Motion graphic video from a concept image

Motion graphic video from a concept image

An n8n workflow. Upload one concept image through its form and it treats the image as the end frame of a short clip. An
AI writes an edit that gives the start frame, Seedream applies it, GPT-4o compares the two frames and writes the motion,
and Hailuo 02 on WaveSpeedAI renders the video between them.

What is inside

• workflow.json: the workflow as n8n exports it, seventeen nodes. It holds no credential. The five Authorization
  headers still read Bearer <token>, as in the source, for you to fill in.
• README.md: these notes.

One expression differs from the source. In Upload a file, the file name was cleaned with /s+/g and (.[^.]+)$, which
replace runs of the letter s and treat any character as the dot. They now read /\s+/g and (\.[^.]+)$, which
replace spaces and put a timestamp before the extension.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key that can use gpt-4.1-mini and gpt-4o.
• A Dropbox account, a credential for Upload a file and a Dropbox token that can create shared links.
• A WaveSpeedAI API key and credit.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add an OpenAI credential and select it in OpenAI Chat Model and Analyze image.
3. Add a Dropbox credential and select it in Upload a file. In HTTP Request, replace <token> in the Authorization
   header with a Dropbox token that may create shared links. If the link call is refused, check that the token's
   permissions include files.content.write and sharing.write.
4. In image-upscaler, Seedream-v4-edit, Hailuo-i2v-pro and Hailuo-i2v-pro-result, replace <token> in the
   Authorization header with your WaveSpeedAI key. A Header Auth credential does the same job in one place.
5. Save, open the form's test address and send a test image. Check each node's output under Executions.
6. Press Publish at the top of the canvas (or Shift+P) when it works.

How a run goes

The form feeds two lines at once, and Merge waits for both.

1. The image line: Upload a file saves the image in Dropbox, HTTP Request makes a public shared link, Code turns
   the link's dl=0 into raw=1 so it points at the file itself, and image-upscaler returns a 4K version.
2. The planning line: Input Validation keeps the description and the aspect ratio, and Aspect Ratio Conversion turns
   the ratio into a 4K width and height. Next, Retrieve Binary File passes the image on. Last, AI Agent on gpt-4.1-mini
   writes an edit prompt for a start frame that would lead to the uploaded image.
3. Merge joins the upscaled image and the prompt. Seedream-v4-edit applies the prompt to the upscaled image, and
   Edit Fields names the result startFrame and the upscaled image endFrame.
4. Analyze image sends both frames to gpt-4o and gets one paragraph that describes the motion between them.
5. Hailuo-i2v-pro submits the video job, Wait pauses and Hailuo-i2v-pro-result asks WaveSpeedAI for the result.

The workflow stops there. The clip's address is in data.outputs of the last node's output under Executions, and the
file does not download it.

Check before you rely on it

• The aspect ratio you pick is converted and carried to Edit Fields, but no later node reads it. Seedream-v4-edit
  always asks for 2880*2160. To make the choice count, set its size to
  {{ $('Aspect Ratio Conversion').item.json.resolution }} inside the quotes of the JSON body.
• WaveSpeedAI's page for minimax/hailuo-02/i2v-pro, read on 2 November 2026, names the optional final-frame field
  last_image. The file sends end_image. If the clip does not end on your image, rename the field in Hailuo-i2v-pro.
• Wait stores a unit of minutes and leaves the amount at n8n's default, 5, so it pauses five minutes. The same page gives a
  typical generation time of 30 to 90 seconds. If data.status in the result is not completed, run
  Hailuo-i2v-pro-result again a little later, or raise the wait.
• Seedream-v4-edit and image-upscaler ask WaveSpeedAI to answer within the request. If one takes too long, its
  outputs come back empty and the next node fails. Run it again.

Change the look

• Motion and style: the prompt in AI Agent shapes the start frame, and the text in Analyze image shapes the motion.
• Sharpness: target_resolution and creativity in the body of image-upscaler.
• Frame shapes: the options in the form trigger, the sizes in Aspect Ratio Conversion, and size in Seedream-v4-edit.
• After the clip: add a node after Hailuo-i2v-pro-result to download it and save or send it.

Good to know

• WaveSpeedAI's pages, read on 2 November 2026, price a start frame at $0.027 and a clip at $0.49, and describe the clip
  as five seconds at 1080p. The upscaler and the OpenAI calls are billed on top.
• HTTP Request asks for a public shared link, so the image stays in your Dropbox with a link anyone can open.
• Once the workflow is published, anyone with the form's address can start a run on your keys. The form trigger can ask
  for a username and password.
• Five nodes hold a token in their Authorization header, which saves it in the workflow and any export. Move them into
  credentials before you share the workflow.
• The image goes to Dropbox, OpenAI and WaveSpeedAI.