Changelog
New model additions, performance fixes, and pricing updates.
0.9.0
18 October 2026
Speech from text, Multi Generate, Brand Kit, and four new models
- The catalog grows from thirty-four models to thirty-eight, from nine makers.
- Gemini 3.8 Flash TTS turns text into speech, with 70 voices for a single narrator or a two-speaker dialogue. On Run, 8.2 credits (Flash) or 5.4 credits (Flash Lite) per 1,000 characters are held first; the real bill is worked out once the audio exists, and the rest goes straight back to your balance. Results play and download from History.
- Vidu Q4 for 3 to 16-second video up to 4K, from one first frame or up to 15 reference pictures, with built-in dialogue and sound effects. From 7.2 credits a second at 540p up to 62.4 at 4K.
- Wan 2.6 Video from text, one picture, or the subject of reference videos. The Standard tier always has sound, at 14 credits a second at 720p and 21 at 1080p. The Flash tier covers pictures and reference videos, at 8 and 12 credits a second with sound, or half that with audio off.
- Nano Banana 2.1 for making and editing pictures up to 4K: 4.4 credits at 1K, 6.6 at 2K and 9.9 at 4K, cheaper than Nano Banana 2 at every size.
- Seedream 5 adds a Flash tier at 3.24 credits a picture from 1K to 2K, and Layer (Flash), which splits a picture into layers at that same price per layer. The old Layer tier is now called Layer (Pro).
- Veo 3.1 now has three tiers. Lite starts at 36 credits a video, Fast keeps its old prices, and Quality starts at 300 credits. Reference mode is on Lite and Fast. It is now called Veo 3.1, without Fast.
- Multi Generate runs one prompt on several models at once, up to five, straight from the console, with the results side by side.
- Brand Kit keeps up to five kits of colours, logos and brand notes. One button adds the whole kit to the prompt and its pictures to the console.
- Improve prompt is now a button in the console toolbar, 0.2 credits a use, and writes its result straight into the prompt box.
- Results can be shared through a private link. Whoever opens it can press Make one like this to open the console on the same model and settings.
- Workflows come with sixteen starters under plainer names, and one workflow can now run over many photos in a single batch.
- The picture picker and gallery open in fixed-size dialogs laid out as a masonry wall, and long option lists are no longer cut off.
- Cancel is now only offered while a generation is still queued, with every credit returned. Once it has been sent to the model it runs to the end, and its result still lands in History.
0.8.0
13 October 2026
Workflows: chain several models into one flow
- Workflows, a new section in the studio. You lay out nodes on a canvas (media, text, generate and scene), connect them, then run the whole thing with one button.
- Before any credits are taken, Run all shows what each node is expected to cost and the total. A node whose result still matches its inputs isn't run again and isn't charged.
- The System Prompt node: the AI answers in exactly the structure you define, down to fields inside fields and lists with a fixed number of items. Every field becomes its own port you can connect to other nodes.
- A System Prompt can read up to 10 pictures and costs 0.2 credits a run. If the answer fails or doesn't fit the structure, the credits come back.
- A User Prompt node for the request itself, with variables and pictures, and a Condition node that picks the next path from a value. Nodes on a path that wasn't taken don't run, so they aren't charged.
- Ran out of balance halfway? The run waits instead of failing, and carries on after a top-up.
- 25 ready-made workflow templates on Home, from a studio product shot and background removal up to a three-scene storyboard and a social media content pack. Pick one and the workflow is made in your account.
- Home now has two tabs, Models and Workflows, each with its own template gallery.
- A workflow can be downloaded as a JSON file and imported into another one. An import can be undone with Ctrl+Z.
0.7.0
29 September 2026
The store opens
- The store in the studio now has 12 products: PDF guides and one asset pack, all paid for in credits.
- The Prompt Basics guide is free for every verified account: eight chapters and 50 before-and-after prompts.
- Harbolnas 11.11 & 12.12 Content Kit (149 credits): a production calendar, promo mechanics, 28 prompts and 13 captions.
- For online sellers: Product Photos Without a Studio (149), Marketplace Banners and Thumbnails (99), and Virtual Model Lookbook (199).
- For video: 30 Short Video Hooks 9:16 (149), 15-Second Ad Storyboards (199), and Camera Moves Dictionary (99).
- For design and brands: Visual Style Dictionary (99), Consistent Mascot (149), and Pitch Moodboard Kit (149).
- Seamless Patterns and Textures Pack (99): a ZIP of 100 pattern tiles as SVG and 2048 px PNG, drawn in code rather than by AI.
- Anything you buy can be downloaded again at any time, and when a file is updated you get the new version automatically.
- Credit package prices on the landing page now appear as five cards, with QRIS payment explained in three steps.
- Paid QRIS orders now turn into credits sooner.
- Support is now on Telegram at t.me/XPlor3.
0.6.0
22 September 2026
Ten new models and new prices
- The catalog grows from twenty-four models to thirty-four, from eight makers.
- Qwen 3, Qwen 2, Qwen 2.1, and Z-Image for images. Qwen 2.1 can make pictures on a transparent background, and Z-Image is now the cheapest way to make a new picture, at 0.8 credits.
- Happy Horse 1 and Happy Horse 1.1 for video. Happy Horse 1 can also edit a video you already have; an edit is billed per second for the clip going in and the clip coming out, so a 5-second clip counts as 10 seconds.
- MiniMax H3 for video up to 2K, from text, first and last frames, or reference pictures, video and audio. A reference clip's seconds are billed on top of the output's, and every reference picture after the fifth costs 4 credits.
- Veo 3.1 Fast with built-in sound and a price per video.
- Recraft Remove Background and Recraft Crisp Upscale: upload a picture, no prompt needed.
- New prices for seventeen models, most of them lower. Nano Banana goes from 6 to 4.8 credits, Seedance 2 at 720p from 49.2 to 45.1 credits a second, Kling 2.6 with sound from 24.2 to 22 credits a second.
- Three go up: Seedance 2 on the Mini tier (480p from 4.56 to 4.94 credits a second, 720p from 9.84 to 12.3; with a reference video, 480p from 2.88 to 3.12 and 720p from 6 to 7.5), Grok Imagine Video at 1080p (from 7.2 to 9.6 credits a second), and Seedance 2.5 at 1080p (from 136.8 to 173.8 credits a second; with a reference video from 82.2 to 104.5). Seedance 2.5 at 480p and 720p goes down.
- Seedance 2.5 with Generate Audio on is now billed at twice the price, reference video seconds included. Generate Audio is on by default; turn it off when the video needs no sound and the price goes back to the usual rate.
- Wan 3 Video and Wan 2.7's reference mode now bill a reference clip's seconds too, at the same rate per second, on top of the output's: a 5-second clip behind a 5-second output counts as 10. Before, only the output was billed. Without a reference video, the price is unchanged.
- An info button beside the variation count explains how each model works: what sets its price, what is billed on top, and the limits on its files.
- Reference files are now checked before any credits are taken, against each model's own limits: size, format, dimensions, and the length of video and audio.
- GPT Image 2 can now make a transparent background at 1K, and GPT Image 2.5 adds four ratios at 1K: 27:16, 16:27, 9:8 and 8:9.
- Grok Imagine Video now takes up to seven reference pictures, or one at 1080p.
- Wan 2.7's edit mode can now keep the video's original sound and follows the edited video's ratio. Wan 2.7 Image Standard goes up to 2K; 4K is on the Pro tier.
- Each Wan 2.7 Video mode now shows only the uploads it uses. Before, text mode still offered picture and video uploads, and the request was then refused when run.
- Kling 3 elements with pictures now work. Before, the pictures never reached the model.
- The maker filter in your history now works for every maker. Picking Kuaishou, xAI or Alibaba used to empty the list.
- Videos you generate from this release on can be used again straight away as a reference video.
0.5.0
14 September 2026
Twelve new models, and Kling 3 with shots
- The catalog grows from twelve models to twenty-four.
- Kling 3, Kling 3 Turbo, Kling 3 Motion Control, and Kling 2.6 for video.
- Wan 3 Video with Standard and Prime tiers, plus Wan 2.7 for video and images.
- Grok Imagine: two image models and two video models.
- GPT Image 2.5 with Sunburst and Flare styles.
- Kling 3 builds a video shot by shot: each shot carries its own prompt and duration, and can draw on named elements reused across shots.
- Reference files collapse to one tile per field, with the contents shown on hover.
0.3.0
27 August 2026
Ten models land in the catalog
- Nano Banana, Nano Banana 2, and Nano Banana Pro.
- Seedream 4, 4.5, and 5 with lite and pro tiers.
- GPT Image 1.5 and GPT Image 2.
- Seedance 1 and Seedance 1.5 Pro for video.
- Per-tier pricing: one card can carry three different prices.
0.2.0
26 August 2026
Buying credits and reference images
- Three credit packages, paid through DOKU.
- Reference image upload for image-to-image modes.
- Reference image order can be changed from the keyboard.
0.1.0
25 August 2026
First release
- Sign up, write a prompt, and watch the result appear with no refresh.
- Credits are debited when a job is created and refunded when a job fails.
- A gallery that keeps every result in our own storage.