Image, video and audio models overview

How to choose, call and pay for BigModel.io's 49 image, video, avatar, speech, music and sound-effect models. One API key, OpenAI-compatible endpoints, priced per image, second, character or call.

On this page
  1. Picking a model
  2. Rules shared by every media endpoint
  3. See what your key can call
  4. Troubleshooting
  5. Related

BigModel.io serves image, video, avatar, speech, music and sound-effect models with the same API key you use for text. The endpoints follow OpenAI conventions, so existing SDKs and tools work; prices are in US dollars, you pay for what is produced, and failed or cancelled tasks are free.

The full list with live prices is in the model directory. This table shows where to start for each job.

Job Endpoint Mode Priced Guide
Text to image, edits, upscale, background removal POST /v1/images/generations, POST /v1/images/edits Synchronous Per image Image generation
Image to video, first & last frame, reference video POST /v1/videos Async task Per output second Video generation
Talking avatars, lip-sync, character animation POST /v1/videos Async task Per second Avatars & lip-sync
Video upscaling, clay render POST /v1/videos Async task Per second Video generation
Text to speech POST /v1/audio/speech Returns audio Per character or audio second Text to speech
Music, sound effects, audio extraction POST /v1/audio/generations Async task Per call or second Music & sound effects
Script to finished clips POST /v1/workflows/short-drama Async workflow Each step at its model's price Short-drama workflow

#Picking a model

Use case Try Why
Posters, images with text qwen-image-2.1, gpt-image-2 Accurate English and Chinese text; qwen-image-2.1 can output transparent PNGs
High volume, low cost qwen-image-2.1-turbo, z-image-turbo Fastest and cheapest
Photoreal, polished gpt-image-2.5-sunburst, omni-image-pro, gpt-image-2 Quality first
Story clips with dialogue and sound seedance-2.0-fast, seedance-2.0, minimax-h3 Native audio, first/last frame and references
Video on a budget seedance-2.0-mini, minimax-h3 Lowest price per second
High-resolution finals seedance-2.0 (up to 4K), minimax-h3-2k For final output
Animate a single image kling-2.6-pro, vidu-q3-pro, grok-imagine-video-1.5 First frame only; some take a last frame
Talking presenter infinitetalk One portrait plus a voice track
New lines on existing footage latentsync (speech), latentsync-music (songs) Redraws only the mouth
Voice-over minimax-speech-2.8-hd, indextts-2.5 Many voices; indextts-2.5 controls emotion and duration
Songs with vocals suno-v6, suno-v4.5 Lyrics give vocals; none gives an instrumental
Background music, sound effects ace-step-1.5, stable-audio-3 Low cost, good for batches

#Rules shared by every media endpoint

  • Inputs: pass images, video and audio as public https:// URLs or base64 data URIs (data:image/png;base64,...), up to 49 MB per file.
  • Output links: url is a signed link on bigmodel.io, valid for 24 hours. Fetch the task again for a fresh link, and copy files you keep to your own storage.
  • Quote first: add "dry_run": true to any create request to get the estimated price without generating or paying.
  • Safe retries: send an Idempotency-Key header; repeating the same value returns the same task instead of a second generation.
  • Billing: a finished task is charged at the model's unit price and shows up in the console's Logs. The estimate is reserved when you submit; if your balance can't cover it you get 402. Failed and cancelled tasks are free.
  • Moderation: prompts and scripts are checked; disallowed requests are refused.

#See what your key can call

The models available depend on your key's group:

bash
curl https://bigmodel.io/v1/models -H "Authorization: Bearer $BIGMODEL_API_KEY"

#Troubleshooting

Symptom Fix
404 model_not_available Your key's group doesn't offer the model, or it went to the wrong endpoint (for example a video model sent to /v1/images/generations).
402 insufficient_balance Your balance can't cover the estimate, including reserves for unfinished tasks.
400 image is required Video models need a first frame; make one with an image model first.
A link won't open Signed links last 24 hours; get the task again for a new one.

Updated · View as Markdown