Character consistency
Lock a face and identity across an entire generation set.
POST /v1/workflows/character-consistencyOperational·99.97% uptime · 30d·p95 2.3sThe human image & video API.
The human image & video API.
Drop-in cheaper.
The human image & video API. Every workflow your AI companion, influencer, or profile-photo product needs — plus discounted access to every major image and video model.
Generate an imagecURL
curl https://api.humanshot.dev/v1/images/generations \
-H "Authorization: Bearer $HUMANSHOT_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "flux-pro-1.1",
"prompt": "studio portrait of a woman, soft rim light",
"n": 4,
"size": "1024x1024"
}'Drop-in for Replicate, fal, OpenAI Images, and Google AI. Same model names, same params, swap the base URL.
Cheaper on every model. Provider price shown next to ours, with no black-box markup.
SLA-backed with multi-provider failover. p95 image latency target 2.3s.
Wraps the providers you already use
ReplicatefalRunPodOpenAIGoogle AIComfyUIBFLIdeogram
50+
image and video models
50+
human-image workflows
99.95%
uptime SLA standard
48h
custom workflow hosting
What you get
One key for every human-image pipeline.
Replace three or four upstream accounts with one. Replace ten in-house deploys with three workflow endpoints.
One unified API
Call Nano Banana, GPT Image, Flux, Imagen, Ideogram, SDXL, Wan, Veo, Kling, and Runway through a single REST endpoint. One key, one bill, same param schema as the upstream provider.
Hosted human workflows
InstantID, PuLID, IP-Adapter, ReActor, character-consistency packs, face-swap, profile-studio, full-body pose packs, fashion try-on, and video lipsync — all production-ready endpoints. Bring your own ComfyUI JSON, hosted in 48 hours.
Volume pricing under cost
First three months at 10% under our cost. After, volume tiers below the upstream price on every model. Provider price shown next to ours on every model row.
Workflows
The endpoints AI-companion, AI-influencer, and profile-photo apps actually ship.
Production-ready human-image pipelines. Same SLA and same volume pricing as the raw model catalog.
InstantID portrait
One-shot face plus style transfer, no training.
POST /v1/workflows/instantidPuLID portrait
High-fidelity identity preservation with prompt control.
POST /v1/workflows/pulidFace swap (ReActor)
Swap a face onto a target image or video.
POST /v1/workflows/face-swapProfile photo studio
LinkedIn-grade headshot pack from a few selfies.
POST /v1/workflows/profile-studioFull-body pose pack
20 full-body poses with locked identity, in one call.
POST /v1/workflows/pose-packCompanion scene pack
Multi-scene character renders for AI companion apps.
POST /v1/workflows/scene-packFashion try-on
Garment on a model, controllable pose and lighting.
POST /v1/workflows/tryonVideo lipsync
Audio-driven mouth and head animation.
POST /v1/workflows/lipsync+ 40 more in the catalog
Don't see what you ship today? Send a ComfyUI JSON to hello@humanshot.dev. We host it within 48h under the same SLA and pricing.
Foundation models
Every major image and video model. Lower price than the provider.
Indicative landing prices. Real rates lock in before launch and live on each model's page.
| Model | Type | Provider | Our price | Provider | Δ |
|---|---|---|---|---|---|
Nano Banana (Gemini 2.5 Flash Image) per 1k images | image | $34 | $39 | −13% | |
GPT Image 1 (low) per 1k images | image | OpenAI | $9 | $11 | −18% |
GPT Image 1 (high) per 1k images | image | OpenAI | $158 | $190 | −17% |
Flux Pro 1.1 per 1k images | image | BFL via Replicate / fal | $36 | $40 | −10% |
Flux Dev per 1k images | image | Replicate / fal | $24 | $30 | −20% |
Flux Schnell per 1k images | image | Replicate / fal | $2.40 | $3.00 | −20% |
Imagen 4 per 1k images | image | $36 | $40 | −10% | |
Ideogram v3 per 1k images | image | Ideogram | $72 | $80 | −10% |
SDXL per 1k images | image | Replicate / RunPod | $1.80 | $2.30 | −22% |
Wan 2.5 per video second | video | fal | $0.36 | $0.40 | −10% |
Veo 3 per video second | video | $0.45 | $0.50 | −10% | |
Kling 2.0 per video second | video | Kling | $0.27 | $0.30 | −10% |
+ 40 more models in the catalog. Pricing parity is enforced — we rate-track every upstream provider weekly and pass changes through.
Drop-in
Replace one line. Keep the rest of your code.
The schema mirrors the upstream provider. If you already call Replicate, fal, or OpenAI Images, point at our base URL with the same params.
api.humanshot.dev/v1
curl https://api.humanshot.dev/v1/images/generations \
-H "Authorization: Bearer $HUMANSHOT_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "flux-pro-1.1",
"prompt": "studio portrait of a woman, soft rim light",
"n": 4,
"size": "1024x1024"
}'Compare
Same providers, better terms.
We sit between you and Replicate, fal, OpenAI Images, Google AI, and the open-source ComfyUI ecosystem. The job is to give you everything Astria and Runninghub do, plus the foundation models, plus a price that beats the upstream list.
| Capability | HumanShot | Replicate | fal | Astria | Runninghub |
|---|---|---|---|---|---|
| Unified image + video API | image only | image only | |||
| Foundation models (Nano Banana, GPT Image, Imagen) | |||||
| Hosted human workflows (InstantID, PuLID, ReActor, ComfyUI) | |||||
| Bring-your-own ComfyUI workflow hosting | 48h | manual | |||
| Pricing below provider list | published | list price | list price | premium | volume only |
| Multi-provider failover | |||||
| Launch credit | $300 | ||||
| Published SLA | 99.95% |
Pricing
First 3 months at 10% under our cost.
New accounts pay 10% below what we pay the upstream provider for any model. After three months, you graduate into volume tiers. Every model on the catalog shows the upstream provider's price next to ours. No black-box markup.
Starter
SLA 99.95%New accounts, first 3 months
10% under our cost
- Every model and workflow
- Pay-as-you-go billing
- Multi-provider failover included
- Prompts and outputs purged after 7 days
Scale
SLA 99.95%$1k+/mo spend
Volume tiers, 5% to 20% under cost
- Volume pricing on all models
- Priority queue and dedicated capacity
- Custom workflows hosted in 48h
- Slack support channel
Enterprise
SLA 99.99%$10k+/mo, custom terms
Custom rate card
- Dedicated capacity per region
- Zero-retention mode
- Single-tenant deployment option
- Solutions engineer on call
Reliability
Built so you can put it on the critical path.
One vendor between you and several upstream providers should improve your uptime, not damage it. We hold ourselves to that.
99.95% standard, 99.99% enterprise
Credits issued automatically when we miss. Status page is the source of truth, not an apology email.
Multi-provider failover
Every model has at least two healthy upstream paths. Example: Flux Pro: Replicate primary → fal fallback.
p95 2.3s on image, 45s on a 5s video clip
Warm pools on the highest-traffic models, region-aware routing. Live numbers on the status page.
Predictable data retention
Prompts and outputs stored 7 days for billing audit, then purged. Opt-out on Enterprise.
Use cases
Built for the products people actually ship.
AI companion apps
Character consistency, multi-scene rendering, video lipsync — at the cost structure that lets you keep paying users instead of providers.
AI influencer studios
Pose packs and identity lock for a single virtual person, plus video clips on Wan, Veo, or Kling — one API for the whole pipeline.
Profile photo SaaS
InstantID and the profile-studio workflow give you LinkedIn-grade output without training a model per user.
Fashion and on-model commerce
Try-on and full-body pose packs on top of Flux and Imagen. Run the catalog without a studio shoot.
FAQ
The questions developers ask first.
How are you cheaper than the providers themselves?
We commit to volume across Replicate, fal, RunPod, OpenAI, and Google, and we negotiate. Most usage on this kind of platform sits on a small number of models, which is exactly the volume that gets discounts. You get the rate card we get, minus a small operating margin, minus another 10% during your first three months.
What about cold starts and queue time?
We keep warm pools on the highest-traffic models and route you to whichever provider is healthiest in your region. The status page shows live p95 latency per model.
Can I bring my own ComfyUI workflow?
Yes. Send the JSON to hello@humanshot.dev and we host it within 48 hours under the same SLA and the same volume pricing as the catalog.
What is your policy on NSFW and identity?
We support adult-oriented apps and require explicit consent for any face used in identity-locked workflows. We block CSAM and impersonation of real public figures by default and surface the controls you need for your own moderation.
How long do you store prompts and outputs?
Seven days for billing audit, then purged. Enterprise can opt into zero-retention with prompts and outputs dropped immediately after generation.
How does billing work?
Single invoice across providers, metered per call. Pay-as-you-go on Starter, monthly commit on Scale, custom on Enterprise. We do not charge for failed generations.
Get started
See exactly what you'd save.
Tell us which models and workflows you run today and your monthly volume. We return a written quote with your savings, the launch credit, and a migration plan — usually within one business day.