
An image you keep revising
Tencent built Hy Image 3.5 Preview around multi-round editing: the result of one turn becomes the input to the next, with the earlier image and the instructions carried forward instead of restarted.
Start Creating NowOnly live routes are selectable. Your prompt stays in place when you switch.
Your image will appear here
Tencent released Hy Image 3.5 Preview on September 22, 2026. It is built for work that does not finish in one shot: send a prompt and reference images, get a result, then keep revising that result across rounds while the model carries the earlier image and the conversation forward. Tencent lists up to 20 reference images, direct output at 4K, and roughly a 30% gain over Hy Image 3.0. This independent Hy Image 3.5 workspace tracks the model and puts its lineage in your hands today: generation currently runs on Hunyuan Image 3.0 and the rest of the image catalog, each named as itself, while the Hy Image 3.5 route is connected. Every new account gets one free image.
Three ways Hy Image 3.5 gives you more control over the result you walk away with.

Tencent built Hy Image 3.5 Preview around multi-round editing: the result of one turn becomes the input to the next, with the earlier image and the instructions carried forward instead of restarted.
Start Creating Now
Hy Image 3.5 raises the reference limit from three images to twenty and generates directly at up to 4096 × 4096, so a brief can carry a subject, a palette and a layout at once.
Tencent Cloud lists Hy Image 3.5 Preview at $0.024 per 2K image and bills only for output images. Generation and instruction editing sit in a single model rather than two separate endpoints.
What Tencent published about Hy Image 3.5 Preview at release. These are their figures for the model, not measurements taken in this workspace.
Announced by Tencent Hunyuan as Hy Image 3.5 Preview, model id hy-image-v3.5-preview.
Tencent's figure. Hy Image 3.0 accepted three.
A result can be revised over successive turns, carrying the earlier image and the conversation forward.
Tencent lists direct 4K generation when it is explicitly requested.
Tencent's own comparison against Hy Image 3.0. Not verified in this workspace.
Tencent Cloud API list price, charged on output images. This workspace prices the routes it actually runs in credits.
Hy Image 3.5 Preview is listed in the model selector and disabled. Generation here runs on Hunyuan Image 3.0 and the rest of the catalog, each under its own name.
Describe the subject, the composition and the light in one clear paragraph.
Pick an available model, check its credit quote, and submit once.
Inspect the result at full size and download it from the workspace or your account history.
See the credit cost before every generation, then choose a subscription or a one-time top-up when you need more.
One free image per new account, enough for a single Hunyuan Image 3.0 generation in the Hy Image 3.5 workspace.
For regular individual creative work.
For campaigns and higher-volume production.
For larger workloads, with adjustable credit capacity.
Capacity uses the lowest-cost live setup for this site. Your exact credit cost changes with the selected model and settings and is shown before generation.
Hy Image 3.5 is Tencent Hunyuan's image model, released as Hy Image 3.5 Preview on September 22, 2026 under the model id hy-image-v3.5-preview. It generates from text and reference images and, unlike a one-shot generator, keeps editing its own result across rounds. Tencent lists up to 20 reference images, output up to 4096 × 4096, and about a 30% improvement over Hy Image 3.0.
Three things, by Tencent's account. The reference limit goes from three images to twenty. Generation and instruction editing live in one model instead of two endpoints. And a request can carry the previous image and the conversation so far, so a fourth revision still knows what the first one was for. Hy Image 3.0 handled each request on its own.
Not yet. Hy Image 3.5 Preview appears in the model selector marked coming soon and cannot be submitted. Tencent currently serves it through Tencent Cloud's own API, which this independent workspace does not run on. Generation here uses Hunyuan Image 3.0 by default, with the rest of the image catalog available; every route is labelled with its real name and nothing is served under the Hy Image 3.5 name.
Each new account receives a one-time grant of 3 credits, which is exactly one Hunyuan Image 3.0 image at the current price. It is granted once per account and does not reset daily. Other models cost their own displayed amount, so the grant may not cover one run on a more expensive route.
Yes. Pick an editing route such as Hunyuan Image 3.0 Instruct · Edit, upload a reference you have permission to use, and describe the single change you want along with what should stay. Reference limits and pricing depend on the model you select. Multi-round editing is a Hy Image 3.5 capability and is not part of the routes running here.
Tencent Cloud lists Hy Image 3.5 Preview at $0.024 per 2K image and bills only for output images, with reference images free. That is Tencent's price for their own API. This workspace charges credits for the models it actually runs, and each one shows its quote before you submit.
Sign in and open your account history. Completed images stay there with a download action, and they are visible only to you. Keep a local copy of anything you want to retain.
No. This is an independently operated workspace with its own accounts and billing. It is not affiliated with or endorsed by Tencent or the Hunyuan team. Everything stated here about the Hy Image 3.5 model comes from Tencent's own release announcement and documentation.
Hy Image 3.5 Preview is Tencent's released image model: 20 reference images, multi-round editing and 4K output. Explore it in the Hy Image 3.5 workspace and get one free image per account.
Start Creating Now