Images & Video API

Authenticate with Authorization: Bearer $SCALEMAX_API_KEY. Match the billing surface: credit keys use https://api.scalemax.pro/v1; token keys use https://api.scalemax.pro/token/v1. Image models appear in GET /v1/models with family: "image". Video models use family: "video" when listed.

Image model ids and prices

Flat per-image pricing by quality tier (same rates as ScaleMax Image 2). Token keys debit the derived token units at the published $2.50 = 1M rate. Keys other than Claude, Codex and combined keys pay 2x these image prices.

POST /v1/images/generations

OpenAI-compatible image generation. Returns a synchronous JSON body with image URLs or base64 data.

curl -X POST "https://api.scalemax.pro/v1/images/generations" \
  -H "Authorization: Bearer $SCALEMAX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"grok-imagine-image","prompt":"A red circle on white","size":"1:1","quality":"medium","n":1}'

@'
{"model":"grok-imagine-image","prompt":"A red circle on white","size":"1:1","quality":"medium","n":1}
'@ | Set-Content -Path body.json

curl.exe -X POST "https://api.scalemax.pro/v1/images/generations" `
  -H "Authorization: Bearer $env:SCALEMAX_API_KEY" `
  -H "Content-Type: application/json" `
  --data-binary "@body.json"

POST /v1/images/edits

Edit an existing image. Grok Imagine models accept a reference image URL or upload (model-dependent).

Video model ids and prices

Priced per clip by length and resolution. The prices below are for a 6-second 720p clip; a clip costs that price × seconds ÷ 6 × 0.7 at 480p, 1.0 at 720p or 1.6 at 1080p, rounded up. A request with noresolution is generated and billed at 720p. The full price is reserved before the job starts, and a key that cannot cover it gets a 402 naming the amount needed. Trial keys can make one 480p clip of up to 6 seconds on grok-imagine-video. Duration 1–15 seconds. 1080p is grok-imagine-video-1.5 only. Edit is grok-imagine-video only. Extend is not available on the current hop.

Seedance and Kling are priced per second at each resolution: a clip costs that rate × seconds. Video tokens are 1 per micro-dollar, so a $1.00 clip debits 1,000,000 tokens from a pack. These models make video from text only (no input image, edit or extend). Seedance takes 4–15 seconds, Kling 3–15.

Video routes

Native video generation is async: create returns 202 with a task id, poll until status is terminal, then download from the ScaleMax url field (never a third-party host). Completed jobs include usage with total_tokens and frontend_cost_microusd.

POST /v1/videos/generations

Start text-to-video or image-to-video generation.

POST /v1/videos/edits

Edit an existing clip with a new prompt. grok-imagine-video only. grok-imagine-video-1.5 returns 400.

POST /v1/videos/extensions

Route exists. The current Imagine hop rejects extend (HTTP 405). Do not call it.

GET /v1/videos/:id

Poll task status and progress. Optional ?wait=0..20 long-poll seconds.

GET /v1/videos/:id/content

Stream the finished mp4. Supports Range after the task completes.

curl -X POST "https://api.scalemax.pro/v1/videos/generations" \
  -H "Authorization: Bearer $SCALEMAX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"grok-imagine-video","prompt":"Ocean waves at sunset","aspect_ratio":"16:9","duration":6}'

@'
{"model":"grok-imagine-video","prompt":"Ocean waves at sunset","aspect_ratio":"16:9","duration":6}
'@ | Set-Content -Path body.json

curl.exe -X POST "https://api.scalemax.pro/v1/videos/generations" `
  -H "Authorization: Bearer $env:SCALEMAX_API_KEY" `
  -H "Content-Type: application/json" `
  --data-binary "@body.json"

Key scope

• Credit keys: any general-purpose scope may call every image and video model.
• Token keys on Claude, OpenAI, combined, Gemini, GLM, Local, and Qwen packs may call every published image model and every published video model.
• grok-imagine-image* and grok-imagine-video* are also on every other general-purpose key (DeepSeek, MiMo, Grok, reseller, …). Auto API and Voice keys cannot. Grok / DeepSeek / MiMo / reseller token packs still cannot call Azure or Qwen image models.