Use the WaveSpeed AI platform to generate and process videos, including image-to-video, video continuation, video chain generation, and 4K upscaling.
When to Use
Applicable to the following scenarios:
- Generate dynamic videos from images (image-to-video)
- Video upscaling to 4K/2K/1080p
- Create video chains (splicing multiple video segments)
- Hero background video production
- Video continuation and extension
Do NOT use this skill if:
- Only need a simple selfie video (use zhuoran-video-selfie or clawra-video-selfie)
- Need real-time video generation (the API is asynchronous)
- No reference images or video materials
- Extremely poor network conditions (need to upload/download large files)
Trigger Keywords / Trigger Phrases:
- AI video generation / AI video generation
- image to video / image to video
- video upscaling / video upscaling
- 4K video / 4K video
- video continuation / video continuation
- video chain / video chain
- video processing / video processing
- generate video / generate video
Workflow
1. Image-to-Video
import requests
import time
API_KEY="***"
BASE_URL = "https://api.wavespeed.ai/api/v3"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}
JSON_HEADERS = {**HEADERS, "Content-Type": "application/json"}
# 1. 上传图片
def upload_image(image_path):
with open(image_path, 'rb') as f:
r = requests.post(
f"{BASE_URL}/media/upload/binary",
headers=HEADERS,
files={"file": f}
)
return r.json()["data"]["download_url"]
# 2. 提交视频生成任务
def submit_i2v_task(image_url, prompt, model="wavespeed-ai/wan-2.2/i2v-480p"):
payload = {
"image": image_url,
"prompt": prompt,
"duration": 5, # 5 或 8 秒
"enable_prompt_expansion": False
}
r = requests.post(f"{BASE_URL}/{model}", headers=JSON_HEADERS, json=payload)
return r.json()["data"]
# 3. 轮询结果
def poll_result(task_id, get_url=None):
poll_url = get_url or f"{BASE_URL}/predictions/{task_id}"
while True:
r = requests.get(poll_url, headers=HEADERS)
data = r.json()["data"]
if data["status"] == "completed":
return data["outputs"][0]
elif data["status"] == "failed":
raise Exception(f"Task failed: {data}")
time.sleep(5)
# 4. 下载视频
def download_video(video_url, output_path):
r = requests.get(video_url, stream=True)
with open(output_path, 'wb') as f:
for chunk in r.iter_content(chunk_size=8192):
f.write(chunk)
2. Video upscaling to 4K
# 1. 上传视频
video_url = upload_video(video_path) # 同 upload_image
# 2. 提交超分任务
payload = {
"video": video_url,
"target_resolution": "4k" # 可选: 720p, 1080p, 2k, 4k
}
r = requests.post(
f"{BASE_URL}/wavespeed-ai/video-upscaler-pro",
headers=JSON_HEADERS,
json=payload
)
task_data = r.json()["data"]
# 3. 轮询并下载（同上）
3. Video chain generation
A video chain refers to connecting multiple short videos into one coherent long video through video continuation technology.
Core concepts:
- First frame extraction: Use FFmpeg to extract the last frame of the video as the starting point of the next segment
- Video continuation: Use the same prompt style to generate a coherent sequel
- Video splicing: Use FFmpeg for lossless splicing
# 提取最后一帧
def extract_last_frame(video_path, output_path):
import subprocess
cmd = [
"ffmpeg", "-y", "-sseof", "-0.1",
"-i", video_path,
"-vframes", "1", "-q:v", "2",
output_path
]
subprocess.run(cmd, check=True)
# 生成视频链
def generate_chain(initial_image, prompt, num_segments=4):
videos = []
current_image = initial_image
for i in range(num_segments):
# 上传当前图片
image_url = upload_image(current_image)
# 生成视频
task = submit_i2v_task(image_url, prompt)
video_url = poll_result(task["id"], task["urls"]["get"])
# 下载视频
video_path = f"v{i+1}.mp4"
download_video(video_url, video_path)
videos.append(video_path)
# 提取最后一帧用于下一段
current_image = f"v{i+1}_last_frame.jpg"
extract_last_frame(video_path, current_image)
return videos
# 拼接视频
def concat_videos(video_list, output_path):
import subprocess
# 创建文件列表
with open("concat_list.txt", "w") as f:
for v in video_list:
f.write(f"file '{v}'\\n")
cmd = [
"ffmpeg", "-y", "-f", "concat", "-safe", "0",
"-i", "concat_list.txt", "-c", "copy",
output_path
]
subprocess.run(cmd, check=True)
Supported Models
Image-to-Video
| Model | API path | Duration | Features |
|---|---|---|---|
| Wan-2.2 | wavespeed-ai/wan-2.2/i2v-480p |
5s/8s | High quality, supports long videos |
| Wan-2.2 720p | wavespeed-ai/wan-2.2/i2v-720p |
5s/8s | Higher resolution |
| Hailuo 2.3 | minimax/hailuo-2.3/i2v-standard |
6s | Fast, good results |
| Kling 1.6 | kuaishou/kling-v1.6/i2v-pro |
5s/10s | Stable, rich in detail |
Video Upscaling
| Model | API path | Resolution options |
|---|---|---|
| Video Upscaler Pro | wavespeed-ai/video-upscaler-pro |
720p, 1080p, 2k, 4k |
Guardrails
Prompt writing suggestions
Good prompt structure:
[动作描述], [氛围/风格], [细节], [效果]
Example:
"Gentle flowing motion with glowing particles moving slowly across the scene,
subtle pulsing neon light effects, ethereal and dreamy atmosphere,
smooth ambient movement, futuristic cyberpunk feeling"
Notes
- Avoid excessive post-processing: Do not use FFmpeg to force stretch the resolution; use native 4K output
- Keep the style consistent: Use the same prompt style for each segment of the video chain
- Reserve buffer: Upscaling tasks may take 10-15 minutes; set a sufficient polling timeout
- Back up the original files: Back up the original video before upscaling
Common issues
| Issue | Solution |
|---|---|
| Edges look blurry after upscaling | Use target_resolution: "4k" native output |
| Video chain is incoherent | Ensure that the true last frame is extracted |
| Task timeout | Increase max_attempts or check the API status |
| Upload failed | Check the file size limit (usually 500MB) |
API key configuration
Store the API key in an environment variable or .env In the file:
# .env
WAVESPEED_KEY=your_api_key_here
Note: Do not hardcode API keys in scripts or commit them to version control.
Project Scripts
| Script | Features |
|---|---|
generate_hero_video.py |
Generate Hero background video from a poster image |
upscale_hero_video.py |
Upscale video to 4K |
generate_latest_chains.py |
Generate the complete video chain |
wan_v1.py ~ wan_v4.py |
Wan model single-segment generation |
upscale_intro_native_4k.py |
Intro video upscaling |
upscale_loop_native_4k.py |
Loop video upscaling |
Related Skills
| Skills | Relationships | Description |
|---|---|---|
| zhuoran-video-selfie | Dedicated version | Selfie video generation for the Zhuoran character |
| clawra-video-selfie | Dedicated version | Selfie video generation for the Clawra character |
| voice-clone | Companion | Can add cloned voice to videos |
| zhuoran-selfie | Reference | Photo generation skill |
About UniqueClub
Part of UniqueClub toolkit - AI-powered creative tools for professional video generation. Learn more: https://uniqueclub.ai
Last updated: 2026-02-07