Seedance on Felo OpenAPI: Four Video Models, One Endpoint
Felo OpenAPI now supports video generation. Four ByteDance Seedance models, per-second pricing, and the same API key you already use.
Your app can draft the script, write the shot list, and summarize the review notes in one API call. But making the actual video meant leaving the stack: a second provider, a second key, a second billing dashboard, and a different async pattern to babysit.
That gap is closed. Felo OpenAPI now supports video generation. Four ByteDance Seedance models run on the same platform as the LLM routes you already call, behind the same base URL and the same FELO_API_KEY. The lineup runs from the cheapest route for high-volume work to Seedance 2.5, ByteDance's newest model, which renders 30-second reference-guided clips with audio.
Try Seedance 2.5 on Felo OpenAPI: https://openapi.felo.ai/models/bytedance/seedance-2.5
The four models at a glance
Felo API Platform lists 27 models. Four of them generate video, and they split the work by what you optimize for: length, resolution, speed, or cost.
| Model | Best for | Resolution | Clip length | From |
|---|---|---|---|---|
| Seedance 2.5 | Long-form storytelling, video editing, video extension, multilingual audio | 480p / 720p | 4–30s | $0.1028/sec |
| Seedance 2.0 | Flagship quality: character consistency, style transfer, 4K output | 480p – 4K | 4–15s | $0.06725/sec |
| Seedance 2.0 Fast | Fast iteration and draft renders | 480p / 720p | 4–15s | $0.04035/sec |
| Seedance 2.0 Mini | High-volume generation and social clips | 480p / 720p | 4–15s | $0.03363/sec |
Here is how to think about each one.
Seedance 2.5 — the long-form route
Seedance 2.5 is ByteDance's newest Seedance model. It renders clips up to 30 seconds in a single pass and accepts up to 50 image, video, and audio reference assets in one request — enough to pin every recurring character, product, or set element to its own reference and keep them consistent across a long take.
It also goes beyond generation: it handles video editing (fix a region instead of regenerating the whole clip) and video extension (continue a clip past its last frame or bridge two shots). Audio is generated alongside the picture, with multilingual support, which matters when one campaign video has to run in several markets.
Reach for it when: the story needs more than a few seconds, when references need to hold a whole cast in place, or when you need to revise a clip rather than reshoot it.
Seedance 2.0 — the 4K flagship
Seedance 2.0 is ByteDance's flagship route and the only Seedance model here that renders up to 4K. It is strongest at preserving character consistency, visual style, and camera movement from reference material: the face, the product, and the framing carry over from the images you feed it.
Reach for it when: the output has to survive a big screen or aggressive cropping — product films, brand spots, anything that ships at delivery resolution.
Seedance 2.0 Fast — the iteration route
Seedance 2.0 Fast trades maximum quality for speed and cost. It renders up to 720p and is built for the part of the workflow where you explore directions rather than finish them.
Reach for it when: you want five takes of the same shot before committing to one, or when a rough cut is good enough for review.
Seedance 2.0 Mini — the volume route
Seedance 2.0 Mini is the lowest-cost Seedance route on the platform, built for everyday generation at scale: social clips, shop variations, A/B creative, and anything where per-unit cost decides the budget.
Reach for it when: volume is the point and 720p is enough.
How the video API works
Every model on the platform uses one endpoint and one pattern: submit a task, poll it, download the result.
POST https://openapi.felo.ai/api/v1/videos/generations
GET https://openapi.felo.ai/api/v1/videos/generations/$TASK_ID
Video generation takes longer than a chat completion, so the API is asynchronous by design. You submit a task with your prompt and parameters, poll until the status reads succeeded, and download the signed output URL. There is nothing to stitch together and no webhook to wire up before you can ship a first version.
Three generation routes are available across the family:
- Text to video — describe the shot and render it.
- First and last frame — supply the opening and closing frames so the clip lands exactly where your edit needs it.
- Multimodal reference — give the model images, video, or audio as reference material, which is how you keep characters and products consistent from clip to clip.
Aspect ratios cover the standard set: 16:9, 9:16, 1:1, 4:3, and 3:4, plus 21:9. The 2.0 routes also offer 9:21. Audio is included: ask for generate_audio and the model returns a video with a matching audio track, at no extra rate.
Your first video request
Step 1: Get an API key
Create a key on the platform's API keys page and export it. Keys are shown once, so save it where your backend can read it, not in client-side code.
export FELO_API_KEY="YOUR_API_KEY"
Step 2: Submit a video task
Set the model and parameters, and send the request:
curl -X POST "https://openapi.felo.ai/api/v1/videos/generations" \
-H "Authorization: Bearer $FELO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance-2.5",
"prompt": "a cinematic drone shot over a coastal city at sunrise",
"resolution": "720p",
"duration": 5,
"aspect_ratio": "16:9",
"generate_audio": true
}'
The response includes a task ID. Swap seedance-2.5 for seedance-2.0, seedance-2.0-fast, or seedance-2.0-mini to route the same request to a different model.
Step 3: Poll the task
curl "https://openapi.felo.ai/api/v1/videos/generations/$TASK_ID" \
-H "Authorization: Bearer $FELO_API_KEY"
Step 4: Download the output
Once the status is succeeded, the response carries a signed output URL. Fetch it and save the MP4:
curl -o output.mp4 "$OUTPUT_URL"
If you would rather click than code, every model page has a playground that sends a real request with your own API key and watches the task until the signed URL is ready. It is a fast way to compare the four models on the same prompt before you wire anything into production.
What it costs
Video routes are billed per second of output, which makes estimates straightforward: a 5-second clip costs five times the per-second rate for its model and resolution.
| Model | 480p | 720p | 1080p | 4K |
|---|---|---|---|---|
| Seedance 2.5 | $0.1028 | $0.2311 | — | — |
| Seedance 2.0 | $0.06725 | $0.1512 | $0.3742 | $0.7776 |
| Seedance 2.0 Fast | $0.04035 | $0.09072 | — | — |
| Seedance 2.0 Mini | $0.03363 | $0.0756 | — | — |
All rates are per second of output. Two things move the number:
- Audio is free. Turning on
generate_audiodoes not change the rate. - Video references get a discount. Passing a video as reference material uses a lower rate, roughly 40% below the standard one: Seedance 2.5 at 720p drops from $0.2311 to $0.1382 per second.
Put together, a few reference points:
| Clip | Cost |
|---|---|
| 5s, 720p, Seedance 2.0 Mini | $0.38 |
| 5s, 720p, Seedance 2.0 | $0.76 |
| 30s, 720p, Seedance 2.5 | $6.93 |
| 15s, 4K, Seedance 2.0 | $11.66 |
Which model should you pick?
| If you need... | Use |
|---|---|
| A 30-second narrative clip, dialogue, or a localized version per market | Seedance 2.5 |
| The same character or product to look identical across many clips | Seedance 2.0 (or 2.5 with references) |
| Final delivery at 4K | Seedance 2.0 — the only route that renders it |
| Twenty social variations without blowing the budget | Seedance 2.0 Mini |
| Five directions explored before lunch, then one finished properly | Seedance 2.0 Fast, then 2.0 or 2.5 |
| Extending or editing a clip you already generated | Seedance 2.5 |
Where it fits in an agent stack
The bigger win is placement. Video now lives where the rest of your pipeline lives: an agent that already writes the brief, picks the shots, and drafts the copy can hand the same job to a video model through the same key and the same Bearer authentication as every other route on the platform.
That turns a multi-vendor pipeline into a single one: the LLM writes the scene description, Seedance renders it, the output drops back into your storage as an MP4. The model pages list "use with agents" as a first-class scenario for exactly this reason.
FAQ
Do I need a separate account or key for video?
No. Video models run on the same Felo OpenAPI platform as the LLM routes. One API key covers both, and the same Bearer authentication applies to every request.
What is the difference between Seedance 2.5 and Seedance 2.0?
Seedance 2.5 is the newer, long-form route: clips up to 30 seconds, up to 50 reference assets, video editing and extension, and multilingual audio. Seedance 2.0 is the flagship quality route and the only one that renders up to 4K.
Can I generate video with sound?
Yes. Set generate_audio in the request and the output includes a matching audio track. Audio generation does not change the billing rate, and Seedance 2.5 supports multilingual audiovisual generation.
Is the video API synchronous?
No. Video generation is asynchronous: submit a task, poll the task ID until the status is succeeded, then download the signed output URL.
Why is billing per second?
Because output length is the variable that actually drives compute. Per-second pricing means a 4-second test and a 30-second hero clip are priced proportionally, and you can estimate a campaign's cost before you render it.
Can I try it before writing any code?
Yes. Each model page includes a playground that runs a real generation request with your own API key and tracks the task to completion, so you can compare models on your own prompts first.
Video generation, without the second stack
Text, code, images, and now video: all reachable from one API key, one base URL, and one billing line. The four Seedance models cover the full range from a $0.38 draft to a 4K hero clip, and the async pattern is the same three steps for every one of them.
Open a model page, grab a key, and send one request. The first clip tells you more than any spec sheet.
Explore the Seedance models on Felo OpenAPI →
This post is also available in 简体中文, 日本語, 한국어, 繁體中文, हिन्दी, Français, العربية, Русский, اردو, Bahasa Indonesia, Deutsch, Tiếng Việt, Türkçe, Italiano, ไทย, Español, বাংলা and Português.