polym-eval-generate-seedance

name	polym-eval-generate-seedance
description	Generate videos with BytePlus Seedance. Use for Seedance text-to-video, image-to-video, camera control, and draft video workflows.

Generate Seedance Videos

Generate videos using BytePlus Seedance via the Ark SDK.

API Reference: https://docs.byteplus.com/en/docs/ModelArk/1366799

Pre-flight Check

Before generating, verify in .env:

• BYTEPLUS_API_KEY (Required)
• BYTEPLUS_BASE_URL (default: https://ark.ap-southeast.bytepluses.com/api/v3)
• SEEDANCE_MODEL (optional, default from env)
• For reference images: TOS_BUCKET, TOS_ACCESS_KEY, TOS_SECRET_KEY

Quick Start

# Text-to-Video
python skills/polym-eval-generate-seedance/scripts/generate.py \
  --prompt "A cat playing with a ball of yarn" \
  --model seedance-1-5-pro-251215

# Image-to-Video (first frame)
python skills/polym-eval-generate-seedance/scripts/generate.py \
  --prompt "The scene comes alive with gentle motion" \
  --first-frame photo.jpg \
  --model seedance-1-5-pro-251215

Parameter Reference

Parameter	CLI Flag	Type	Description	Default
prompt	--prompt, -p	string	Text prompt	Required
model	--model, -m	string	Model ID	env SEEDANCE_MODEL
first_frame	--first-frame	string	First frame image path/URL (I2V)	None
last_frame	--last-frame	string	Last frame image path/URL	None
seconds	--seconds, -d	int	Duration (2-12s, 1.5-pro: 4-12s)	8
size	--size, -s	string	Resolution: 480p, 720p, 1080p	720p
aspect_ratio	--aspect-ratio	string	16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptive	auto-detect
watermark	--watermark	flag	Add watermark	False
camerafixed	--camerafixed	flag	Fix camera position	False
draft	--draft	flag	Draft mode (480p only, faster)	False
seed	--seed	int	Random seed	None
output	--output, -o	string	Output path	generated/{id}.mp4

Available Models

Model	Resolution	Duration	Notes
seedance-1-5-pro-251215	480p/720p/1080p	4-12s	Latest, native audio
seedance-1-0-pro-250528	480p/720p/1080p	2-12s	Strong visual consistency
seedance-1-0-pro-fast-251015	480p/720p/1080p	2-12s	Low latency
seedance-1-0-lite-t2v-250428	480p/720p/1080p	2-12s	T2V only
seedance-1-0-lite-i2v-250428	480p/720p	2-12s	I2V only

Generation Modes

• T2V — text only, no reference images
• I2V (First Frame) — --first-frame photo.jpg
• I2V (Last Frame) — --last-frame photo.jpg
• First & Last Frame — both --first-frame and --last-frame

Aspect ratio is auto-detected from reference images. Params are auto-validated/adjusted per model capabilities.

Dependencies

byteplussdkarkruntime
Pillow
requests
python-dotenv
tqdm
tos  # for reference image upload

Troubleshooting

"BYTEPLUS_API_KEY is not set"

Set it in .env.

"参考图模式不支持 1080p"

Use 720p for lite I2V models.

Draft mode forces 480p

This is by design — draft mode only supports 480p for faster generation.