Phersa runs Nano Banana 2, GPT Image 2.5 Sunburst, Grok Imagine Image 2.0, Gemini Omni 1.1 Flash and Grok Imagine Video 1.5 on your own fal (fal.ai) API key. fal bills in US dollars, and Phersa adds no markup.
Get your fal API key
fal.ai/dashboard/keys Open | Model | Setting | Price |
|---|---|---|
| Nano Banana 2 Up to 14 reference images | 1K | $0.08 |
| 2K | $0.12 | |
| 4K | $0.16 | |
| GPT Image 2.5 Sunburst Text to image or image to imageUp to 16 reference images | Low · 9:16, 1152×2048 | $0.00471 + inputs |
| Medium · 9:16, 1152×2048 | $0.01101 + inputs | |
| High · 9:16, 1152×2048 | $0.04239 + inputs | |
| Extra high · 9:16, 1152×2048 | $0.07533 + inputs | |
| Max · 9:16, 1152×2048 | $0.1695 + inputs | |
| Grok Imagine Image 2.0 Up to 3 reference images+$0.01 per reference image | 1K · Low | $0.04 |
| 2K · Low | $0.06 | |
| 1K · Medium | $0.06 | |
| 2K · Medium | $0.08 | |
| Gemini Omni 1.1 Flash 3–10 s clipsUp to 10 reference images | 720p · 8 s clip | $0.80 ($0.10/s × 8s) |
| 1080p · 8 s clip | $1.20 ($0.15/s × 8s) | |
| 4K · 8 s clip | $2.40 ($0.30/s × 8s) | |
| Grok Imagine Video 1.5 1–15 s clipsUp to 7 reference images+$0.01 per reference image | 720p · 8 s clip | $1.12 ($0.14/s × 8s) |
| 1080p · 8 s clip | $2.00 ($0.25/s × 8s) |
Each model's price on fal next to the cheapest other providers Phersa supports. Video is priced for an 8-second clip without reference images.
| Provider | 1K | 2K | 4K |
|---|---|---|---|
| fal | $0.08 | $0.12 | $0.16 |
| Agent Platform and Google AI Studio | $0.067 | $0.101 | $0.151 |
| Kie | 8 creditsLowest $0.04 | 12 creditsLowest $0.06 | 18 creditsLowest $0.09 |
| WaveSpeedAI | $0.07 | $0.105 | $0.14 |
Kie is cheapest at 1K, 2K and 4K: $0.04, $0.06 and $0.09, vs $0.08, $0.12 and $0.16 on fal.
Also on Atlas Cloud and Krea, at the same or higher prices.
| Provider | Low | Medium | High | Extra high | Max |
|---|---|---|---|---|---|
| fal | $0.00471 + inputs 1152×2048 | $0.01101 + inputs 1152×2048 | $0.04239 + inputs 1152×2048 | $0.07533 + inputs 1152×2048 | $0.1695 + inputs 1152×2048 |
| Atlas Cloud | $0.00377 + inputsLowest 1152×2048 | $0.00881 + inputsLowest 1152×2048 | $0.03391 + inputsLowest 1152×2048 | $0.06026 + inputsLowest 1152×2048 | $0.1356 + inputsLowest 1152×2048 |
Atlas Cloud is cheapest at Low, Medium, High, Extra high and Max: $0.00377–$0.1356, vs $0.00471–$0.1695 on fal.
Kie and WaveSpeedAI price GPT Image 2.5 Sunburst by other settings: from $0.03 on Kie and from $0.01 on WaveSpeedAI. fal bills GPT Image 2.5 Sunburst by tokens at OpenAI's rates. Prices shown are the output image at 9:16 (1152×2048); the prompt and each reference image add input tokens, about $0.008 for a 1024×1024 reference by fal's measurements. Atlas Cloud bills GPT Image 2.5 Sunburst by tokens at 80% of OpenAI's rates. Prices shown are the output image at 9:16 (1152×2048); the prompt and each reference image add input tokens.
| Provider | 1K · Low | 2K · Low | 1K · Medium | 2K · Medium | 1K |
|---|---|---|---|---|---|
| fal | $0.04Lowest | $0.06 | $0.06 | $0.08 | Not offered |
| Kie | Not offered | Not offered | Not offered | Not offered | 4 credits $0.02 |
| Atlas Cloud and SpaceXAI | $0.04Lowest | $0.06 | $0.06 | $0.08 | Not offered |
| WaveSpeedAI | $0.05 | $0.05Lowest | $0.05Lowest | $0.05Lowest | Not offered |
fal is tied for lowest at 1K · Low. WaveSpeedAI is cheapest at 2K · Low, 1K · Medium and 2K · Medium: $0.05 each, vs $0.06–$0.08 on fal. The lowest price overall is 1K on Kie, at $0.02.
Also on Krea, at the same or higher prices.
fal, Atlas Cloud, SpaceXAI and WaveSpeedAI add $0.01 per reference image.
| Provider | 720p · 8 s | 1080p · 8 s | 4K · 8 s |
|---|---|---|---|
| fal | $0.80 $0.10/s × 8s | $1.20 $0.15/s × 8s | $2.40 $0.30/s × 8s |
| Kie | 105 creditsLowest $0.525 | 105 creditsLowest $0.525 | 189 creditsLowest $0.945 |
| Atlas Cloud | $0.792 $0.099/s × 8s | $1.152 $0.144/s × 8s | $2.232 $0.279/s × 8s |
Kie is cheapest at 720p, 1080p and 4K for an 8-second clip: $0.525–$0.945, vs $0.80, $1.20 and $2.40 on fal.
Also on Agent Platform, Google AI Studio, WaveSpeedAI and Krea, at the same or higher prices.
| Provider | 720p · 8 s | 1080p · 8 s |
|---|---|---|
| fal | $1.12 $0.14/s × 8s | $2.00 $0.25/s × 8s |
| Kie | 36 creditsLowest 4.5/s × 8s · $0.18 | 64 creditsLowest 8/s × 8s · $0.32 |
| Atlas Cloud | $0.64 $0.08/s × 8s | $0.64 $0.08/s × 8s |
Kie is cheapest at 720p and 1080p for an 8-second clip: $0.18 and $0.32, vs $1.12 and $2.00 on fal.
Also on WaveSpeedAI, Krea and SpaceXAI, at the same or higher prices.
fal adds $0.01 per reference image. Kie lists no 1080p rate for Grok Imagine Video 1.5; the Kie 1080p price shown is its rate for the sibling Grok Imagine model.
Sign in at fal.ai, open the API keys page and create a key. fal takes generation costs from your fal balance, so add funds to your fal account too. Choose an Admin-scope key if you want to see that balance in Phersa.
fal.ai/dashboard/keys →In Phersa, open Settings, go to API keys and choose fal. Paste the key and click Save & verify.
Before saving, Phersa tests the key with fal, and a key fal rejects is not stored. Saved keys are encrypted, and with an Admin-scope key the fal panel in Settings shows your balance in dollars.
In the app, pick a model that runs on fal. You see the cost in dollars before you generate, such as $0.08 for a 1K Nano Banana 2 image.
fal (fal.ai) is a generative media platform for developers that runs image, video, audio and 3D models through an API. It bills per output from prepaid USD credits, which expire 365 days after purchase. fal was founded in 2021 by Burkay Gur and Gorkem Yurtseven in San Francisco.
Visit fal.ai ↗On fal, Nano Banana 2 costs $0.08 for a 1K image, $0.12 for 2K and $0.16 for 4K.
No. Generations bill to your own fal balance at fal's price, and Phersa adds no markup. The Phersa plan is a separate $99 per month.
See Phersa pricing →No published sign-up credit. fal's pricing page mentions a free tier without an amount, and credits come through programs such as the partner-referred Builder Grant ($25–$250) and the Startup Program ($1,000–$5,000). Terms can change.
Check fal.ai pricing ↗Images: Nano Banana 2, GPT Image 2.5 Sunburst and Grok Imagine Image 2.0. Video: Gemini Omni 1.1 Flash and Grok Imagine Video 1.5. Voice Swap runs on ElevenLabs Voice Changer, so it needs an ElevenLabs key.
Voice Swap with ElevenLabs →Create a key at fal.ai/dashboard/keys. In Phersa, open Settings, go to API keys, choose fal and paste it. Phersa checks the key with fal before saving it.
See the setup steps →fal bills it by tokens. A 9:16 image (1152×2048) costs $0.00471 at Low, $0.01101 at Medium, $0.04239 at High, $0.07533 at Extra high and $0.1695 at Max for the output, plus input tokens for the prompt and each reference image (about $0.008 per 1024×1024 reference). Larger sizes and higher quality use more tokens. fal and Kie do not share a size setting.
See Kie prices →fal only shares the balance with Admin-scope keys. An API-scope key still runs every model, but Phersa cannot read the balance with it. Use an Admin-scope key to see it in the fal panel in Settings.
fal refuses the job, and Phersa shows the error, usually “Out of credits”. Add funds to your fal account, then generate again.
Phersa tests the key with fal before saving it, stores it encrypted and uses it for generation and your balance. AI agents connected over MCP cannot read or change it. Removing the key in Settings deletes it permanently.
