Pricing & Cost Optimization
Understand how FOTOhub API pricing works and how to minimize costs for your use case.
How Billing Works
The API is prepaid, in US dollars. You top up a wallet, every request is priced in USD and deducted from that balance, and when the balance cannot cover a request the API refuses it with 402 and charges nothing.
Wallet balance ($) − price of each request = remaining balanceThree things follow from that, and they surprise people coming from the web app:
- Credits cannot pay for API usage. Credits belong to a fotohub.app subscription. An account holding 5 000 credits and $0.00 in its wallet gets a 402 on every API call.
GET /v1/billing/balancereturnsbilling_model: "prepaid_wallet_usd"and no credit field at all. - There is no invoice at the end of the month and no overage credit line. You cannot spend money you have not already deposited.
- Prices are the provider's own rate, 1:1. The margin multiplier is
1.0today, and every response that quotes a price also returnsprovider_cost_usdbeside it so you can check that yourself.
Prices are held to 6 decimal places, so a single cheap request can legitimately cost $0.000398.
Reading a price
GET /v1/pricing publishes every rate, and each one carries the unit it is denominated in — you cannot infer the meter from the model name:
| Unit | Applies to |
|---|---|
per_second | most video models |
per_1k_tokens | Seedance video, and chat when quoted per 1K |
per_1m_tokens | chat / LLM |
per_piece | images |
per_1k_chars | text-to-speech |
per_minute | transcription, music, audio |
per_gb_month | storage |
Each entry also carries verified. true means the figure was read off the provider's own price list or invoice; false means it is our recorded copy of their published rate and has not been reconciled against an invoice. GET /v1/pricing/audit lists which models are in which bucket.
Image Generation Costs
Charged per delivered image. Models marked with a range charge per resolution — pass image_size and you are billed that tier, not the highest one.
| Model | 1K | 2K | 4K |
|---|---|---|---|
flux-2-klein-4b | $0.015005 | $0.015005 | $0.015005 |
grok-imagine-image | $0.02 | $0.02 | $0.02 |
minimax-image-01 | $0.03 | $0.03 | $0.03 |
flux-2-pro | $0.03 | $0.075 | $0.255 |
seedream-5-0-260128 | $0.0315 | $0.0315 | $0.0315 |
gemini-3.1-flash-lite-image | $0.0336 | $0.0336 | $0.0336 |
gemini-2.5-flash-image | $0.039 | $0.039 | $0.039 |
flux-1.1-pro | $0.04 | $0.04 | $0.04 |
flux-kontext-pro | $0.04 | $0.04 | $0.04 |
grok-imagine-image-pro | $0.05 | $0.07 | $0.07 |
gemini-3.1-flash-image | $0.067 | $0.101 | $0.151 |
imagen-4-standard | $0.080386 | $0.080386 | $0.080386 |
gemini-3-pro-image | $0.134 | $0.134 | $0.24 |
imagen-4-ultra | $0.160772 | $0.160772 | $0.160772 |
BEST VALUE
seedream-5-0-260128 at $0.0315 delivers premium quality for a third of what the Imagen tier costs. Use it as your default.
If you are rendering at 1K, flux-2-pro is $0.03 — but check the tier column before scaling up: the same model is $0.255 at 4K, an 8.5x step.
Video Generation Costs
Most video models are metered per second of output, so the price of a clip is rate × duration. Durations are snapped to what the provider will actually render before you are billed, so you pay for the clip you receive.
| Model | ID | $/second | 5s clip | Provider |
|---|---|---|---|---|
| Hailuo O2 | hailuo-o2 | $0.017 | $0.085 | MiniMax |
| Wan 2.2 Plus | wan2.2-t2v-plus | $0.02 | $0.10 | Alibaba |
| Kling v2.5 Turbo | kling-v2-5-turbo | $0.026 | $0.13 | Kuaishou |
| Veo 3.1 Lite | veo-3.1-lite-generate-001 | $0.03 | $0.15 | |
| Kling v2.6 | kling-v2-6 | $0.038 | $0.19 | Kuaishou |
| Hailuo 2.3 | hailuo-2.3 | $0.047 | $0.235 | MiniMax |
| Veo 3.1 Fast | veo-3.1-fast-generate-001 | $0.08 | $0.40 | |
| Gemini Omni Flash | gemini-omni-flash | $0.1014 | $0.507 | |
| OpenAI Sora 2 | sora-2 | $0.13 | $0.65 | OpenAI |
| Grok Video 1.5 | grok-imagine-video-1.5 | $0.14 | $0.70 | xAI |
| Google Veo 3.1 | veo-3.1-generate-001 | $0.20 | $1.00 | |
| OpenAI Sora 2 Pro | sora-2-pro | $0.30 / $0.50 / $0.70 | $1.50 at 720p | OpenAI |
sora-2-pro prices per resolution ($0.30 at 720p, $0.50 at 1024p, $0.70 at 1080p). Send resolution and you are billed that tier. Omit it and you get the 1080p rate, so always send it.
See the full catalog for every variant, or GET /v1/pricing?model=<id> for one.
BEST VALUE
hailuo-o2 / wan2.2-t2v-plus are the lowest cost per second with solid quality for social content. Step up to veo-3.1-generate-001 for cinematic output with native audio, or gemini-omni-flash for automatic native audio at half the per-second rate.
Seedance: priced per output token
The Seedance family is the exception. ByteDance bills it per 1000 output tokens, and the token count is a function of resolution and duration:
tokens = floor(tokens_per_frame × (24 × seconds + 1))
tokens_per_frame = frame pixel count / 1024| Model | $/1K tokens | 5s @ 480p | 5s @ 720p | 5s @ 1080p |
|---|---|---|---|---|
seedance-1-5-pro-251215 | $0.0012 | $0.060766 | $0.13068 | $0.29403 |
seedance-2-0-mini | $0.0035 | $0.177233 | $0.38115 | — |
seedance-2-0-fast | $0.0056 | $0.283573 | $0.60984 | — |
seedance-2-0-pro | $0.007 (480p/720p) · $0.0077 (1080p) · $0.004 (4K) | $0.354466 | $0.7623 | $1.886693 |
seedance-2-5 | $0.0107 | $0.518276 | $1.16523 | — |
seedance-1-5-pro-251215 charges $0.0024 per 1K when you ask for generated audio — double the video rate, so a 5-second 720p clip with audio is $0.26136. Every other Seedance tier includes audio at no extra cost.
Only seedance-2-0-pro reaches 4K ($3.9204 for 5 seconds). You do not have to do this arithmetic yourself: GET /v1/pricing?model=seedance-2-5 returns token_formula and worked examples, and POST /v1/billing/estimate prices a specific duration.
Long clips
seedance-2-5 is the only model that reaches 30 seconds in a single request — 648 900 output tokens, $6.94 at 720p. Draft at 480p first: the same 30 seconds is $3.09 there, and a 5-second test is $0.52.
Native audio is included at both resolutions. Attaching a source video makes the render cheaper, not dearer — input frames bill at $0.0064 per 1K instead of $0.0107, so that same 30-second 720p clip is $4.15 as an edit.
Chat / LLM Costs
/v1/ai/chat/completions bills the tokens you actually use. Output tokens cost 4-8x input, so the two are rated separately.
| Model | $/1M in | $/1M out |
|---|---|---|
gemini-flash | $0.30 | $2.50 |
claude-haiku-4.5 | $0.80 | $4.00 |
gemini-pro | $1.25 | $10.00 |
gpt-4o | $3.00 | $15.00 |
claude-sonnet | $3.00 | $15.00 |
A 75-token prompt with a 150-token answer is $0.000398 on gemini-flash.
Pass the alias (gemini-flash, not gemini-2.5-flash) — the aliases are what the rate table is keyed on. For a per-request cost_breakdown, use the premium endpoint (/v1/ai/chat/claude). See the Chat/LLM reference.
Storage Costs
FOTOhub-managed buckets are metered hourly against the bytes you are holding, and the accrual is deducted from the same wallet:
| Rate | |
|---|---|
| Per GB-month | $0.0245 |
| Per GB-hour | $0.00003356 |
Holding 100 GB for a full month costs $2.45. Every storage class prices at the Standard rate, because managed buckets live on S3 Standard in eu-central-1 whatever the class label says. Bring-your-own-bucket destinations are metered at the real per-class rate of the region they are in.
Free Operations
These cost $0.00 and never touch the wallet:
- Gabriel AI classify / suggest / recommend
- Translation (
/v1/ai/translate, up to 10 000 characters per request) - Model catalog queries (
/v1/models,/v1/pricing,/v1/plans) - Billing, balance and usage queries
- Webhook management
They still consume your rate limit.
When a Request Fails
Nothing is charged for a request that never produced output. Two distinct cases:
Empty wallet — 402, before the provider is called, so there is nothing to refund:
{
"error": "insufficient_funds",
"message": "Insufficient funds: this request costs $0.381150 but your balance is $0.120000. Top up your wallet with at least $0.261150 to continue. The FOTOhub API is prepaid: no credits or subscription plan can pay for API usage.",
"required_usd": 0.38115,
"balance_usd": 0.12,
"shortfall_usd": 0.26115,
"charged": false,
"charged_usd": 0,
"topup_url": "https://fotohub.app/console/wallet"
}Read shortfall_usd to size the top-up.
Provider failure after the charge — the wallet is refunded and the error says so explicitly ("No charge was made for this request."). If you do not see that sentence, reconcile against GET /v1/billing/usage rather than assuming.
Cost Optimization Tips
1. Estimate before you spend
POST /v1/billing/estimate prices a batch from the same rate table and the same quote builders that bill it, so the estimate cannot disagree with the charge:
estimate = client.estimate_cost([
{"type": "generate_image", "model": "seedream-5-0-260128", "count": 100},
{"type": "generate_video", "model": "veo-3.1-generate-001", "duration": 8, "count": 20},
{"type": "chat", "model": "gemini-flash",
"input_tokens": 400_000, "output_tokens": 100_000},
])
print(f"${estimate['total_usd']} — affordable: {estimate['sufficient']}")
# $36.1213 — affordable: True
# ($3.7513 for the images + $32.00 for 20 x 8s of Veo 3.1 + $0.37 of chat)Read sufficient rather than comparing numbers yourself: it is the server's own answer, and it degrades honestly when part of the batch has no published rate (priced: false, amount_usd: null — never a misleading 0).
Pass the fields the meter needs: duration for a per-second model, input_tokens/output_tokens for chat, image_size for a resolution-stepped image. An operation that omits them comes back unpriced with a reason instead of a guess.
2. Draft cheap, finish expensive
The spread within a family is large enough that previews should not use the final model:
# Preview: $0.02/s
preview = client.generate_video(prompt="...", model="wan2.2-t2v-plus")
# Final: cinematic quality with native audio, $0.20/s
final = client.generate_video(prompt="...", model="veo-3.1-generate-001")Ten 5-second previews cost $1.00 on Wan, the price of a single 5-second Veo 3.1 clip. The same applies within Seedance: draft at 480p, deliver at 720p.
3. Send the resolution you actually want
Twelve image models and sora-2-pro price per resolution, and omitting the field bills the top tier. A 1K flux-2-pro render is $0.03 with image_size set and $0.255 without it.
4. Use Gabriel AI for model selection
Gabriel is free and picks the cheapest model that can do the job:
route = client.gabriel_classify(prompt="Generate a simple product photo")
print(route["model_selected"])Then price what it chose with GET /v1/pricing?model=<id> before dispatching.
5. Cap your own spending
A wallet is already a hard ceiling, but you can set a lower monthly one — useful for a shared key or an experiment you do not want draining a funded account:
client.set_overage_limit(15) # stop at $15 of spend this calendar monthPass 0 or null to disable it and let the wallet balance be the only cap.
6. Monitor the balance you actually spend from
balance = client.get_balance()
if balance["wallet"]["balance_usd"] < 5:
# Top up, or stop dispatching work
passThere is no credit field to read here. spend.this_month_usd and spend.monthly_limit_usd show the running total against any cap you set.
Account Tiers
Tiers set rate and concurrency limits. They do not include an allowance and they do not change prices — every account pays the same USD rate for the same model and pays it out of its own wallet.
The pay-as-you-go tiers activate automatically from wallet balance or lifetime spend, with no subscription:
| Tier | Unlocks at | RPM | Concurrent jobs | Max upload |
|---|---|---|---|---|
| PAYG Basic | $0 | 30 | 3 | 25 MB |
| PAYG Standard | $25 balance or $50 lifetime spend | 120 | 10 | 100 MB |
| PAYG Premium | $120 balance or $500 lifetime spend | 500 | 30 | 500 MB |
Above that, Enterprise is provisioned by sales:
| Tier | Price | RPM | Concurrent jobs | Support |
|---|---|---|---|---|
| Enterprise | Custom | 5000 | 200 | Custom SLA |
Apply at POST /v1/tiers/enterprise/apply — it cannot be self-served.
Paid API plans were retired on 2026-08-13. There is nothing to subscribe to and nothing to cancel: every tier is free, and it unlocks on your own wallet. So the only way to raise your limits is to top up — which is also the cheaper one, since from $500 up a top-up earns a volume bonus of 5–20% in extra spendable dollars that a monthly fee never gave you.
See Rate Limits for the burst windows and the X-RateLimit-* headers.
Top-Up Packages and the Volume Bonus
Fund the wallet with a package or any custom amount between $10 and $15,000. From $500 up, the top-up earns a volume bonus — extra real dollars credited in the same transaction as the payment, spendable on any operation, with no expiry:
| Slug | You pay | Bonus | Credited to wallet |
|---|---|---|---|
topup-50 | $15 | — | $15 |
topup-100 | $25 | — | $25 |
topup-250 | $60 | — | $60 |
topup-500 | $120 | — | $120 |
scale-500 | $500 | +$25 (5%) | $525 |
scale-1000 | $1,000 | +$100 (10%) | $1,100 |
scale-2000 | $2,000 | +$240 (12%) | $2,240 |
scale-3000 | $3,000 | +$390 (13%) | $3,390 |
scale-5000 | $5,000 | +$750 (15%) | $5,750 |
scale-7500 | $7,500 | +$1,275 (17%) | $8,775 |
scale-10000 | $10,000 | +$1,800 (18%) | $11,800 |
scale-15000 | $15,000 | +$3,000 (20%) | $18,000 |
The bonus is a function of the money, not of the package: typing 2500 into a custom top-up earns the same 13% as any $2,500 preset would. Rungs do not stack — only the highest one at or below your amount applies — and the bonus is floored to the cent, so a $499 top-up earns nothing. Aim at the round numbers.
The four starter slugs are not their amounts
topup-50 charges $15, topup-100 $25, topup-250 $60 and topup-500 $120 — historical names from the pre-USD PLN pricing. The scale-* slugs do match their dollar amounts. Read amount_usd and total_usd from GET /v1/billing/topup/packages rather than hardcoding any of it.
Above $15,000, apply at POST /v1/tiers/enterprise/apply for invoicing. See Billing for the full endpoint contract and the machine-readable bonus ladder.