Toucan Alternative: Ambient Blending That Is Still Maintained

If you are reading this you probably installed Toucan at some point, liked it, and then noticed it had gone quiet. That instinct is correct. Toucan still works, it is now completely free, and it has not received an update in over a year. This post covers what actually happened to it, what word-level swapping could never do even when the team was active, and what a maintained version of the same idea looks like today. Toucan's approach was right. The product around it stopped moving.

What Toucan got right#

Toucan installs as a browser extension and rewrites a small number of words on whatever page you load. You read an English article and a handful of nouns and verbs come back in Spanish, visually marked so you notice them. Hover over one and the original returns, usually with pronunciation. There is no session to open, no streak, and no decision about what to study today. The learning is a byproduct of browsing you were doing anyway.

That is a real design achievement and it is why Toucan collected fans who had bounced off every other app. The adoption cost is close to zero. For an adult who has failed to build a study habit four times, an approach that requires no habit at all is not a compromise, it is the entire product.

The mechanic also has a research history. Anthropologist Robbins Burling described it in 1968 as the diglot weave: embed target-language words inside sentences a learner already understands, so meaning is carried by context rather than by a gloss. It is a direct application of comprehensible input, because the surrounding sentence keeps the whole thing readable. The diglot weave method and comprehensible input explain the underlying evidence in more depth. Toucan did not invent that idea, but it was the first product to put it on the live web at scale.

What happened to Toucan#

The company wound down in 2023. Babbel acquired the technology in September 2023, and TechCrunch reported on 19 September 2023 that the deal covered the tech only, with none of the founders or employees joining Babbel. It relaunched as "Toucan by Babbel" and everything that had previously been behind a subscription was moved into the free tier.

For existing users that was a decent outcome. A product that would otherwise have disappeared kept running, at no cost. But it is now a funnel for Babbel subscriptions rather than a product with a roadmap, and the numbers show it. The Chrome Web Store listing reports 200,000 users and a last update of 15 July 2025, while Toucan's own homepage still advertises "750,000+ users". Those two figures may not measure the same thing, but the gap and the year-plus without a release both point the same way.

None of this is a moral failing and it is not something to celebrate. Small teams run out of runway. The practical consequence is what matters: if you are using Toucan today, you are using software that nobody is fixing.

Where word-level swapping hits its ceiling#

Three limits were structural, not bugs, and they were there before the wind-down.

Grammar. Swapping single words out of a native-language sentence means the swapped word has no sentence around it in its own language to agree with. A documented Toucan output reads "on their primero solo excursion", where Spanish requires primera to agree with excursión. Research into Toucan's stack indicates it relied on a local dictionary rather than a translation model, and the open-sourced aracari, pos-js and finch-graphql libraries are consistent with that; this is strong inference rather than a documented fact. Either way, dictionary-form substitution produces exactly this failure, and it teaches a form that does not exist in the sentence you are looking at.

Dosage. Density is deliberately low, because a page full of unknown words stops being readable. Incidental vocabulary acquisition is real but slow: a word usually needs many separate encounters before it sticks, and Nation (2006) puts the vocabulary needed for comfortable unassisted reading at roughly 8,000 to 9,000 word families. A few swapped words per page is a long arithmetic problem.

Follow-up. Ebbinghaus mapped the forgetting curve in 1885 and the shape has held for 140 years. Passive exposure gives you the first encounter for free. It does not give you the scheduled second, fifth and twentieth encounters that move a word into production. That is the job of spaced repetition, and it has to happen somewhere.

What LingoBlend's extension does on a page#

The previous version of this article argued that Toucan was still the better tool for anyone wanting exposure with zero deliberate study time, and conceded the ambient category to it. That concession no longer holds, because LingoBlend now blends pages in the browser.

You read a page in a language you already know and a share of its words are swapped in place into the language you are learning. You infer them from context, click one to see the original, and save it. The differences from Toucan are mostly about control and safety:

• A slider from 5% to 100%. You set the density. At 10% it is ambient. At 60% it is a reading session.
• Viewport-driven. It blends what is on screen and keeps up as you scroll, so page length does not matter.
• Click for the full picture. A popover shows the target word, the original beneath it, and "Save word". Pro adds "Base form", which gives the dictionary form plus a grammar note, so escogido resolves to escoger with the form labelled.
• Copy returns the originals. Select and copy blended text and you get your own language back, so nothing strange lands in an email.
• Off by default, everywhere. Turning blending on triggers Chrome's permission prompt for page access; the install itself asks for nothing broad. You can also hit "Blend this page" for a single tab and grant nothing.
• Off by default on tools you work in. Gmail, Docs, Drive, Notion, Slack, GitHub, GitLab, Overleaf, Figma, LinkedIn, WhatsApp Web, Teams, Outlook and Discord. It never touches text you are typing, form fields, or code blocks.
• It checks the page language rather than trusting the markup, and refuses to blend a page already in the language you are learning.
• It skips words it is not confident about instead of risking a wrong one.

The deliberate half stays in the mobile app: paste an article, a chapter, a PDF or a URL, blend it, read it in a paginated reader, and every saved word feeds five practice games on an Anki-style SM-2 schedule. Ambient in the browser, deliberate in the evening, review the next morning.

Where the translation happens#

This is the part that is easy to skip and probably matters most.

LingoBlend blends pages using Chrome's built-in on-device translator, available in Chrome 138 and later on desktop. Page text never leaves the machine. It is not uploaded, not logged, and not sent to LingoBlend or to anyone else. The extension contains no analytics and no telemetry. The only thing that ever leaves your device is a single word you explicitly choose to save, plus your sign-in.

Credit where it is due: if the local-dictionary inference above is right, Toucan was also doing its lookups locally, which means it had the same privacy property. It simply never made an argument out of it. Anyone evaluating browser extensions that read every page should be asking this question of every candidate, and most of them answer it worse than either of these two.

Toucan and LingoBlend, side by side#

Feature sets change on both sides. Check each product's own listing before deciding.

The limits we share, stated plainly#

It would be dishonest to present LingoBlend as having solved Toucan's platform problem, because we have the same one.

Blending is desktop Chrome and Edge only. There is no mobile browser version, no Firefox, and no Safari. It needs Chrome 138 or later, because that is where the built-in translator lives; right-click word saving still works on any recent Chrome. The first time you use a language pair you download a language pack once from the popup, after which it works offline. And Serbian does not blend at all, because Chrome's on-device translator does not support it, so Serbian learners get right-click saving and the mobile app but no page blending.

If you need Firefox or Safari, neither of these tools is your answer. The wider field is covered in browser extensions for language learners, and Readlang vs Toucan compares the active-lookup approach if you would rather read target-language pages and click what you do not know.

Moving over from Toucan#

The transition is smaller than it looks, because the daily interaction is nearly identical. Install the extension, sign in, pick your pair, and start at 10% or 15% so pages stay comfortable. Leave it off on the sites you work in; the defaults already do this for the obvious ones. Save the words you actually stop on rather than everything, since a review queue you dread is a review queue you abandon.

The part that is genuinely new is what happens afterward. Words you save in the browser show up in the app's dictionary and enter the review schedule, which is the encounter loop passive exposure alone cannot supply. If you want to see what a full blended text looks like before installing anything, the extension page shows the browser flow and the languages page has a real blended sentence with glosses for each of the 17.