Percify APIfrom $0.05 a second
Make any face speak
Send a portrait and an audio file and get a lip synced video back. Priced by the second of audio, so a short clip costs cents.
InfiniteTalk, InfiniteTalk Fast
Browse 2 models
Percify API
130 AI models, one API key
Talking video, image and voice from one endpoint. You pay per run.
API keys come with the Scale and Ultra plans
curl https://api.percify.io/v3/playground/v1/run \
-H "Authorization: Bearer $PERCIFY_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"modelId": "infinitetalk-fast",
"input": {
"image": "https://example.com/portrait.jpg",
"audio": "https://example.com/voice.mp3"
}
}'Veo 3.1Sora 2Kling V3Seedance 2.5Wan 3.0Hailuo 2.3Luma Ray 3.2Pika 2.2LTX-2InfiniteTalkGPT Image 2Nano Banana ProImagen 4Flux 2MidjourneyIdeogram V4Recraft V4Seedream 5.0ElevenLabsLyria 3 Pro
What you can build
Everything Percify makes, as an API
The models behind the Percify app, callable from your own code.
from $0.05 a second
Generate video from text or a still
The current video models behind one request shape. Switch from a fast draft model to a cinematic one by changing one string.
Veo 3.1, Sora 2, Kling V3, Seedance 2.5, Wan 3.0
Browse 68 models
from $0.025 a runCreate and edit images
Text to image, reference edits, inpainting and head swaps.
GPT Image 2, Nano Banana Pro, Flux 2, Midjourney
Browse 41 models
from $0.05 a runSpeech, cloned voices and music
Text to speech, voice cloning from a short sample, songs and sound effects.
ElevenLabs V3, Zonos 2, Lyria 3 Pro
Browse 17 models
from $1.80 a runDub and translate
Translate a finished video into another language, voice and lips included.
ElevenLabs Dubbing, HeyGen Video Translate
Browse 2 modelsUse cases
78 jobs, ready to run
Each one opens a playground with the request already filled in.
See every use case- UGC style video ads from a scriptA script and a face become a talking head ad, so a new hook costs a request instead of a shoot.
- Turn a product photo into a videoA product photo becomes a short video for the listing or an ad.
- Dub a video into another languageA finished video comes back speaking another language.
- Rewrite what people say in a videoThe clip keeps its scene and performance while the speaker says your new lines, with the lips matched to them.
- Clone a voice from a short sampleA short sample and a line of text return that line in the sampled voice.
- Personalized sales videos at scaleEach prospect gets a video that says their name and company, generated from one photo and a template script.
Console
See every dollar before it leaves
Balance, keys and every request in one place, in dollars.
Open the consoleA price before every run
The estimate call prices your real input, length and resolution included, and charges nothing.
A limit on every key
Give a key a monthly limit and an expiry date. At the limit it stops, and you are told at 80%.
Failed runs cost nothing
A run is charged when it starts. If it fails or times out, the money returns to your balance by itself.
Balance$54.95Sample
InfiniteTalk FastSucceeded$0.60
GPT Image 2Succeeded$0.10
Veo 3.1 FastRefunded$0.00
Built for production
The parts you would build yourself
Long jobs, retries and bulk work are handled on our side.
{
"modelId": "veo-31-fast-t2v",
"input": { "prompt": "…" },
"webhook": "https://you.dev/hooks/percify"
}
# later, on your server
X-Percify-Signature: sha256 of timestamp.bodyResults come to you
Pass a webhook URL and the finished job is posted to it, signed with HMAC SHA256. Or poll the generation and wait up to 55 seconds a call.
{
"modelId": "gpt-image-2",
"prompts": ["…", "…", "…"],
"budgetCredits": 400,
"concurrency": 5
}Fifty prompts in one call
A batch runs up to 50 prompts on our servers, five at a time, and stops when it reaches the budget you set.
Authorization: Bearer pk_live_your_key
Idempotency-Key: order_8841_clip_2
# the same key twice returns the same job,
# and is charged onceSafe to retry
Send an Idempotency-Key with a run. If your request is sent twice, you get the same job back and pay for it once.
MCP server
Your agent can use it too
One URL gives Claude, ChatGPT or Cursor the same models as tools.
{
"mcpServers": {
"percify": {
"url": "https://mcp.percify.io/mcp"
}
}
}- Works on every plan, Free includedThe agent signs in with OAuth. No API key to paste.
- The same balance and pricesAn agent call spends the credits a REST call would.
- Tools for whole jobsGenerate, estimate a cost, make an avatar speak, check usage.
Prices
Priced by the run
130 models in the catalog. These are the ones developers start with.
See every model and priceModelPriceOutputWhat it does
InfiniteTalk Fast$0.05 a secondVideoFast lip-sync: animate a portrait to speak any audio in seconds.
InfiniteTalk$0.10 a secondVideoLip-sync any portrait to any audio for a natural talking-avatar video.
Veo 3.1 Fastfrom $1.50VideoGoogle Veo 3.1 Fast creates text-to-video with native 1080p and synchronized audio, delivering high-quality videos for creators.
Sora 2 (Text-to-Video)from $2.00VideoOpenAI Sora 2 is a state-of-the-art text-to-video model with realistic visuals, accurate physics, synchronized audio, and strong steerability.
GPT Image 2from $0.10ImageOpenAI image generation: excellent prompt-following and clean text rendering; takes reference images.
Nano Banana Pro$0.625 eachImageGoogle's Nano Banana pro (Gemini 3.0 Pro Image) is a cutting-edge text-to-image model enabling high-res 4K image generation optimized for phones.
Z-Image Turbo$0.05 eachImageZ-Image Turbo is a super fast text-to-image model of 6B parameters developed by Tongyi-MAI.
ElevenLabs Eleven V3$0.25 eachAudioElevenLabs eleven-v3 is a text-to-speech model available as a hosted endpoint.
Zonos 2from $0.05AudioVoice cloning + text-to-speech: clone a voice from a short sample and make it say anything, multilingual.
Dollar prices are the credit price at 200 credits for $5. Prices that start with "from" change with length or resolution, and the estimate call returns the exact figure. Full reference in the docs.
Questions, answered
How do I get an API key?
API keys come with the Scale and Ultra plans. Sign in to the console and create one. The key is shown once and stored hashed, so copy it when you make it.
How is a run priced?
Every model has a price in credits, and the estimate call returns the exact price of your input before you run it, free. Credits are sold in packs from 200 credits for $5, so one credit is $0.025. The dollar prices on this site use that rate.
What happens when a run fails?
A run is charged when it starts. If it fails or times out, the credits go back to your balance automatically.
Can I cap what a key spends?
Yes. A key can carry a monthly limit and an expiry date. At the limit the API answers 429 until the next month, and you are told at 80% and at 100%.
What are the rate limits?
Each key can make 60 requests a minute, and an account can start 60 generations a minute.
Can an AI agent use it?
Yes. The MCP server at https://mcp.percify.io/mcp works on every plan, including Free. The agent signs in with OAuth and its calls spend your credits.
Do you send webhooks?
Pass a webhook URL with a run and the result is posted to it once, signed with HMAC SHA256 in the X-Percify-Signature header. You can also poll the generation.
