Pinned Tweet
I built a workflow that renovates a room from one photo 👇
GPT Image 2 x Seedance 2.0.
A dated fixer-upper turns into a finished renovation, then a cinematic walkthrough of the new room. Before and after included.
One wireflow workflow does the whole thing end to end.
I built a growth team inside Claude Code, with 50 roles that all read the same brand file.
I ran the team once for Cinderpan, a cast-iron skillet brand we made up. All 50 roles finished in 12 minutes 43 seconds.
2
1
89
Claude Code plans the work and hands each job to the right role, and each role has its own job, what it reads, what it writes, a method and a rule it never breaks.
Claude fills in the brand file from your website once, and every role reads it.
1
16
ChatGPT plugin monitoring is live in @sitefireai.
Jochen's point is that ChatGPT now organically recommends plugins, and getting picked works a lot like SEO.
Made this from that launch image and their dashboard.
ChatGPT now organically recommends plugins - and getting picked works a lot like SEO.
OpenAI says to treat your plugin's description like product copy. If it's good enough, people skip your website and start using your plugin.
ChatGPT plugin monitoring is live in @sitefireai!
71
AndyAI retweeted
Kling 4.0 Flash is now live on Wireflow.
→ Set Model to Flash on the Kling 4.0 card
→ 5 second 720p clip: 54 credits, vs 91 on Kling 4.0
→ 3 to 20 seconds from a prompt, a start frame or up to 10 reference images
→ Edits clips up to 20 seconds
wireflow.ai/models/kling-4-0
What would you draft first at 41% fewer credits?
AndyAI retweeted
The quiet big deal in Seedream 5.5 is native transparency. Every cutout step in an ad pipeline just disappears. Wiring it into Wireflow the day BytePlus opens the API. What would you build with transparent output?
Introducing Seedream 5.5.
More realistic images. More stable editing across 100 rounds.
1. Keep refining, with pixel-level preservation of unedited regions and reduced detail loss across successive edits.
2. Generate transparent and semi-transparent assets with native RGBA output, ready for compositing and reuse.
3. More natural lighting, materials, and skin textures, with significantly improved small-text rendering.
Seedream 5.5 is coming to Dreamina, CapCut, and BytePlus. We can't wait for you to try it!
Explore what's new: seed.bytedance.com/seedream5…
Claude and Gemini Omni 1.2 Flash made a 5-second talking ad for Vorlune from a product photo, a photo of a made-up actor and a single prompt. Vorlune isn't a real brand, and the man isn't real either.
2
1
124
Google's Gemini Omni 1.2 Flash has sound built in, so the made-up actor says his line to camera in the same run that makes the picture, then tilts the can he's holding, and its label stays right.
1
31
Claude Opus built a workflow that draws, voices and edits cartoon history Shorts 👇
Opus designed its own cast of brand-new characters, then wrote true stories told to the viewer as "you".
1
108
Opus split the cartoon Shorts workflow into small flows, one for each step from plan to final edit.
On the Pig War Short, Nano Banana Pro drew each picture from the cast sheets, Wan turned it into a clip and ElevenLabs read the story.
1
40
I built a Claude skill pack that turns one product photo into ad creative 👇
13 of our skills across 4 stages, from the first packshot through statics and video to making ads in bulk.
5
5
262
AndyAI retweeted
I spent $0 and 55 of 70 free credits testing @wireflowai.
The ad images were good and my product stayed the same. Then one run charged 16 credits when the node said 12.
Access: free account with Google sign-in. The app gave me 70 credits on the Free plan (pricing says up to 60). No card, no referral, no promo code.
What I tested
Easy: text to image, water bottle ad with the headline "Stay cold for 24 hours". 18 s, 5 credits
Typical: my own bottle photo into a lifestyle ad. 28 s, 16 credits (shown 12)
Hard: 5 s video with Wan 3.0 from my ad. About 2 min, first video free (shown ~29 credits)
Extra: ask the chat for square, story and wide versions. 31 s, 34 credits for 2 of 3
What it does well
→ Text spelled right in every image
→ Kept my bottle, color, bamboo band and logo
→ Every download free, no watermark
→ One free first video, 480x854
What it does badly
→ Charged 16 credits when the node showed 12
→ The AI chat dropped my headline from the prompt
→ Adds badges and taglines I did not ask for
→ Asked for square, got 4:3
→ Headline fades out of the free video
→ Your next video needs a plan
→ Story size skipped: 3 x 17 = 51 credits, I had 49
Pricing as listed
Free $0 (up to 60 starter credits)
Starter $36/mo (2,660 credits)
Professional $59/mo (4,370 credits)
Team $299/mo (22,140 credits)
Annual: 2 months free
Scores
Text to image 7/10
Product photo to ad 7/10
Image to video 6/10
Batch variants 6/10
Free tier value 6/10
Credit clarity 4/10
Speed and UI 7/10
Verdict 6.1/10
Use it for quick product ad images on a free account.
Skip it if you need more than one free video or exact costs before you run.
What I would fix first
→ Charge what the node shows
→ Keep my prompt and headline
→ Use the size written in the prompt
→ Make the pricing page match the app
Tip: run one node at a time on Free, and check the credit balance after each run. Bookmark this before you sign up.
Comment Any AI platform you want me to review next and follow along.
AndyAI retweeted
No video models were used in this scene.
Our agent built the whole room in Blender code: the rain on the window, the warm lamp, the light bouncing off the walls, the slow push-in.
On a cloud GPU that's about 4 cents a clip. Five seconds from a video model runs about $1.80.
We're putting this behind an API: describe a scene, get the shot back.
What would you build first?
new YT vid live
link to full video below
you'll learn
how kling 4.0 and seedance 2.5 compare when they get the same prompt and the same images
how every clip got scored blind out of five, and what counts as usable
how each one reads 24 words in 12 seconds for a ugc talking ad
how wind moves her hair and clothes in each model
how well they keep a face and a bottle from two reference photos
how both models make up a line when you don't write one
how fast action and a whip pan to the sea hold up
how close each gets to a true point of view camera move
how the clay look and the mouth shapes come out
how five hard cuts fit inside one 13 second clip
how they fill the gap between a start frame and an end frame
how much of a busy product ad each one gets through in four seconds
how to pick between them for each kind of ad
how to run the same nine tests on your own product with a public wireflow flow
4
4
210
Gemini Omni 1.2 Flash made this talking ad from two photos and one prompt.
The can and the man are made up, and his voice came out of the same Wireflow run as the picture.
The 5-second clip cost $0.70, and the model took 34.5 seconds to make it.
1
78
I put the prompt, the brand file template and the Wireflow board in a free kit: assets.wireflow.ai/playbooks…
13