ChatGPT Try On or Sume for product ads? Different jobs
ChatGPT Try On is a shopper's feature inside ChatGPT results. Sume is an API for the images and videos you publish yourself. What each gives a clothing seller.
Two different products
TechCrunch, read 2026-10-03, describes ChatGPT's Try On as a button in shopping results that lets a person upload a selfie or full-body photo and see clothing or accessories on themselves, with the result and photo kept in their ChatGPT Library. It is a shopper's tool, and the shopper owns the photo and the result.
Sume does not put a button in ChatGPT. It is an API and agent workspace for making the images and videos you publish: a model photo, a lookbook clip, a product ad. The two do not compete for the same job.
Side by side
| Question | ChatGPT Try On | Sume |
|---|---|---|
| Who uses it | A shopper, on their own photo | You, on your garment and a model or creator |
| Output you can publish | Not described in the article | Images and video you generate, on media.sume.com |
| Image model | ChatGPT Images 2.5 | openai/gpt-image-2.5 and others, up to 16 references |
| Video | Not described | Video Router models, including Recast for person swaps |
| Control over the result | The shopper's | Yours, through prompt and references |
What a seller can do now
A shopper's try-on starts from your product listing, so the listing's images matter more than before: a clean garment photo, a front and back, and a size chart. Sume's image API can make an on-model version from a flat garment photo with GPT Image 2.5, and the Video Router can animate it.
Whatever tool made the picture, put your size chart and return policy on the page next to it.
A reasonable plan
- Keep your garment photos sharp and consistent so a shopper's try-on has good input.
- Generate your own on-model images for ads from rights-cleared photos.
- Do not present a generated image as proof of fit.
- Label synthetic people where a platform or law asks for it.
What to watch for next
The launch coverage does not say how merchants take part, which products qualify or what controls a seller has, so there is nothing to configure on that side from the article alone. Revisit when OpenAI publishes merchant documentation, and plan on your own images and video in the meantime.
Sources
Related posts
More in Comparisons
- Claude Code mod vs MCP server vs skill vs hook: where Sume fits
Claude Code mods, MCP servers, skills and settings hooks overlap. Which one gives an agent Sume's image and video tools, and which one only guards them.
- Colossyan 20-30 minute video cap vs Sume avatar videos of 4-60 seconds
Colossyan's pricing page lists a per-video limit of 20 to 50 minutes and 40 scenes. Sume's avatar video takes 4 to 60 seconds per job.
- Comfy Agent Ask or Auto mode vs unattended Sume Format runs
Comfy Agent asks before each run or runs on its own. Sume API runs never ask. See which controls replace the approval prompt when a batch goes overnight.
- Comfy Agent vs an MCP agent for image and video work
Comfy Agent builds and runs ComfyUI graphs for you in Comfy Cloud; an MCP agent calls hosted tools like Sume's. What differs in control, billing and location.
Written by Sume