Pick a model and the dashboard writes the whole build prompt — every endpoint, every parameter, the polling loop, the gotchas. Paste it into whichever AI assistant you use.
Free credits on signup. No card. Usage-based after that.
Around 50 distributed CosyFlow workflows — Z-Image, Qwen Image Edit, Wan 2.2, ChronoEdit, InfiniteTalk dubbing, ACE-Step music, F5-TTS voice, background removal, upscaling — and you can publish your own.
A saved ComfyUI graph called over REST, on the same key and balance as the hosted models rather than on a separate platform.
FLUX.3 video, FLUX.2, Kontext max, Veo 3.1, Seedance 2, Nano Banana Pro, Seedream 5.0 Pro, GPT-Image-2.
Swap between any of them on one line of JSON. No separate vendor contracts, no separate billing relationships, no separate SDKs.
Juggernaut XL, DreamShaper, Pony Diffusion XL, Realistic Vision, AlbedoBase, Animagine, ICBINP and the rest of the ecosystem through AI-Horde.
Seed, steps, CFG, sampler, scheduler, denoise, VAE and checkpoint are all yours to set. Most hosted APIs decide those for you.
Live count from the dashboard, not a roadmap. Text-to-video, image-to-video, first-and-last-frame, video extend, inpainting, multi-image edit, audio-driven dubbing, music and voice.
Live roster from the dashboard. Availability and per-call cost are shown against each model before you call it.
Pick a model in the dashboard and it writes you a complete build prompt — every endpoint, every parameter, the polling loop, the three image-format cases, the known gotchas. Paste it into whichever assistant you use — ChatGPT, Claude, Gemini, Copilot, Cursor — and you get a working studio app back. No SDK to learn and no docs to read first.
# 1 — submit curl -X POST https://api.promptus.ai/api/generate \ -H "Authorization: Bearer$PROMPTUS_API_KEY " \ -d '{ "prompt": { "cosyflow":"(cosy) Z-Image: Text to Image" , "text_0":"neon alley, rain, 35mm" , "width": 1024, "height": 1024, "steps": 25 }, "accountNo":"$PROMPTUS_ACCOUNT" }' # -> { "uid": "job_abc123", "error": null } # 2 — poll every 2s GET /api/generate/progress?uid=job_abc123 # -> { "progress": 100, "output": [{ "uid": "out_xyz" }] } # 3 — retrieve GET /api/generate/output/job_abc123/out_xyz
Submit, poll, retrieve. Switching from Z-Image to Veo 3.1 is a change of one string — not a new integration.
Seed, steps, CFG, sampler, scheduler, denoise, VAE and checkpoint are all yours to set. Most hosted APIs decide these for you.
One balance spends across every provider. No per-model subscription, no minimum, no seat licence.
| Plan | Credits | Best for |
|---|---|---|
| Free | Included on signup | Testing a model, prototyping an agent tool |
| Pay as you go | Top up any amount | Production apps — a Veo 3.1 call and a Z-Image call cost different amounts, and you only pay the difference |
| Volume | Committed, discounted | Platforms embedding generation at scale, and white-label partners |
Per-call cost varies by model, resolution and duration — a 10-second Veo 3.1 video is not priced like a 1024×1024 Z-Image. Live rates are shown per model in the dashboard before you call.
No. Call a CosyFlow by name like any other model. Knowing ComfyUI just means you can publish your own graphs as endpoints too.
Replicate and fal host open models and can run ComfyUI graphs; the closed labs host their own. What is different here is the union on one key and one balance — graphs you publish, the hosted frontier models, and ~130 open checkpoints — plus a dashboard that writes the integration for you. If you only need one of those three, they are fine choices.
Each provider keeps its own policy — a call to a hosted model is subject to that model's filter. Open checkpoints do not add one. Illegal material is prohibited on every route regardless.
Yes, on the volume plan. That is a conversation, not a checkbox — talk to us.
Yes. One login, one credit balance across the desktop app, the web app and the API.
Generate your first image in about two minutes — or paste one prompt into your AI assistant and have it build you the whole app.
Get an API key