eroq api · public beta
Every medium,
one key.
Roleplay chat, image, video, voice and music behind one OpenAI-shaped API, metered in flat credits. 16 named video engines, 13 image models, voices you can clone, songs with vocals — and a written policy where everyone else has a filter.
Private and uncensored · a written policy, not a filter
40 free credits on signup · no card required
curl https://eroq.ai/v1/chat/completions \
-H "Authorization: Bearer $EROQ_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "eroq-rp-plus",
"context": "Mira: sardonic starship mechanic, dry humor, hates small talk. Scene: engine bay, mid-shift.",
"messages": [
{
"role": "user",
"content": "The reactor is making that noise again."
}
]
}'35
models, one key
1 cr
cheapest completion
SSE
token streaming
0
token math
Partnering with global industry leaders to power your creative output
The catalogue
35 models, one store. Flat prices.
eroq RP+
chateroq-rp-plus
Our production roleplay engine: stays in character across long scenes, tuned sampling against repetition and filler, streams over SSE. The model behind millions of conversations on our consumer network.
3 credits / completion
eroq RP mini
chateroq-rp-mini
The economical tier for high-volume products. Same API shape and same persona handling as eroq RP+, sized for latency and cost.
1 credit / completion
eroq Voice One
speecheroq-voice-one
Natural speech synthesis with a curated voice roster. Returns MP3 audio directly.
3 credits / 100 characters
eroq Voice Turbo
speecheroq-voice-turbo
The high-volume voice tier: lower latency and a lower price, same roster and cloned voices as eroq Voice One.
2 credits / 100 characters
Scribe One
transcriptioneroq-scribe-one
Transcription with automatic language detection, built for voice-message-length audio.
1 credit / minute
eroq Music One
musiceroq-music-one
Describe a song, or bring your own lyrics and style: Music One writes the arrangement and sings it. Verses, choruses and a bridge, vocals and instruments, up to four minutes, MP3 out.
3 credits / 30 s of song
Eroq Uncensored
imageeroq-uncensored
eroq's own uncensored photoreal model: the scene renders as written within the acceptable-use policy, with CFG, in-place editing and up to four reference pictures, fully private.
10 credits / image
Krea2
imageeroq-krea2
Krea 2 is the studio's clean photoreal base: fast turbo sampling, CFG control, up to four reference pictures, in-place editing and a warm render in about fifteen seconds — the everyday character image engine.
10 credits / image
Eroq One
imageeroq-one
Eroq One is the reference engine: generation and plain-language instruction editing from up to ten reference pictures, for consistent characters, scenes and products.
10 credits / image, +1 per extra picture
Anima
imageeroq-anime
Anima renders anime and stylized illustration, uncensored within policy — a two-pass hi-res detail render with CFG and negative-prompt control.
10 credits / image
eroq Motion One
videoeroq-motion-one
eroq’s own video engine: 5 to 15 second cinematic clips with native audio baked into the picture — uncensored within the acceptable-use policy, private, and priced per second.
29 credits / second
Motion Pro
videoeroq-motion-pro
Motion Pro: 5, 10 or 15 second clips at 720p or 1080p, uncensored within the acceptable-use policy, from a prompt or a start frame — fully private, on every account.
30 credits / second
Seedance 2.5
videoseedance-2-5
ByteDance’s newest Seedance: clips from 4 to 30 seconds with native sound, rendered at up to 720p and taken to 1080p or 2K by our upscaler. Your characters and elements ride as references (up to six pictures, their look sheets) — a character’s voice sample too — and a reference video can set the motion or be extended. First/last-frame control and seeds. Moderated upstream by its provider.
98 credits / second
Veo 3.1
videoveo-3-1
Veo 3.1 renders 4, 6 or 8 seconds at 720p, 1080p or 4K with dialogue, ambience and effects generated with the picture, up to three reference pictures for identity, and first/last-frame control. Moderated upstream by its provider.
169 credits / second
Kling 3.0 Pro
videokling-3-pro
Kuaishou’s premium Kling 3.0: remarkably fluid movement and strong faces, 3 to 15 seconds at 720p with optional sound, up to four reference pictures, first/last frames and a native negative prompt. Moderated upstream by its provider.
71 credits / second
Seedance 2.0
videoseedance-2-0
Seedance 2.0 holds a cast together from your references (up to six pictures, plus a character’s voice sample), renders 4 to 15 seconds at up to 4K with native sound, and takes first/last frames and seeds. Moderated upstream by its provider.
64 credits / second
Hailuo 3
videohailuo-3
MiniMax’s Hailuo 3 renders 5 to 15 seconds at 2K with sound, convincing physics and camera language, up to three subject references and a character’s voice sample, first/last frames. Moderated upstream by its provider.
55 credits / second
Runway Gen-4.5
videorunway-gen-4-5
Runway’s Gen-4.5: strong motion quality and prompt adherence, 2 to 10 seconds at 720p, image-to-video from a start frame, seeds. No soundtrack. Moderated upstream by its provider.
51 credits / second
Veo 3.1 Fast
videoveo-3-1-fast
The fast Veo 3.1 tier: the same 4 to 8 second clips with native audio, reference pictures and first/last-frame control, rendered quicker and for far less. Moderated upstream by its provider.
43 credits / second
Kling 3.0
videokling-3
Kling 3.0 Standard: the fluid motion and controls of Pro — references, first/last frames, negative prompt, optional sound — at a lower price. Moderated upstream by its provider.
54 credits / second
Seedance 2.0 Fast
videoseedance-2-0-fast
The speed tier of Seedance 2.0: the same references, voice sample, native sound and frame control at 480p or 720p — for drafts and volume. Moderated upstream by its provider.
39 credits / second
Hailuo 3 Max
videohailuo-3-max
The fast Hailuo 3 tier: 5 to 15 second clips at 480p or 768p, image-to-video with first/last frames, no soundtrack. Moderated upstream by its provider.
34 credits / second
Grok Imagine
videogrok-imagine
Grok Imagine Video 1.5: quick 1 to 15 second clips at up to 1080p from a prompt or a start image. Moderated upstream by its provider.
60 credits / second
Seedance 2.0 Mini
videoseedance-2-0-mini
Seedance 2.0 Mini is the entry engine: 4 to 15 seconds at 480p or 720p with sound, reference pictures, a voice sample and frame control. Moderated upstream by its provider.
15 credits / second
Veo 3.1 Lite
videoveo-3-1-lite
Google’s most economical Veo: 4 to 8 seconds at 720p or 1080p with native audio and first/last-frame control — no reference pictures. Moderated upstream by its provider.
22 credits / second
Wan 3.0
videowan-3
Alibaba’s Wan 3.0: 2 to 30 seconds at up to 1080p with sound, reference-guided generation and image-to-video, seed control — from the Hobby plan. Moderated upstream by its provider.
43 credits / second
Flash Image
imageeroq-flash-image
Flash Image is the fast multimodal engine: broad style range and batch support.
10 credits / image
Flux Image
imageeroq-flux-image
Flux Image is the high-fidelity photoreal engine: sharp detail and natural light.
12 credits / image
GPT Image 2.5 Sunburst
imagegpt-image-2-5-sunburst
GPT Image 2.5 Sunburst is OpenAI's flagship image model: strong prompt adherence, clean composition and the sharpest in-image text of the lineup. Text-to-image, five formats, from the Creator plan.
31 credits / image
GPT Image 2.5 Flare
imagegpt-image-2-5-flare
GPT Image 2.5 Flare is OpenAI's expressive image model: the same precision as Sunburst with a bolder, more dramatic rendering for posters, key art and cinematic stills. Text-to-image, five formats, from the Creator plan.
31 credits / image
Nano Banana 2.1
imagenano-banana-2-1
Nano Banana 2.1 is Google's fast multimodal image model: quick, style-flexible renders with strong natural-language understanding. Text-to-image, five formats, from the Creator plan.
31 credits / image
Seedream 5.0 Pro
imageseedream-5-0-pro
Seedream 5.0 Pro is ByteDance's flagship image model: dense detail, rich texture and strong photographic realism. Text-to-image, five formats, from the Creator plan.
36 credits / image
Nano Banana Pro
imagenano-banana-pro
Nano Banana Pro is Google's highest-fidelity image model: the sharpest detail, the most accurate text and the best prompt understanding of the Nano Banana line. Text-to-image, five formats, from the Creator plan.
124 credits / image
Flux 3 Image
imageflux-3-image
Flux 3 Image is Black Forest Labs' latest FLUX: natural photoreal rendering, strong lighting and reliable in-image text. Text-to-image, five formats, from the Hobby plan.
32 credits / image
Grok Imagine 2.0
imagegrok-imagine-2-0
Grok Imagine 2.0 is xAI's image model: quick, characterful renders with a distinct look. Text-to-image, five formats, from the Hobby plan.
32 credits / image
Store
storageeroq-store
Object storage on a global CDN, under the same written policy as the engines: host what you generate next to where you generate it. Included in your plan's storage, served forever, fair use included.
Free within your plan's storage, then 2 credits / 10 MB
Why eroq
Built for character products, not chatbots.
One key, every medium
Chat, images, video, speech, music and transcription answer to the same key and spend from the same balance. No per-service provisioning, no second invoice to reconcile, one error envelope to handle.
Private and uncensored
Fiction renders instead of refusing — dark, violent or provocative themes included — on eroq's own engines, under a written acceptable-use policy (no minors, no real people, ever) you can quote in your own terms. Send private: true and an image or clip is returned once and kept nowhere.
Credits, not invoices
Prepaid credits with flat per-call costs. An image on a given model costs the same every time, from 10 credits — no token math, no surprise bill — and a generation that fails refunds itself into your usage ledger.
Engines you can name
Video models are published by name with their prices, lengths and capabilities, because the engine is the decision. The engines endpoint reports what each one accepts, so you can branch on capability instead of guessing.
Streaming and webhooks
Chat streams over server-sent events with the standard chunk format. Video renders are jobs — take a signed webhook when one finishes rather than holding a request open for a minute.
Docs that cannot drift
Reference pages, openapi.json and llms.txt are generated from the same spec the endpoints charge from. If the docs say 3 credits, the meter says 3 credits. An MCP server and a CLI read the same surface.
Pricing
Top up. Build. Repeat.
Starter
$10
1,000 credits
Try every model properly.
Plus
+10%$25
2,750 credits
A real week of creating, bonus included.
Builder
+15%$50
5,750 credits
Enough to ship a beta.
Scale
+20%$100
12,000 credits
Production traffic, best rate.
Prefer a subscription? Monthly plans from $15 carry credits that roll over and 2–8× rate limits.
FAQ
The short version.
What is a credit worth? +
Packs start at $10 for 1,000 credits — about a cent each, with bonus credits on larger packs. A chat completion costs 1–3 credits, an image 10 to 124, a five-second video clip between 75 and 845 depending on the engine, and a song 3 credits per started 30 seconds, per take.
Is the API OpenAI-compatible? +
The request and response shapes follow the OpenAI convention (chat/completions, images/generations, audio/speech), so most client code ports by changing the base URL, key and model id. We are not affiliated with OpenAI.
Do you keep what I generate? +
Images, clips and speech are saved to your own creation library with the prompt and settings that produced them — that is what makes a later remix possible. It is private to your workspace, never published unless you publish it, and you can delete any of it. Chat completions are not stored.
What happens when a generation fails? +
The call is refunded automatically and the refund appears in your usage ledger. You only pay for output, and in a batch each failed take is refunded individually.
Is the API uncensored? +
Yes — that is the point. Chat, images and voice are uncensored within a written acceptable-use policy, and so is Motion One for video; the other video engines are moderated by their providers. The hard limits are absolute: no minors, no real-person impersonation, nothing illegal.
Are there embeddings or tool calling? +
No. There is no embeddings endpoint and no tool-calling surface. Keep a mainstream API for retrieval, agents and structured tool use, and route the content work here.
From the engine room.
all articles →Oct 23, 2026 · 6 min
Halloween AI trends — the costume, pumpkin head and zombie ideas to make this week
The Halloween AI trends people are making right now, from the costume try-on to the pumpkin head photoshoot and the zombie love story, each with a prompt that opens in the studio.
Oct 19, 2026 · 10 min
Best Higgsfield alternatives in 2026 — 7 tools compared
Seven Higgsfield alternatives for short-form video and ads in 2026: multi-engine studios and the engine makers' own apps, with prices, policies and limits.
Oct 19, 2026 · 11 min
Best Krea AI alternatives in 2026 — 7 tools compared
Krea AI alternatives compared on what Krea is known for — named image and video engines, real-time generation, upscaling — plus policy, privacy and price.
First call in five minutes.
Sign up, create a key, paste the quickstart. 40 free credits to start, no card needed.