WaveSpeed API Setup
Connect WaveSpeed to unlock additional AI video models in Cuppa, starting with Higgsfield Director-of-Photography (DoP) image-to-video. WaveSpeed is an aggregator that hosts hundreds of third-party models behind a single, fast API.
What is WaveSpeed?
WaveSpeed AI is a unified API for running 700+ open and proprietary AI models - image, video, and audio. Cuppa uses WaveSpeed as the gateway to Higgsfield DoP, a video model that turns a still image into a cinematic 5-second clip with director-style camera moves (orbit, dolly, parallax push, vertigo zoom, anamorphic flare).
You bring your own WaveSpeed API key - Cuppa adds zero markup on generation costs.
What You Get
| Model | Mode | Length | Resolution | Best For |
|---|---|---|---|---|
Higgsfield DoP (wavespeed:higgsfield-dop-i2v) | Image-to-video | 5s | 720p | Cinematic camera work on a still you already love |
Wan 2.7 (wavespeed:wan-2.7) | Text or image | 5-15s | 720p/1080p | Cinematic clips from prompt or animate a reference photo |
Wan 2.7 uses text-to-video by default; pass
start_image_urlto switch to image-to-video (optionalend_image_urlfor start/end frame control).
Plan Availability
| Feature | Solo | Studio | Scale | Enterprise |
|---|---|---|---|---|
| Video Generation | ❌ | ✅ | ✅ | ✅ |
| Higgsfield DoP via WaveSpeed | ❌ | ✅ | ✅ | ✅ |
| Video Script Generator | ❌ | ✅ | ✅ | ✅ |
| Wan 2.7 via WaveSpeed | ❌ | ✅ | ✅ | ✅ |
Video generation requires a Studio or higher plan.
Legacy plans: Business, Agency, and Agency+ include the same access (mapped to Studio, Scale, and Enterprise).
Step 1: Create a WaveSpeed Account
- Go to wavespeed.ai
- Click Sign up
- Sign in with Google, GitHub, or email
Step 2: Get Your API Key
- Go to wavespeed.ai/accesskey
- Click Create new key
- Name it “Cuppa”
- Copy the key
Your key looks like: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Step 3: Add Billing
WaveSpeed bills per generation. Higgsfield DoP runs roughly 0.30 per 5-second clip depending on settings.
- Go to wavespeed.ai/billing
- Add a payment method or top up credits
Step 4: Add the Key to Cuppa
- Go to Team Settings → API Keys
- Find the WaveSpeed section
- Paste your API key
- Click Save - Cuppa pings WaveSpeed to validate the key immediately
You can store multiple keys (e.g. one per client). Cuppa picks the right key automatically based on the model you choose.
Generating Videos with Higgsfield DoP
In the Web App
- Go to Creative Studio → Video Generation
- Open the model selector (bottom-left of the prompt area)
- Pick Higgsfield DoP (via WaveSpeed)
- Drop in or pick a Start Image from your Media Library - DoP is image-to-video, so this is required
- (Optional) Pick an End Image to lock in a transition
- Write a director-style prompt (see prompt tips below)
- Hit Generate - typical render is 1–3 minutes
Via the CLI
cuppa video create
--prompt "Slow vertigo zoom revealing the city skyline behind, anamorphic flare"
--model wavespeed:higgsfield-dop-i2v
--start-image-url "https://your-cdn.com/hero.jpg"
--duration 5
--aspect-ratio 16:9Then poll status:
cuppa video status <video-id>Via MCP (Claude / ChatGPT / Cursor)
Tell your agent:
Use
create_videowithmodel: "wavespeed:higgsfield-dop-i2v",start_image_url: "https://…",prompt: "Orbit camera around the subject, soft volumetric light", then pollget_video_statusuntil complete.
Via REST API
curl -X POST https://app.cuppa.ai/v1/video
-H "Authorization: Bearer $CUPPA_API_KEY"
-H "Content-Type: application/json"
-d '{
"model": "wavespeed:higgsfield-dop-i2v",
"prompt": "Anamorphic dolly-in on the steaming espresso, warm golden hour lighting",
"start_image_url": "https://your-cdn.com/hero.jpg",
"duration_seconds": 5,
"aspect_ratio": "16:9"
}'Writing Good Higgsfield DoP Prompts
DoP is camera-led, not scene-led. Your image is the subject - your prompt is the director.
Structure
[Camera move] + [Subject focus] + [Lighting/mood] + [Optional environmental motion]What works
- Camera verbs: orbit, dolly in/out, pull back, arc left/right, parallax push, vertigo zoom, crane up/down
- Lens language: anamorphic flare, shallow depth of field, soft volumetric light
- Subtle environmental motion: drifting steam, swaying foliage, golden-hour shimmer, light shifts
Examples
Product shot → cinematic ad:
Slow dolly-in on the bottle, warm cinematic golden hour, subtle steam drifting upwardPortrait → editorial reveal:
Slow orbit around the subject, soft volumetric light from the left, shallow depth of fieldLandscape → trailer beat:
Vertigo zoom revealing the mountain range behind, slow drifting clouds, anamorphic flare on the sunWhat to avoid
- No multiple shots - DoP is one continuous camera move
- No dialogue or sound effects - DoP is silent
- No scene changes - the still is your subject
- No technical render terms (“4k 60fps blender octane”) - describe what the camera and world should do
Brand DNA Integration
When you have Brand DNA configured for the selected site, Cuppa automatically enriches your prompt with brand colors and visual style - exactly the same way it does for Replicate models. Your prompt stays clean while the generated video reflects your brand.
Troubleshooting
”WaveSpeed API key required for this video model”
Add a WaveSpeed key in Team Settings → API Keys. Make sure the key is active (validation runs at save time).
”Higgsfield DoP requires a start image”
DoP is image-to-video only. Pick an image from your Media Library or paste a public URL into the Start Image field.
Video stuck on “Creating your video…”
Higgsfield DoP usually completes in 1–3 minutes. If it’s stuck beyond 10 minutes, the job will auto-fail. Check that:
- Your WaveSpeed key has billing/credit
- Your start image URL is publicly reachable (not behind auth)
- The image isn’t blocked by content moderation (try a different shot)
“Failed” generation
Click Edit & Retry in the gallery to try again. Common fixes:
- Simplify the prompt - DoP responds best to one camera move per generation
- Try a different start image (clean composition, clear subject works best)
- Switch aspect ratio if the original image’s framing doesn’t fit
Where do my generated videos live?
Same place as every other model. Generated videos appear in:
- Creative Studio → Video Generation gallery
- Media Library (filter to Videos)
WaveSpeed keeps the source video for 7 days, but Cuppa downloads and re-hosts it on your account immediately, so you keep the file forever.
Cost Reference
WaveSpeed bills per generation. Approximate per-clip cost for Higgsfield DoP at 5s / 720p is 0.30. Check wavespeed.ai/pricing for current rates.
Cuppa adds zero markup - you pay WaveSpeed directly via your API key.
Related Docs
- Video Generation: Full overview of all video models
- Replicate API Setup: For Runway, Seedance, Kling, Veo, Grok, P-Video
- Brand Asset Library: Source images to use as DoP start frames
- Brand DNA: Lock in colors and style across every generation