Models
Every model in the studio. Open one to see what it makes, its options and its credit prices.
Video
- Veo 3.1 FastTOP
Fast cinematic video with native sound.
Google's Veo 3.1 Fast: text or image to video with synchronised native audio, up to 1080p. One of the quicker high-end video models. Costs less than Veo 3.1 because it trades a little detail for speed.
from 41
- Talking Avatar Premium
Lifelike lip-synced performance from one photo.
Wan 2.7 image-to-video on Alibaba Cloud Singapore: a portrait driven by your audio, with more natural motion. Slower than P-Video Avatar. Costs more for the premium model, charged by the second.
from 21
- Veo 3.1 Lite
The most affordable Veo, with native sound.
Veo 3.1 Lite (Preview): quick video with native audio, but no reference images. One of the fastest video models here. The lowest-cost Veo, because it's Google's lightest version.
from 21
- Veo 3.1
Google's flagship video model with native audio.
Veo 3.1: the highest-quality Veo, with synchronised native audio and reference images. Slower than Veo 3.1 Fast. One of the most expensive video models, because Google charges more for its full-quality version.
from 161
- Wan 3.0NEW
Long takes up to 30 s, with audio references.
Alibaba's Wan 3.0 (Singapore): text, image and reference video with native audio, and the longest takes here. Medium speed. Charged by the second, so long or high-resolution clips cost more.
from 11
Image
- Nano Banana 2TOP
Fast, sharp images and edits.
Google's Gemini 3.1 Flash Image: generation and multi-image editing, up to 4K. Fast. Costs less than Nano Banana Pro because it's Google's Flash model; larger sizes cost more.
from 7
- Nano Banana ProTOP
Studio-grade images with legible text.
Google's Gemini 3 Pro Image: the most capable image and editing model here, up to 4K. Slower than Nano Banana 2. Costs more because it's Google's Pro model.
from 14
- FLUX.2 Pro
Black Forest Labs' detailed, faithful images.
Black Forest Labs' FLUX.2 Pro on DeepInfra: high-quality images that follow the prompt closely. Medium speed. Charged by image size, so larger images cost more.
from 4
- Seedream 4.5
ByteDance's crisp, high-resolution images.
ByteDance's Seedream 4.5 on DeepInfra: high-quality images from text. Medium speed. One flat price per image, whatever the settings.
from 5
- Qwen-Image
Strong multilingual text rendering.
Alibaba's Qwen-Image 3.0 (Singapore): detailed images with strong text rendering. Medium speed. One of the lower-cost image models, with one flat price per image.
from 4
Audio
- Lyria 3 Clip
30-second music clips from a prompt.
Google Lyria 3: a short, high-fidelity music clip. Fast. One of the lowest-cost music models, because clips are short.
from 5
- Gemini TTS
Expressive speech with style control and two speakers.
Gemini 3.8 Flash TTS: natural speech with preset or replicated voices, style instructions and two-speaker dialogue. Fast. Charged by the length of your text; costs more than Kokoro for the richer voices.
from 1
- Lyria 3 Pro
Full songs with vocals, up to about 3 minutes.
Google Lyria 3 Pro: full songs with vocals, with an optional inspiration image. Slower than Lyria 3, because songs are longer. Costs more for the full length.
from 9
- Voice Replication
Replicate your own voice with Gemini TTS.
Gemini 3.8 TTS voice replication: speak in a voice copied from a short sample plus a consent statement by the same speaker. Fast. Charged by the length of your text.
from 1
Safety & helpers
These aren't charged separately; their cost is included in every model's price.
Gemini Flash-Lite
Prompt enhancer and safety check #1
Rewrites prompts when you ask it to, and is the first check on every prompt and upload.
Llama Guard 4
Safety check #2
Checks prompts and images when the first check can't answer.
Qwen
Safety check #3
Checks prompts and images when the first two can't answer.
gpt-oss-safeguard-20b and Qwen3Guard-Gen-8B
Safety check #4, text
Open models on our own GPUs that check prompts when every other check is down.
Qwen3-VL-8B and ShieldGemma 2
Safety check #4, images
Open models on our own GPUs that check images when every other check is down.
Freepik NSFW detector
Second image check
Looks at every image a second time, whichever check passed it.
Gemma 4 12B
Backup prompt enhancer
Rewrites prompts on our own GPUs when the main enhancer is unavailable.
Whisper large-v3-turbo
Transcription
Turns speech into text so spoken audio can be checked and re-voiced.