What prompts can I paste into Claude to make images and videos with Wireflow?
Paste these prompts into Claude to make images and videos with Wireflow: image sets, photo to video, product shots, an AI influencer, music and batch runs.
The prompts below are ready to paste into Claude with Wireflow connected, one for each job: a set of images built as a workflow, a video from your photo with Seedance 2.5 or Kling O3, product images, an AI influencer, background music, and a batch run over a list in Claude Code. Every prompt that spends credits has Claude name the model and its credit cost and wait for your yes before it runs. Claude looks up that price with list_models, which spends nothing. Replace everything in square brackets with your own details.
Not connected yet? Click Add Wireflow to Claude. It opens Claude's connector form with the name Wireflow and the URL https://www.wireflow.ai/api/mcp already filled in. Claude walks you through Wireflow's sign-in and a consent screen, with no API key to paste. The full steps are in Claude app setup, and the Claude Code steps are in Use Wireflow from your agent.
To use your own photo, give Claude a link to a file that is already online at an https:// URL, or ask Claude for an upload link. create_upload_link makes a no-login, phone-friendly link that lands one image, video or audio file on your board as an Import node, and get_upload_link checks whether you used it and finds the Import node it filled. create_upload_link needs the opt-in workflows:write permission, so select it on the consent screen when you connect.
How do I get Claude to build a set of images as a Wireflow workflow?
Claude uses create_workflow_from_prompt to save the set as a private draft workflow, which runs nothing. The draft uses the Nano Banana Pro node to make each image from text, and Claude runs it with run_workflow once you say yes.
Use Wireflow. With create_workflow_from_prompt, build a draft workflow that makes [number] images of [your subject] in [style, light and colors], with one Nano Banana Pro node per image and these shots: [shot 1], [shot 2], [shot 3]. Show me the nodes and their prompts, tell me the model and the credit cost of one run, and wait for my yes before you run it with run_workflow. When it finishes, give me every image link.
Claude reads the finished run with get_execution and gives you one link per image. The takes also live on the draft's board in Wireflow, where you can change one value and run it again.
How do I get Claude to turn my photo into a Seedance or Kling video?
Claude builds a draft workflow that animates your photo with one of two nodes. On the Kling O3 node, your photo is the required start frame and your prompt directs the motion. On the Seedance 2.5 node, a photo on Start Frame is the exact image the video opens on.
Use Wireflow to turn my photo into a video: [link to your photo]. Build a draft workflow with [Kling O3 or Seedance 2.5], my photo as the start frame, [length in seconds] long, in [aspect ratio, such as 16:9]. The motion: [what moves, how the camera moves and the mood]. Tell me the model and the credit cost, and wait for my yes before you run it. When it finishes, give me the video link and check the clip with check_video.
You get a link to the clip. check_video costs no credits and answers pass, review or fail after checking for frozen regions, silent audio, black or flash frames and the clip's length. A Seedance 2.5 clip has sound when its Generate Audio setting is on.
What should I ask Claude to make product images from one photo?
Ask Claude to put your product photo on Image 1 of the Nano Banana Pro node, which is the image the model edits, and to give each shot its own scene prompt.
Use Wireflow to make product images from my photo: [link to your product photo]. Build a draft workflow with Nano Banana Pro, my photo wired to Image 1, and one prompt per shot: [shot 1, such as a white background with a soft shadow], [shot 2, such as a lifestyle shot on a kitchen counter], [shot 3, such as held in a hand]. Write each shot as the setting, the surface the product stands on, the light, two or three props at most and the camera angle, and in every prompt keep the product's shape, colors, label, logo and text exactly as photographed. Tell me the model and the credit cost, and wait for my yes before you run it. Then compare each image with my photo and tell me anything that changed.
You get one image link per shot. Open each one next to your photo and check the shape, label and logo before you use it. In Claude Code, the product photography skill on the agent skills page also turns one product photo into new shots.
How do I get Claude to create an AI influencer photo and video?
Claude makes one still with the GPT Image 2.5 Sunburst node, then animates it with Seedance 2.5 in reference mode. These are the nodes the node map on the MCP connector page lists for a realistic AI talking-head avatar, and the prompt follows the rules in Make a Realistic AI Avatar.
Use Wireflow to make an AI influencer: [who they are, their look and style], in [the setting], doing [what they do in the clip]. Step 1: make one still with GPT Image 2.5 Sunburst from text alone, waist-up or closer, at custom size 1152x2048 with quality max, no film grain, sensor noise or pores in the prompt, and no upscale. Show me the still. Step 2: animate the still I approve with Seedance 2.5 in reference mode: leave Start Frame empty, wire the still into Reference Images, and keep the raw output with no video upscale. Before each step, tell me the model and the credit cost and wait for my yes.
You get a link to the still, then a link to the clip. Reference mode is what holds the face steady through a head turn, so check the face at crop zoom as well as at the full frame.
How do I get Claude to make background music for my video?
Claude makes the track with the Lyria 3 Pro node, one of the three music nodes in the same node map. It turns one style and mood prompt into a full, structured song of up to about three minutes and writes its own lyrics.
Use Wireflow to make background music for my video with Lyria 3 Pro. The video shows [what your video shows]. The track: [genre], [mood], [instruments], [tempo in BPM] and [vocal style]. Steer the mood with this frame from my video as the Mood Reference: [link to a still from your video]. Tell me the model and the credit cost, and wait for my yes before you run it. When it finishes, give me the audio link.
A song takes about three minutes to render. You then get a link to the Generated Audio. Play the track against your video before you use it.
How do I make Claude Code run my workflow once per item in a list?
Claude Code calls POST /api/v1/workflows/{id}/run once for each item, with that item's value in inputs and its own Idempotency-Key, and keeps each executionId to poll. Each item is a separate run, as run one workflow over a list describes. The script uses a Wireflow API key, which starts with sk- (see API Overview), while the lookups in the prompt use your Wireflow connection.
Use Wireflow. Find my saved workflow [workflow name] with list_workflows and get_workflow, then tell me its id, the inputs it accepts, the models it uses and, from list_models, the credit cost of one run and of the whole list in [path to your list file, one item per line]. Wait for my yes. Then read https://www.wireflow.ai/docs/api/run and write a script that calls POST /api/v1/workflows/{id}/run once per line, with that line in the [input to fill] input, its own Idempotency-Key, and my API key from the [name of the environment variable that holds your key] environment variable, never printed. Before the first run, call whoami and GET /api/v1/me with the key, and stop if they are different accounts. If a call returns 429, wait and retry it. Run the first two lines, show me their output links and wait for my yes before the rest. Keep every executionId, poll each run until it finishes, and save results.csv with each line, its executionId, its status and its output links.
Each item comes back as its own run with its own executionId, and results.csv lists every line with its status and output links. Open the first two before you say yes to the rest. In Claude Code, the ai-workflow-runner skill also runs your workflow once per row of a CSV.
For connection steps, see Claude app setup. For how many runs each plan gives a month, see Credits & Billing.