SeeDance 2.5 (BudgetPixel)
SeeDance 2.5 — the flagship, where references become a creative medium
- end_image
- image
- reference_audios
- reference_images
- reference_videos
- video
SeeDance 2.5 is ByteDance's flagship multimodal video model and the most interesting node in the BudgetPixel video lineup - because the reference inputs stop being a fixed few images and become a full media palette. Up to 15 reference images (free), up to 5 reference video clips for editing and extension, up to 5 reference audio clips for sound and voice guidance, 4–30 second clips at up to 1080p. It's the closest a hosted model gets to a video editing bay. It's also 330 credits per second at the headline rate, which is how you know it's the flagship.
Why you'd reach for it
The reference-video workflow is the reason. Feed it a clip plus a prompt and it edits, extends, or borrows motion from it - there's no separate "edit endpoint," video editing and extension are just reference-video modes. Reference audio means you can drive voice/sound direction from a sample. The multi-asset mode where the prompt can address "Image 1", "Video 1" by order is where the real creative power lives. And audio-only input is supported, which almost nothing else in the pack can do.
The inputs that matter
- prompt - required; the edit or motion instruction in reference mode.
- reference_images - up to 15, free, for identity/style guidance.
- reference_videos - up to 5 clips (2–15s each, 30s combined, ≤50MB each) for editing/extension/motion. These bill at half the output rate per input second on top of the output - the one input that costs you.
- reference_audios - up to 5 clips (30s combined, WAV/MP3), free; audio-only input is supported.
- image / end_image - start/end frames for classic image-to-video; can't combine with any reference media.
- length_seconds - 4–30.
- resolution - 480p/720p/1080p.
- generate_audio - off by default, no price impact.
Output is video (VIDEO), core type - 2025+ ComfyUI needed.
How it works
Same hosted mechanics as every BudgetPixel node - serialize, upload, create a job, poll, download the MP4 into the temp folder. The polling loop on a 30-second multimodal clip is the slowest in the pack, so budget your patience (and your credit balance).
Install and key setup
ComfyUI Manager → BudgetPixel, or:
cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt
Only requests. API key from a paid plan via BUDGETPIXEL_API_KEY or budgetpixel.json; never in workflow JSON.
Where people get burned
The input-video billing is the one to watch - reference videos bill at half the output rate per input second in addition to the output, so a long reference clip can dominate the cost of a short output. BudgetPixel Cost Estimate will tell you before you commit. And note the reference-media-versus-frames exclusivity: connect an image and all your reference media silently stops being part of the job. Interrupted runs still finish and bill server-side, as everywhere in this pack.
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | Text description of the video. | |
| aspect_ratioopt | COMBO | 16:9 | Aspect ratio for text-to-video. Ignored (adaptive) when a start frame or any reference media sets the frame. |
| end_imageopt | IMAGE | Optional end frame, used together with `image` (the start frame) to interpolate the video between the two frames. Same input forms as `image`. | |
| generate_audioopt | BOOLEAN | false | Generate audio with the video (no price impact). |
| imageopt | IMAGE | Optional start frame for image-to-video. Provide a public image URL, a data URI, raw base64, or an uploaded-file URL from POST /v1/uploads. Omit for text-to-video. Can't be combined with reference media. | |
| length_secondsopt | INT | 54–30 | Output video length in seconds (4-30). |
| reference_audiosopt | AUDIO | Optional reference audio clips (up to 5; each 2-15s, 30s combined; WAV/MP3) that guide sound/voice. Audio-only input is supported. Each item is a public audio URL or an uploaded-file URL from POST /v1/uploads. Free — no input billing. Can't be combined with `image`/`end_image`. | |
| reference_imagesopt | IMAGE | Optional reference images (up to 15, free) that guide identity/style in reference-to-video mode. Each item is a public image URL, a data URI, raw base64, or an uploaded-file URL from POST /v1/uploads. Can't be combined with `image`/`end_image`. | |
| reference_videosopt | VIDEO | Optional reference video clips (up to 5; each 2-15s, 30s combined; MP4/MOV, ≤50MB each) for editing, extension, and motion/identity guidance. Each item is a public video URL or an uploaded-file URL from POST /v1/uploads (videos are passed by URL, not inlined). Billed on the INPUT duration at half the output per-second rate (80/sec at 480p, 160/sec at 720p, 375/sec at 1080p) in addition to the output — estimate via POST /v1/cost. Can't be combined with `image`/`end_image`. | |
| resolutionopt | COMBO | 720p | Output resolution. Pricing varies by resolution — see `resolution_pricing` in GET /v1/models. |
| force_rerunopt | BOOLEAN | false | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |