IMAGE
Nano Banana 2
Google's Gemini image model. Generates and edits images from a text prompt and up to 14 reference images, at 1K, 2K or 4K.
From $0.04 per image · 1K on Kie
Available on 7 providers
Bring your own API key. Pay only the provider.
Phersa is the platform for building and monetizing AI personas. With BYOK, every image, video and voice generation runs on your own API key from one of 9 providers, including Google AI Studio, fal, Kie and ElevenLabs, and bills to your account at that provider's published rate. Phersa adds no markup.
Connect your API keysPick a provider for your API key
Each provider page lists its price for every model, compares it with the other providers, and shows how to connect the key.
- Gemini Enterprise Agent Platform Google Cloud's developer platform for Gemini models, formerly Vertex AI. New Google Cloud customers get $300 in free credit.
cloud.google.com - Google AI Studio Google's developer-direct portal for Gemini — one key, no GCP setup, free & paid tier.
aistudio.google.com - SpaceXAI Grok Imagine Image 2.0 + Video 1.5 from SpaceXAI. Images priced by quality × resolution, video per second by resolution, plus $0.01 per reference photo.
x.ai - Kie Universal AI hub aggregating image, video, audio, and chat models behind one key.
kie.ai - fal Serverless GPU platform tuned for the lowest-latency generative inference.
fal.ai - Atlas Cloud Multi-provider AI cloud with USD-native pricing across image, video, and chat models.
atlascloud.ai - WaveSpeedAI Latency-optimized unified API for image and video, USD-billed with a live balance endpoint.
wavespeed.ai - Krea Generative AI studio — real-time canvas, custom style training, and fine-tuned models.
krea.ai - ElevenLabs Speech-to-Speech voice lock — swap a clip's voice to your persona's ElevenLabs voice.
elevenlabs.io
AI models you can run on your own API key
Three image models, two video models and a voice changer, with the lowest published price for each and the providers that run it.
-
- IMAGE
GPT Image 2.5 Sunburst
OpenAI's image model. Creates images from text, or edits them from reference images. Each provider sets its own size, quality and price.
From $0.00377 + inputs per image · Low on Atlas Cloud
Available on 5 providers - IMAGE
Grok Imagine Image 2.0
SpaceXAI's Grok image model. Turns a text prompt or reference images into images at 1K or 2K.
From $0.02 per image · 1K on Kie
Available on 6 providers - VIDEO
Gemini Omni 1.1 Flash
Google's Gemini video model. Makes 3 to 10 second videos with sound from text, images or reference images, at 720p, 1080p or 4K.
From $0.06563 per second · 720p on Kie
Available on 7 providers - VIDEO
Grok Imagine Video 1.5
SpaceXAI's Grok video model. Makes videos with sound, up to 15 seconds long, from text or images, at 720p or 1080p.
From $0.0225 per second · 720p on Kie
Available on 6 providers - VOICE
ElevenLabs Voice Changer
ElevenLabs' speech-to-speech model. Keeps the words, timing and delivery of a recording and swaps in a different voice. Multilingual and English versions.
$0.002 per second · ElevenLabs Pro plan rate
Available on ElevenLabs
BYOK questions
What is BYOK (bring your own key)?
BYOK, short for bring your own key, means you connect your own API key from an AI provider, so generations run and bill on your own account. In Phersa, images, videos and voice swap run on your key at the provider's published price, with no markup.
Does Phersa add a markup to API costs?
No. When you generate, the provider bills your account directly at its published rate, and Phersa adds nothing on top of the provider's price.
Which AI providers can I use with Phersa?
Phersa works with 9 providers: Gemini Enterprise Agent Platform, Google AI Studio, SpaceXAI, Kie, fal, Atlas Cloud, WaveSpeedAI, Krea and ElevenLabs. Each provider has its own page with prices for every model and steps to connect the key.
See all providers →Which AI models can I use with my own API key?
Three image models: Nano Banana 2, GPT Image 2.5 Sunburst and Grok Imagine Image 2.0. Two video models: Gemini Omni 1.1 Flash and Grok Imagine Video 1.5. And ElevenLabs Voice Changer, which Voice Swap uses to give a clip your persona's voice.
See all models →Do I need more than one API key?
For images and video, no: Kie, fal, Atlas Cloud, WaveSpeedAI and Krea each run all 5 image and video models on a single key. For voice swap, yes: add an ElevenLabs key.
Which API provider is cheapest?
It depends on the model and its settings. At the lowest published prices, Nano Banana 2 starts at $0.04 per image (1K) on Kie and Grok Imagine Video 1.5 at $0.0225 per second (720p) on Kie. Each model card shows its lowest price, and each provider page compares every model across providers.
Why does Phersa use these models, and not Kling, Seedance or Veo?
Kling, Seedance and Veo are capable models, but short-form creators make a lot of clips, so Phersa keeps a short list picked for top output quality at a price that works at volume. Nano Banana 2 and Gemini Omni 1.1 Flash both run on Google, and new Google Cloud customers get $300 in free credit. Your AI agent writes the script and prompts for every clip, and these models turn them into realistic images and video out of the box. Fewer models means one key and fewer settings.
How do I create an API key?
Sign up with the provider, open its API keys page and create a key. Then paste it in Phersa under Settings → API keys and choose Save & verify. Gemini Enterprise Agent Platform needs a service account JSON, a project and a region instead of a single key. Each provider page has step-by-step instructions.
What happens if I don't add an API key?
You can still browse the library, track accounts, collect videos, and have your AI agent plan a remake clip by clip. When you generate an image, video or voice, Phersa asks you to add a key for that provider first.
Is my API key safe?
Phersa tests your key with the provider before saving it, encrypts it at rest with AES-256-GCM, only ever shows its first four characters, and keeps keys out of logs. You can remove a key at any time in Settings.
Can AI agents use my API keys?
They can use them, but never see them. An AI agent connected over MCP can't read or change your keys. It can generate on them, but every generation starts as a plan with its price and runs only after you say yes.