Six photos to one morphing transformation video

Six photos to one morphing transformation video

An n8n workflow. Send six pictures through its form, and RunningHub turns each neighbouring pair into a short clip that
morphs one picture into the next. Then fal.ai joins the five clips into one video, in the order you sent the pictures.

What is inside

• workflow.json: the workflow as n8n exports it, 57 nodes and no sticky notes. It holds no credential. Your two keys go
  into one node, API Setup.
• README.md: these notes.

What we changed from the original copy

• Both fields of API Setup held the same words asking for a key. They are now YOUR_RUNNINGHUB_API_KEY and
  YOUR_FAL_KEY, so you can tell them apart.
• Nothing else. Its one Code node, getBinaryFile, needed no repair.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A RunningHub account and its API key. Each clip is a paid task there.
• A fal.ai account and its API key, for the job that joins the clips.
• Six pictures in .jpg, .jpeg, .png or .webp, in the order the video should follow.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Open API Setup and replace YOUR_RUNNINGHUB_API_KEY with your RunningHub key.
3. In the same node, replace YOUR_FAL_KEY with your fal.ai key, and nothing more. The three fal.ai nodes put Key and a
   space in front of it themselves, so a value typed as Key … would reach fal.ai as Key Key … and be refused.
4. Optional: open On form submission and turn Multiple Files off on the fifth and sixth fields, as it already is on the
   first four (see Good to know).
5. Save, open the form's test address and send six pictures through it.
6. Press Publish, then share the form's production address.

The form confirms at once, and the run goes on behind it. The joined video ends up as a file in downloadVideo, so open
the run under Executions to get it, or add a node after downloadVideo to save it to a drive or send it on.

How a run goes

1. On form submission takes the six files. getBinaryFile passes them on, and Merge1 puts them in one item with the
   two keys from API Setup.
2. Ten upload nodes send the pictures to RunningHub. Pictures two to five go up twice, once as the closing frame of one
   pair and once as the opening frame of the next.
3. Merge and Merge2 to Merge5, each followed by an Aggregate node, build the five pairs: 1 and 2, 2 and 3, 3 and 4,
   4 and 5, 5 and 6.
4. For each pair, one of HTTP Request to HTTP Request4 starts the RunningHub AI app 1935261777502601217. The opening
   frame goes into its node 64 and the closing frame into node 112. Two more of its nodes set a vertical frame and a clip
   of two seconds (Change the look).
5. A Wait node pauses one minute, then a checkStatus node asks for the task's state. Its If node sends SUCCESS on to
   getOutput, which reads the clip's address. Any other state goes back to the same Wait for another minute.
6. Merge All links collects the five clips in pair order, and Aggregate5 puts their addresses in one list.
7. Submit merge job sends the list to fal.ai's FFmpeg merge endpoint. Wait5 pauses 30 seconds, Check Status asks for
   the job's state, and If5 loops back to Wait5 until fal.ai reports COMPLETED.
8. Get Final Result reads the joined video's address, and downloadVideo fetches the file.

| Node | Endpoint | Sends | Gets back |
|---|---|---|---|
| The ten upload nodes | RunningHub /task/openapi/upload | the key and one picture | the stored picture's fileName |
| HTTP Request to HTTP Request4 | RunningHub /task/openapi/ai-app/run | the key, webappId and nodeInfoList | a taskId |
| checkStatus to checkStatus4 | RunningHub /task/openapi/status | the key and the taskId | the task's state, such as SUCCESS |
| getOutput to getOutput4 | RunningHub /task/openapi/outputs | the key and the taskId | the clip's fileUrl |
| Submit merge job | fal.ai fal-ai/ffmpeg-api/merge-videos, on its queue | the five addresses in video_urls | a request_id |
| Check Status, Get Final Result | fal.ai fal-ai/ffmpeg-api/requests/<request_id> | the key | the job's state, then video.url |

Change the look

• Frame: node 304 in all five HTTP Request nodes, 1 for vertical and 0 for horizontal. Change all five, or the clips
  will not match when they are joined.
• Clip length: node 307 in the same five nodes. The file sets 2 seconds, so the joined video runs about ten seconds.
• Order: the form's fields. The first picture opens the video and the sixth closes it.
• Check timing: Wait to Wait4 pause one minute before each RunningHub check, and Wait5 pauses 30 seconds before each
  fal.ai check.
• The form: its title reads Start & End Frame Tranformation Creator, spelled as in the file. Rename it in
  On form submission. If you rename a field, change the upload nodes too: they read the files as
  Upload_your_1st_image to Upload_your_6th_image, names n8n makes from the field labels.
• Which app makes the clips: webappId in the five HTTP Request nodes. The node numbers in nodeInfoList belong to
  that app, so check them if you point it at another.

Good to know

• RunningHub bills five clip tasks for each run, and fal.ai bills one merge job.
• A clip that fails never ends the run. The If nodes look for SUCCESS only, so a failed task is checked again every
  minute. When no video has arrived after a long wait, open the run under Executions and stop it.
• n8n's v1 execution order, which this file uses, runs the five branches one at a time. The clips are made one after
  another, each with at least a minute of waiting, before the merge starts.
• The fifth and sixth form fields leave Multiple Files at n8n's default, which is on. Send one picture in each. Two files in
  one field get numbered names, such as Upload_your_5th_image_0, and the upload node finds neither.
• Once the workflow is published, anyone with the form's address can start a run on your RunningHub and fal.ai accounts.
• Both keys sit as plain text in API Setup, so anyone who can open the workflow can read them. Take them out before you
  share it.
• The pictures go to RunningHub, and the clips' addresses go to fal.ai. Use pictures you have the right to use, and show
  a person's face only with their consent.