Overview

Models

Don't want to choose? A model router picks per request based on your cost, latency or quality preference, at no extra charge.

Choosing a video model#

If you need Use Cost
Up to 30 seconds, cinematic, large reference budget seedance2_5 20–30 credits/second
Best motion quality and prompt adherence gen4.5 12 credits/second
Lowest cost per second gen4_turbo 5 credits/second
Editing an existing video with a prompt aleph2 28 credits/second
Native audio with the video veo3.1 40 credits/second with audio
Driving a character's performance act_two 5 credits/second
4K output seedance2 150 credits/second

See video models for resolutions, durations and reference limits.

Choosing an image model#

If you need Use Cost
The cheapest usable image gen4_image_turbo 2 credits, any resolution
A production still gen4_image 5 credits (720p) / 8 (1080p)
Highest fidelity at 4K gemini_image3_pro 40 credits per 4K image
Precise text rendering gpt_image_2 1–41 credits by quality

See image models for reference-image behaviour and quality tiers.

Full model list#

Video generation#

Model Input Output
seedance2_5 Text, Image, or Video Video
grok_imagine_1_5 Text or Image Video
seedance2 Text, Image, or Video Video
seedance2_fast Text, Image, or Video Video
seedance2_mini Text, Image, or Video Video
hailuo3 Text, Image, or Video Video
aleph2 Video + Text/Image Video
gen4.5 Text or Image Video
gen4_turbo Image Video
act_two Image or Video Video
veo3.1 Text or Image Video
veo3.1_fast Text or Image Video
happyhorse_1_0 Text or Image Video
gemini_omni_flash Text, Image, or Video Video

Image generation#

Model Input Output
grok_imagine_image_2 Text/Image (references) Image
seedream5_pro Text/Image (references) Image
seedream5_lite Text/Image (references) Image
gen4_image Text/Image (references) Image
gen4_image_turbo Text + Image (references) Image
gemini_image3_pro Text/Image (references) Image
gemini_image3.1_flash Text/Image (references) Image
gpt_image_2 Text/Image (references) Image
gemini_2.5_flash Text/Image (references) Image

Audio generation#

Model Input Output
seed_audio Text or Audio Audio
eleven_v3 Text Audio
eleven_multilingual_v2 Text Audio
eleven_text_to_sound_v2 Text Audio
eleven_voice_isolation Audio Audio
eleven_voice_dubbing Audio Audio
eleven_multilingual_sts_v2 Audio Audio

Real-time#

Model Input Output
gwm1_avatars Text (conversation) Video + Audio

Upscaling#

Model Input Output
magnific_precision_upscaler_v2 Image Image
magnific_video_upscaler_creative Video Video

Video upscaling always uses model: magnific_video_upscaler_creative on POST /v1/video_upscale, and input videos can be at most 30 seconds.

  • Pricing — exact credit cost of every model above
  • Inputs — per-model aspect ratio and reference limits
  • Making API calls — how to invoke any of them
Try a model against your own prompt

The playground runs every model on this page without writing code.

Updated

Was this page helpful?