
Capabilitiestext-to-video, image-to-video
StatusOperational
InferencePartner-served

Requires your own OpenAI API key.

ParameterTypeRequiredDefaultDescription
openai_api_keystring (password)✅—Your OpenAI API key
promptstring✅—Text description of the video
secondsenum—44, 8, 12
aspect_ratioenum—portraitportrait (720×1280), landscape (1280×720)
input_referencestring (URI)——Reference image or video
{
  "type": "string",
  "format": "uri"
}

Sora 2 Pro

PropertyValue
Namespacesora-2-pro
VendorOpenAI
Pricing$0.002 / video
Capabilitiestext-to-video, image-to-video
StatusOperational
InferencePartner-served

Same as Sora 2 with additional resolution control (standard = 720p, high = 1024p). Requires your own OpenAI API key.


Wan 2.5-i2v

PropertyValue
Namespacewan-2.5-i2v
VendorAlibaba
Pricing$0.06–$0.16 / second (resolution-based)
Capabilitiesimage-to-video, reference, quality, audio, video
StatusOperational

Alibaba's image-to-video model with background audio, prompt expansion, and multi-resolution support.

ParameterTypeRequiredDefaultDescription
imagestring (URI)✅—Input image
promptstring✅—Text prompt
negative_promptstring—""What to exclude
resolutionenum—720p480p, 720p, 1080p
durationenum—55, 10 seconds
audiostring (URI)——Audio file for sync (wav/mp3, 3–30s, ≤15MB)
enable_prompt_expansionboolean—trueEnable prompt optimizer
seedinteger——Random seed
{
  "type": "string",
  "format": "uri"
}
ResolutionPrice per second
480p$0.06
720p$0.11
1080p$0.16

Video Upscale

PropertyValue
Namespacevideo-upscale
VendorTopaz Labs
Pricing$0.10 / 5 seconds
Capabilitiesvideo-to-video
StatusOperational
ParameterTypeRequiredDefaultDescription
videostring (URI)✅—Video file to upscale
target_resolutionenum—1080p720p, 1080p, 4k
target_fpsinteger—30Target FPS (15–60)
{
  "type": "string",
  "format": "uri"
}

Audio Models

BeatFusion 2.0

PropertyValue
Namespacebeatfusion-2.0
VendorSkytells
Pricing$0.75 / prediction
Capabilitiestext-to-audio, music, quality, audio
StatusOperational

Skytells's flagship music generation model. Generate full-length songs with vocals, lyrics, and rich instrumentation. Supports structure tags: [Intro], [Verse], [Pre Chorus], [Chorus], [Bridge], [Outro], [Hook], [Solo], and more.

ParameterTypeRequiredDefaultDescription
lyricsstring✅—Lyrics with structure tags (1–3500 chars). Use \n for line breaks
promptstring—""Music style description (0–2000 chars)
sample_rateenum—4410016000, 24000, 32000, 44100
bitrateenum—25600032000, 64000, 128000, 256000
audio_formatenum—mp3mp3, wav, pcm
{
  "type": "string",
  "format": "uri"
}

Returns an audio file URL.


BeatFusion 2.1

PropertyValue
Namespacebeatfusion-2.1
VendorSkytells
Pricing$0.85 / prediction
Capabilitiestext-to-audio, music, quality, audio, text-to-music
StatusOperational

Skytells's latest music generation model. Generate full-length songs with vocals, lyrics, and rich instrumentation. Adds is_instrumental mode and lyrics_optimizer for automatic lyric generation from a prompt. Supports structure tags: [Intro], [Verse], [Pre Chorus], [Chorus], [Interlude], [Bridge], [Outro], [Post Chorus], [Transition], [Break], [Hook], [Build Up], [Inst], [Solo].

ParameterTypeRequiredDefaultDescription
lyricsstring—""Lyrics with structure tags (1–3500 chars). Required unless is_instrumental is true
promptstring—""Musical style, emotional tone, and context (0–2000 chars). Required when is_instrumental is true
is_instrumentalboolean—falseGenerate instrumental music without vocals. When true, prompt is required
lyrics_optimizerboolean—falseAuto-generate lyrics from prompt when lyrics is empty
sample_rateenum—4410016000, 24000, 32000, 44100
bitrateenum—25600032000, 64000, 128000, 256000
audio_formatenum—mp3mp3, wav, pcm
{
  "type": "string",
  "format": "uri"
}

Returns an audio file URL.


BeatFusion 1.0

PropertyValue
Namespacebeatfusion-1.0
VendorSkytells
Pricing$0.45 / prediction
Capabilitiestext-to-audio, music, quality, audio
StatusOperational

First-generation music model. Supports [intro], [verse], [chorus], [bridge], [outro] tags.

ParameterTypeRequiredDefaultDescription
promptstring✅—Music style description (10–300 chars)
lyricsstring✅—Lyrics with structure tags (10–600 chars)
sample_rateenum—4410016000, 24000, 32000, 44100
bitrateenum—25600032000, 64000, 128000, 256000
audio_formatenum—mp3mp3, wav, pcm
{
  "type": "string",
  "format": "uri"
}

Understanding Model Schemas

Every model on Skytells defines its own input_schema and output_schema. These schemas follow JSON Schema conventions and describe exactly what parameters a model accepts and what it returns.

Why schemas differ

Different models are built for different tasks, so their inputs vary significantly:

  • Simple models like TrueFusion only need a prompt and optional aspect_ratio
  • Advanced models like TrueFusion Pro add controls for guidance, num_inference_steps, seed, and output format
  • Flagship models like TrueFusion Ultra support inpainting (image + mask), style references, and resolution presets
  • Reference models like TrueFusion 2.0 accept tagged reference images you can invoke by name in your prompt
  • Video models add duration, start_image/end_image, and cfg_scale for temporal control
  • Audio models use lyrics and prompt for compositional control with structure tags
  • Partner models (GPT-Image-1, Sora 2) require your own API key via openai_api_key

Common input parameters

Most image models share these parameters:

ParameterDescription
promptAlways required. The text description of what to generate
aspect_ratioControls image dimensions. Available ratios vary by model
seedFor reproducible generations. Set the same seed to get identical outputs
output_formatUsually webp, jpg, or png

Output schema patterns

Models return one of two patterns:

Single output — returns one URL:

{ "type": "string", "format": "uri" }

Multiple outputs — returns an array of URLs:

{ "type": "array", "items": { "type": "string", "format": "uri" } }

Models that support num_outputs or number_of_images typically use the array pattern.

Pricing units

UnitDescription
imageFlat rate per generated image
secondBilled per second of generated video/audio
predictionFlat rate per API call
gpuBilled per GPU second used
computing_secondBilled per compute second
image_megapixelBilled by output resolution in megapixels
5 secondsBilled per 5-second chunk

How is this guide?

On this page