BlendBlend.Get Started

Changelog

New models, faster workflows, and the occasional same-day fix. Everything we ship to Blend, as we ship it.

0.18.0

MAI-Image-2.6 and workflow building

Create and edit with MAI-Image-2.6

Create images from a prompt or edit with up to five reference images using MAI-Image-2.6. Choose an aspect ratio or let the model choose the size.

Build workflows from instructions

Describe a workflow or ask for changes in the workflow editor. Connected assistants can also create and update saved workflow drafts.

Improvements

  • Image cost estimates account for MAI-Image-2.6 reference images and output size.
  • Workflow building covers more patterns, including lists, media batches, choices, tool steps, and image-to-video.
  • Workflow replies include a cost estimate when the draft passes its completion checks.
  • New workflow conversations keep their history for follow-up edits.
  • Agent activity expands while it runs and collapses when it finishes.

Fixes

  • Fixed an issue that blocked app updates.
  • Workflow edits report missing connections instead of silently removing them.
  • Workflow completion checks catch missing inputs and unfinished changes.
  • Fixed workflow layout failures during automatic edits.
  • Z.ai models show the correct icon.

Other

  • Admins can download conversation transcripts for review.
0.17.0

Reliable workflow files

Instagram publishing

Connect your Instagram accounts to publish or schedule photos, Reels, and carousels. Set a comment keyword to send a link in a private reply.

Improvements

  • Upload HEIC, HIF, and supported RAW photos without converting them first. Large photos are resized to fit within 4,096 pixels before upload, and a preparation status shows when they are being processed.
  • Assistant requests can now run for up to 30 minutes, so long tasks have more time to finish.
  • Published workflows now open with a clean custom input form instead of loading the first example automatically.
  • Published workflow inputs entered before login are restored once after login without carrying over to later refreshes.
  • Workflow choice inputs can now show text or image options and use a separate text or media value for each choice.
  • Workflow list outputs can now take their item count from a connected number input.
  • Image choices in workflows can have a title on each option, and hovering an option shows its full name.

Fixes

  • Workflow files now load reliably after refresh and remain available when you duplicate shared workflows.
  • Older workflows now continue to open correctly after workflow updates.
  • The model selector now shows the correct brand logo instead of a generic icon.
  • Files in older published workflow examples now load correctly.
  • Published workflow example images now transition smoothly from preview to full detail.
  • Fork details now load correctly when you reopen a conversation.
  • Landing page links now open the correct workflows, examples, creator earnings, privacy, and sign-up destinations.
  • Plan credit selectors now have clear labels for screen readers.
  • Fixed failures when signing in during checkout, changing subscriptions, resuming workflow selections, and browsing available models.

Other

  • Updated project contribution guidance to keep implementation work focused.
  • Removed retired Studio and preview tools.
0.16.0

A clearer public identity

Improvements

  • BlendAI's official company and founder profiles are now connected more clearly across the web.
  • Clarifying questions now stay beside the prompt box and move through a compact card stack, so answers are easier to review and submit.
  • Published workflow runners now show more compatible examples and use a clearer Examples label.
  • Published workflow runners now fill the available screen height.
  • Library images now show in full by default instead of being cropped.

Fixes

  • Web searches now return results reliably during assistant conversations.
  • Image and video workflows now keep distinct reference files while preventing duplicate inputs.
  • Published workflow examples now load images from direct links correctly.
  • Workflow discovery no longer shows workflows that are not ready to run.
0.15.0

More control, creative models, and flexible plans

Choose your monthly credits

Creator, Growth, and Scale now include several monthly credit options. Compare prices and estimated output for popular models before you choose.

Create images and videos in chat

Ask BlendAI to create, edit, transform, or animate media, then choose from Good, Better, and Best options with a credit estimate before generating.

Improvements

  • Choose whether each assistant tool is off, asks for approval, or runs automatically.
  • Answer structured clarification questions directly in chat and workflow authoring.
  • Approval decisions and clarification answers stay visible in the conversation.
  • Added a wider range of image and video models for fast drafts, high-quality production, video editing, and synchronized audio.
  • Image models that support multiple native outputs can now return several images from one generation.
  • Video settings now expose each model's supported audio, duration, resolution, framing, and advanced controls.
  • Change your plan or monthly credit allowance from billing settings.
  • Review the amount due and start date before you confirm a plan change.
  • See your selected credit allowance throughout checkout and account settings.
  • Follow-up prompts can now reuse images and videos from earlier in the conversation.
  • The analytics dashboard now has clearer navigation and richer chart, metric, and table views.
  • Chart time ranges are easier to inspect and adjust with mouse, keyboard, or touch controls.
  • Cost estimates now appear for image and video models and adjust to the selected output count, duration, resolution, audio, and input media.
  • Text generation cost estimates now reflect the requested response limit and use a consistent default when no limit is selected.
  • Explore image and video concepts in a new Studio workspace with creative controls, reusable references, and a browsable media gallery.
  • Optional theme controls now stay hidden until you enable them.
  • MCP clients can now generate an image or video from a creative brief, with optional prompt handling, model settings, and reference files.
  • Added a Contact page with support, creator program, and partnership channels.

Fixes

  • Failed chat requests now show a clear error state instead of leaving an unfinished response.
  • Your Max Content Width setting now applies to conversations and the prompt box at any width you pick.
  • Attached files now stay inside the prompt box and scroll sideways instead of stretching across the screen.

Other

  • General stability and reliability improvements.
0.14.0

Web search comes to chat

Choose before you run

Published workflows can now let runners choose one or more text, image, video, audio, or file options before a run starts.

Web search in chat

Blend can now search the web for current information when your question needs it. Answers cite their sources inline, and each citation opens a preview with the page title, site, and excerpt.

Workflow links unfurl as rich preview cards

Share a published workflow on X, iMessage, Slack, or anywhere links unfurl, and it now shows a designed preview card: the cover, real example outputs, the workflow's name and description, and a Run workflow button. Every workflow gets its own look, and the card refreshes when you republish.

Workflow links unfurl as rich preview cards

Improvements

  • Published workflow results got a full redesign: images and videos keep their natural size and shape, text results sit alongside them in a tidy grid, and everything opens full screen in one click.
  • Live generations remain available for longer after a temporary disconnect.
  • Connected assistants can now generate and attach polished covers to published workflows.
  • Model choices and selector categories can now be refreshed faster as new releases become available.
  • Image upload controls in workflow nodes now stay compact after the first image is added, leaving more thumbnails visible without scrolling.
  • A browsable library of more than 1,000 Recraft examples now makes it easier to study visual styles and copy prompts when building workflows.
  • Image model cards now show an estimated credit cost per image.
  • Image model names in the picker are now shorter and easier to scan.
  • Connected assistants can now find, edit, run, and publish workflows more reliably.
  • Connected assistants can now upload files directly to Blend and attach them to workflows, including large files.
  • Connected assistants can now read complete information about Blend's marketplace, pricing, company, support, and privacy pages.
  • Connected assistants can now check each model's supported settings, including aspect ratios and reference image limits, before they build a workflow.
  • Normal app use now produces less unnecessary browser console noise.
  • Nano Banana 2 Lite is now available for fast 1K image generation and editing.
  • Model names now stay readable in selectors without animated scrolling or background gradients.

Fixes

  • Using a tool in chat no longer creates a duplicate entry in your request history.
  • Some published workflows could fail to start with a cost estimate error. They now start and price normally.
  • Selected workflow nodes now show a clear blue outline on the canvas.
  • Generated image previews now appear as soon as they are ready, keep their intended shape, and remain visible until the final image loads.
  • Generated SVG images now display and download with the correct file type.
  • Recraft models now display the correct brand icon.
0.13.0

A cleaner workflow canvas

Improvements

  • Workflow editors now use one clear canvas control for hiding the surrounding sidebar and header.
  • A new visual guide makes it easier to understand how Blend workflows, creator earnings, and the marketplace fit together.
  • Connecting AI agents to Blend now uses the shorter /api/mcp address.
  • Workflows with built-in processing steps now move through those steps faster.
  • Workflow nodes now keep clear status colors after a run, making completed, attention-needed, and failed steps easier to spot.
  • Workflow output previews now render consistently across node types.
  • The homepage now has smoother scroll-responsive motion and more varied visual artwork.
  • Homepage and file-preview images now load faster while staying sharp.
  • The sign-in and sign-up pages now use a calmer, focused dark layout.
  • Improved page-loading reliability and responsiveness across the app.
  • AI agents can now import files and media from public links, test workflow runs, and publish workflows with examples.
  • Long-running chats now show a calm live activity indicator with elapsed time and rotating status messages.
  • Thinking and tool calls are easier to scan, with expandable input and output details when you want more context.
  • Conversation titles now complete more reliably while staying short and clear.
  • Workflows created or updated by AI agents now open with neatly arranged nodes.
  • Questionnaire forms now support single-choice, multiple-choice, and written responses with keyboard-friendly navigation.
  • Chat search options are now simpler, with the retired deep-search experience removed.

Fixes

  • Workflow editors no longer crash when entering or leaving the full-canvas view.
  • Publishing a workflow no longer causes the editor to crash after a successful publish.
  • AI agents no longer get stuck while connecting to Blend.
  • Consulted model answers now appear correctly inside workflow runs.
  • Built-in workflow utilities no longer require a model before they can run.
  • Re-publishing a workflow through an AI agent now keeps existing details, pricing, and examples unless they are changed.

Other

  • Improved the safety and stability of routine app updates.
  • Project guidance now stays consistent across supported coding assistants.
0.12.0

Smarter, smoother workflows

Workflow covers, made for you

Blend can now create a polished cover image directly from your workflow while you publish, with framing controls and quick alternatives.

A calmer prompt editor

Expanding a prompt now opens a full screen writing view. Your text sits in a centered column that stays comfortable to read at any length, and the controls stay quietly out of the way.

Improvements

  • Publishing now keeps a clear success panel in place, with quick access to the live workflow and its API.
  • Workflow covers now use a wider 16:9 layout across the editor and workflow library.
  • New workflows open at a closer zoom level, so the first node you add is easier to see.
  • Choose the exact file grid density you want, from 2 to 12 columns.
  • More files now load in rows that match your selected grid density.
  • File uploads now behave consistently across chats, workflows, profile photos, feedback, and the media library.
  • Images, videos, and documents now load and preview more consistently throughout Blend.

Fixes

  • Running a single workflow step now uses every value from connected lists and collections, so batch counts match what actually runs.
  • Workflow editors now open without briefly showing the sidebar.
  • Edits in the expanded prompt editor now save as you type, instead of only when you close it.
  • The scrollbar in the expanded prompt editor no longer sits on top of your text.
  • File thumbnails now render with smooth, clean rounded corners in every state.
  • Workflow connections stay easy to select around grouped canvas areas, and running animations now remain on connection lines instead of appearing inside completed nodes.
0.11.0

More reliable workflows and media generation

Improvements

  • Toast countdown borders now draw clockwise with a cleaner, track-free animation.
  • Image and video models now use clearer, more consistent settings across prompts and workflows.
  • What's new updates now follow a predictable weekly schedule.
  • Media across chats, workflows, uploads, and the library now share a clearer, more consistent preview experience.
  • The media library now offers an adjustable grid and easier selection.
  • Workflow connections can now keep files together in one run or process one file per run, with clear controls directly on the canvas.

Fixes

  • Tool step outputs now match the look and formatting of AI step outputs.
  • Image and video generations now complete more reliably across prompts and workflows.
  • Estimated image and video generation costs are now more accurate across sizes, durations, and batch counts.
0.10.0

A distraction-free workflow canvas

A brand-new home page

Blend's landing page got a full redesign: a cleaner hero, richer workflow showcases, clearer pricing, and a warmer look from top to bottom. Log-in and sign-up were redesigned to match.

Improvements

  • Picking a paid plan while signed out now opens a quick sign-up right on the page, then brings you straight back to checkout for that plan.
  • Checkout now pre-fills your name and email, so there is less to type before subscribing.
  • Subscription plans have clearer names: Creator, Growth, and Scale.
  • The workflow editor now opens in a distraction-free canvas view, with run, publish, history, arrangement, and node controls close at hand.
  • Frames can now grab every workflow step they touch, making grouped sections easy to move together.
  • Batch steps now make it clear that connecting a User Input is optional.
  • Notification toasts got a cleaner close button: a single crisp countdown ring with rounded ends instead of a doubled border.
  • Aspiring creators can now apply through a focused form, share their social profiles and workflow idea, and receive creator onboarding after acceptance.
  • The home page can now open directly to Workflows for signed-in users, making it quicker to start building.
  • The sign-up screen now links directly to the Privacy Policy, making email update terms clearer before account creation.
  • Standalone tool steps now use a compact picker that matches the height of nearby controls.
  • Focus mode now shows your workflow's save status and lets you rename it without leaving the canvas.
  • Short-form content drafting now uses more consistent voice guidance and better-grounded source material.
  • The file picker is now roomier and lets you browse or create collections, select files across sources, resize previews, and switch between cropped and full-image views.
  • Published workflows can now ask for a collection, with a focused picker for choosing or creating the right collection before a run.
  • Workflow node documentation now covers file Set and Explode behavior, reusable collections, structured LLM outputs, standalone runs, and canvas annotations.
  • Monthly AI usage costs are now easier to review in a clear local report.

Fixes

  • Advanced settings in workflow steps now stay open after you connect an input, instead of collapsing as soon as you drop the connection.
  • The countdown ring on notification toasts now stays in sync with how long each toast is actually shown, instead of finishing early.
  • Legal Data Hunter is available again in the assistant tools menu.
  • Published workflows no longer show an error for inputs that are no longer part of the workflow.
  • Creator application answers now stay filled in when submission fails, so you can correct the issue and retry without starting over.
  • Email updates now preserve unsubscribe choices more reliably.
  • Important service alerts now reach our team more reliably, helping us respond faster when something goes wrong.
  • Connected image files now work reliably in workflow tool steps.
  • Workflow sidebar tabs no longer show an unwanted outer border.
  • New workflow connections no longer flash red on the first attempt after adding a step.
  • Workflow step menus now stay aligned with their fields while you zoom the canvas.
  • Pressing Escape now closes an open workflow menu or rename field before leaving focus mode.
  • Image to Prompt now handles large images and complete collages more reliably, instead of sometimes failing or focusing on a single tile.
  • TIFF images can now be uploaded reliably, and failed uploads show a clearer reason.
  • You can now upload up to 20 images at once.
  • Workflow batches now process complete file lists and show accurate generation counts and price totals.
  • Text generations that stop with an error now report the failure instead of appearing complete.

Other

  • Improved reliability across the app.
  • Reduced unnecessary warning noise during production requests.
  • Improved Image to Prompt recreation quality evaluation and refinement, with faster and more reliable multi-image batches, deeper visual reasoning, more thorough guidance, easier-to-follow terminal progress, and clearly highlighted reconstruction scores.
0.9.0

Clearer names for workflow inputs and outputs

Improvements

  • Workflow inputs and outputs have clearer names: the node for what people fill in is now called User Input, and the node for what your workflow returns is now called User Result.

Fixes

  • Workflow canvas swipes now stay inside the editor more reliably, instead of occasionally navigating backward or forward in your browser.
0.8.1

Fixes for adding files and run details

Fixes

  • Adding files from your library works again. Earlier today the picker could show an empty list.
  • Checking a workflow run through the API now includes its result and cost.
0.8.0

A clearer Library for your chats and media

Improvements

  • Your Library is easier to scan, with separate tabs for chats, media, and saved items, expandable previews, and quick links back into any chat.
  • In your Library, media now shows the whole image by default so nothing is cropped, with a toggle to fill the tile instead.
  • When you build a workflow with AI, Blendito handles complex multi-step edits more reliably.
  • When you build a workflow with AI, the canvas now arranges new steps into a tidy layout automatically as the assistant adds and connects them, instead of leaving them to overlap.
  • You can now compare prompt improvements across selected models, repeated improvement passes, and reasoning settings, with judge feedback tied to each model and pass.
  • Your workflow run history now shows a rich preview of each past run, its result, the original input, and any generated text, grouped by date, so you can find the run you want at a glance.
  • Options in a workflow step's model settings, like aspect ratio, now appear in alphabetical order so they are easier to scan.
  • When you build a workflow with AI, you can now choose between a super fast, balanced, or smarter model to match the job.
  • When you build a workflow with AI, you can now attach an image for the assistant to work from, on models that support images.
  • Downloading an image or video now saves it straight to your device with a progress bar, instead of opening in a new browser tab.
  • You can now download images directly from workflow previews, including expanded fan-out results.
  • Notifications now have a close button you can click to dismiss them, with a ring around it that counts down the time left before they disappear.

Fixes

  • Workflow run estimates now explain when a tool step needs a runnable tool selected.
  • The cost badge on a workflow step now shows the total for all of its runs, matching the run count, instead of the price of a single run.
  • When you build a workflow with AI, Blendito now asks its follow-up questions reliably, without the occasional error or pause before they appear.
  • Saved preferences and recent work now stick more reliably across refreshes.
  • When you share a workflow with specific people, the link in their invite email now opens the workflow correctly.
  • Notifications now match your light or dark theme, instead of always showing on a light background.
  • Expanded fan-out results now size to each image, so tall results no longer spill past their card or leave an empty gap at the bottom of the group.
  • Images you download now save with the correct file extension, so they open as an image instead of an unrecognized file.
  • Fetching a workflow run through the API now returns its generated images and files, along with what the run cost.
0.7.0

Sharper image prompts when you build with AI

Improvements

  • You can now let people running your published workflow edit a Batch step's list of values: connect a Published Input to the Batch, and runners can add, change, or remove items before they run it.
  • The price shown on a published workflow now updates live as runners add or remove list items, so the cost always matches what will actually run.
  • Published workflow inputs that drive a Batch list accept a list of values through the API, and the API reference marks them as lists with a matching example.
  • When you build a workflow with AI, image steps now get sharper, more detailed prompts written for you automatically.
  • When you build a workflow with AI, Blendito handles complex multi-step edits more reliably.
  • When you build a workflow with AI, the canvas now arranges new steps into a tidy layout automatically as the assistant adds and connects them, instead of leaving them to overlap.
  • You can now compare prompt improvements across selected models, repeated improvement passes, and reasoning settings, with judge feedback tied to each model and pass.
  • You can now download images directly from workflow previews, including expanded fan-out results.
  • Downloading an image or video now saves it straight to your device with a progress bar, instead of opening in a new browser tab.
  • Your workflow run history now shows a rich preview of each past run, its result, the original input, and any generated text, grouped by date, so you can find the run you want at a glance.
  • Options in a workflow step's model settings, like aspect ratio, now appear in alphabetical order so they are easier to scan.
  • When you build a workflow with AI, you can now choose between a super fast, balanced, or smarter model to match the job.
  • When you build a workflow with AI, you can now attach an image for the assistant to work from, on models that support images.
  • External agents can connect to Blend MCP with your existing API key, and every published workflow is available automatically.

Fixes

  • A workflow run no longer starts, or charges you, when a required input was left empty; you get a clear message naming the missing input instead.
  • Workflow run estimates now explain when a tool step needs a runnable tool selected.
  • The cost badge on a workflow step now shows the total for all of its runs, matching the run count, instead of the price of a single run.
  • When you build a workflow with AI, Blendito now asks its follow-up questions reliably, without the occasional error or pause before they appear.
  • Saved preferences and recent work now stick more reliably across refreshes.
0.6.0

Use several models in one generation step

Improvements

  • Steps that produce several results now carry through: generate 12 images and the next step runs 12 times, once per image, instead of collapsing everything into a single run. Collections and uploaded image sets still travel together as one bundle.
  • Batch nodes now accept images: give each row its own set of reference images, and optionally fan a row out into one run per image, or use a Batch to gather fanned-out results back into a single set.
  • Wire other steps into a Batch and it now tries every combination of its rows automatically, so two lists of options multiply into a full grid of runs.
  • Run counts and cost estimates shown before you run now match what actually runs, including runs fanned out from a collection or a batch.
  • Running a single generation step across several models at once now shows all of their settings together, grouping the options that only apply to some models by which models they affect, so it's clear which setting belongs to which model.
  • Image generation now offers an Auto size: choose it and your result follows the orientation you describe in the prompt (a portrait or a wide landscape) instead of always coming out square.
  • The homepage now highlights content-creator workflows: turn one idea into a week of on-brand posts, repurpose a long piece into a multi-channel pack, or cut a long video into a titled clips pack.
  • The homepage use cases section is now interactive: scroll through the jobs Blend handles (fashion, legal, e-commerce, research, and more) and watch each job's workflow come together step by step.
  • Founding creators keep 100% of every run: the first 100 creators we onboard keep the full price of every run for life, even after the standard split kicks in.
  • Affiliate program: you now earn 20% of every subscription payment from the people you refer, for as long as they stay subscribed.
  • Workflow creators now price a published workflow as a percentage of each run, so the fee scales with how much a run actually uses instead of a single flat amount.
  • Published workflows now show clearer cost ranges before you run them, including variable steps and creator fees.
  • When you publish a workflow, you can now frame its cover image: zoom in or out, drag to reposition, and fill any empty space with a blur of the image or a solid color, so every cover looks the way you want across the marketplace.
  • You can now annotate the workflow canvas: add sticky notes, text, images, arrows, and frames to label, group, and explain a workflow without changing how it runs. Resize any annotation to fit, and grow a frame as large as you need to wrap a whole group of nodes.
  • Visit /new to jump straight into a fresh workflow, no need to open the workflow list and click New first.
  • When a generation step uses several models with similar names, hover any model label to see its full name.
  • You can now stop a workflow while it's running: hit Stop and the run halts, skipping any steps that haven't started yet.
  • When a step produces several results, hovering or selecting one now highlights the exact inputs that produced it and shows their prompts, so you can tell which model and prompt made each output, even while the other results are still generating.
  • The workflow builder chat is smoother: it keeps the newest reply in view as it streams, lets you start typing your next message while it's still working, and adds a button to jump back to the latest when you scroll up.
  • The Choose Output step is now easier to read: view each option full size, flip between them with a thumbnail strip, then pick the one to continue with.
  • The workflow editor now has quick access to Previous Run and Delete in the right-side tool rail.

Fixes

  • Moving a Batch node's output to a different field no longer leaves behind a connection of the wrong type (like an image collection wired into a text field).
  • Republishing a shared workflow no longer accidentally removes existing recipients if you publish before the share list finishes loading.
  • Fixed a glitch where a workflow listed in more than one category could show up twice in the Discover spotlight.
  • Fixed a freeze in the workflow editor that could happen when adding or removing batch items, or when selecting certain nodes on the canvas.
  • Fixed a problem where a file or image chosen from your Library could fail to run in a workflow.
  • Fixed an error that could stop some published workflows from opening.
  • The chat prompt box is now wider, and it stays the same size whether you're starting a new chat or continuing a conversation.
  • The recent chats list in the sidebar now sizes itself to your window and zoom level, so it no longer leaves an awkward gap or gets cut off.
  • The What's new notification now appears only when there's a new update since you last saw it, instead of popping up again on every visit.
  • Reopening a workflow no longer leaves the editor stuck as if a run were still going, so you can start a new run right away.
  • When you run a workflow, image results now show at their real proportions instead of being cropped into a square, so wide or tall images are no longer cut off.
  • The page shown when you open a workflow that isn't shared with you now has the same header bar as the rest of the app, instead of looking unfinished.
  • Workflow result images no longer flash black while they load. A soft blurred preview now shows first and the full image fades in as you switch between a step's results.
  • Long conversations with the workflow builder assistant now save to your chat history reliably, so you can pick them back up later instead of losing the thread.
0.5.0

Workflow controls, batching, and clearer file inputs

Improvements

  • Workflow steps now have a Run and collapse button, so you can start a step and tuck it away in one click while it keeps running.
  • Collapsed workflow steps now keep their Run button, so you can run a step without expanding it first.
  • Workflow auto-arrange is now easier to reach and gives separate branches more breathing room.
  • Workflow file fields can now pick from files you've already uploaded, with file choices narrowed to what the field accepts.
  • Batch workflows that vary models now run once for each selected model and estimate costs more accurately.
  • The homepage now always appears in a light theme, no longer following your device's dark mode.
  • Workflow Batch rows now suggest more useful choices when you add another row, while still restoring the default after the list is cleared.
  • Changed the default chat model from Claude Sonnet 4.6 to Gemini 3.5 Flash.
  • You can now resize the Blendito chat panel to give the conversation more room, and it remembers your preferred width.
  • Blendito now keeps your recent chats, so you can reopen and continue a previous conversation on the same device.
  • Assistant tool results now open in clearer, purpose-built previews, with raw details still available when needed.
  • Workflow canvas selection now works across multiple steps, so you can copy, paste, move, or clear a group at once.
  • New Image to Prompt step: add an image and get back a detailed prompt that recreates its subject, composition, lighting, and style.

Fixes

  • When a workflow node run fails, it now shows a clear error message in its output area instead of unexpected raw text.
  • You can now move between a workflow node's previous runs even when one of them failed.
  • Tool errors now stay visible when a tool result is expanded.
  • Images uploaded to a workflow's published inputs now reappear correctly after you reload the editor.
  • Workflow steps that don't finish no longer leave the Run button stuck.
  • Workflow connections now explain more clearly when an input cannot be reused with another field.
  • Batch workflow results no longer duplicate completed runs after a step finishes.
  • The Batch node no longer shows a results section, it only repeats a setting to create multiple runs.
0.4.0

Faster workflow building and better previews

Expand a step's results into a gallery

When a workflow step produces many results at once, like a batch of generated images, open them all in a single view instead of clicking through them one by one. Hover any result to see which inputs created it, resize the gallery to fit more per row, and drag it anywhere on the canvas.

Improvements

  • Blendito now understands Collection nodes, LLM structured-output fields, and the latest workflow handles more reliably.
  • Workflow nodes can now be added directly from a connection handle, with compatible choices shown where you drop.
  • The What's new notification and update dialog are easier to scan, with clearer styling and fuller release details.
  • Choose Output nodes now show just the outputs themselves, making image and text choices easier to compare.
  • Workflow home now has a denser marketplace-style layout with spotlight picks, category badges, templates, and cleaner card actions.
  • Workflow delete confirmation is cleaner and no longer requires typing the workflow name.
  • Refreshed the logged-out homepage with a new landing page covering how Blend works, use cases, distribution, creator earnings, and pricing.
  • Blendito now gives workflow steps clearer names and keeps new workflows to one connected canvas whenever possible.

Fixes

  • Generated image previews now keep portrait and landscape proportions instead of cropping into a square.
  • Result previews now load reliably even when a step produces dozens of images at once.
  • Workflow connection menus now avoid showing choices that cannot work with the selected field.
  • Models with version numbers now appear in the expected order in the model selector.
  • Choose Output keeps your selection when the same choices are shown again.
  • Swiping across the workflow canvas no longer sends you back to the previous browser page.
  • Workflow steps no longer stay stuck showing a running animation after they have finished.
  • Workflow input labels now stay in place after you connect an input to another step.
0.3.0

Live batch streaming and workflow concept guides

Improvements

  • New Concepts section in the docs that explains the core workflow ideas in short, individually linkable pages: lists vs single values (and why a collection is a value, not a list until you batch it), data types and connection rules, and the difference between running a single node, a full workflow, and the Runner.
  • Improved workflow text previews with a wrap toggle in the full-screen editor, a tighter line-number gutter, and compact copy and expand controls for latest outputs.
  • Workflow canvas nodes now look more consistent, with aligned field labels, Advanced Settings chevrons, collapsed handles, and output cards across node types.
  • Workflow streaming now keeps up smoothly even on slower connections.
  • Assistant tools (web search, model consultations, image prompt engineering) hand their results back faster.
  • The collapsed sidebar is slimmer, with evenly aligned and consistently sized icons.
  • Development tabs now keep the branch or workspace label visible as you move through the app.

Fixes

  • Workflow node outputs now stream live even with many batch items running at once. Text no longer trickles in word by word or appears only after the whole run finishes.
  • Refreshing a workflow during a long run now keeps generated results visible instead of hiding them behind an interrupted-run message.
  • Recent chats no longer run into the screen edge on mobile.
0.2.0

Collections, sharing, and a sharper workflow builder

Collections

Create a collection of images or files once, then reference that same list across multiple requests instead of re-adding everything each time. Pull a collection straight into a workflow with the new Collection node.

Reuse files you've already uploaded

File uploads now work the same everywhere you add files, and you can pick from files you've already uploaded instead of uploading them again.

Share workflows with specific people

Share a published workflow with specific people instead of only making it fully public or private: click Publish and add their email. Anything shared with you shows up under Shared with me.

A clearer way to publish

The Workflow publish dialog has been redesigned around organized section tabs, with a live "try it" panel for your workflow's API so you can publish and test your endpoint in one place.

Improvements

  • Workflow result preview and run history are now combined into a single card.
  • Drag nodes straight from the dropdown controls onto the canvas.
  • Dragging nodes no longer drops the frame rate, even on nodes with a lot of text in them.
  • Bring your own key is now available to paid plans and as a BYOK-only subscription from Provider API Key settings.
  • Workflow page tabs are now animated and remembered in the URL.
  • Cleaner "Recent" date grouping in the sidebar.
  • Added full workflow documentation, including a complete node reference.
  • Updated the workflow docs to focus on Gen Canvas workflows, publishing, item fan-out, structured LLM outputs, and the published API output shape.
  • Batch workflow runs now move through their items with fewer unnecessary waits, keeping results in order.
  • The AI workflow builder now understands Collection nodes, LLM structured-output fields, and the latest workflow handles more reliably.
  • Sign-in and sign-up pages now load faster.
  • LLM workflow settings now use clearer response-format controls for choosing plain text, separate fields, or a list of similar items.
  • Workflow node result cards now show per-item status, consistent output and run navigation, and clearer estimated item counts before runs finish.
  • The plan picker now includes the BYOK plan, clearer credit-access copy, and helpful explanations for credit and feature rows.
  • Choosing a plan now opens the in-app payment form directly instead of showing an extra confirmation step.
  • Provider API Key settings now use the full plan comparison dialog for locked accounts instead of a BYOK-only upsell.
  • Model usage estimates in pricing now account for gifted credits and include video model estimates.
  • The sidebar upgrade button now animates continuously and speeds up on hover.

Fixes

  • Fixed a Send button that could silently stop working after a model change.
  • Long lines now wrap properly in the workflow text viewer.
  • Edges and output stay visible when a workflow node is collapsed.
  • Removed an incorrect login error message that could briefly appear after signing in.
  • Free accounts with previously saved provider keys no longer receive BYOK discounts or use those keys until BYOK is active.
  • Fixed a see-through sidebar panel on iOS Safari.
  • Workflow result previews now keep output-item selection, upstream nodes, connected result nodes, and highlighted batch rows in sync.
  • Image and video results with multiple items now keep flowing into connected result nodes.
  • Result nodes now show the correct upstream media preview.
  • Result cards are less likely to appear empty after batch and parallel runs finish.
  • Batch workflow runs now preserve every generated output when downstream nodes fan out over multi-item results.
  • Workflow runs with many image inputs no longer time out.
  • Workflow nodes no longer look stuck in a running state after their run has finished.
  • Workflow result previews now show the active output item instead of stacking every result together.
  • Workflow debug controls no longer overlap the selected-node result details button.
  • LLM workflow previews no longer collapse reasoning or reset expanded output while streamed text is still arriving.

Other

  • Retired the older model alias (friendly names) feature.
0.1.0

Introducing the changelog

A home for every update

We ship to Blend almost every week: new models, faster workflows, and the occasional same-day fix. From now on, every change lands right here.

When you come back after we've shipped something new, you'll get a quiet heads-up in the app so you never miss what changed.

Improvements

  • Faster model switching in chat.
  • Workflows now remember the inputs from your last run.

Fixes

  • Long messages no longer overflow the composer.