← Back to Talio Homepage
What's new
Changelog
Follow along as we ship improvements to help learners practice speaking and see how they are progressing. Dates below are when each release reached learners.
Conversation playback in transcripts, and a new way to see progress group by group across every metric.
Analytics
- NewProgress by group — Each metric panel gains a view setting a user's earliest conversations against their most recent, group by group, on one shared axis — Pronunciation, Mean Length of Run, Speaking Speed, and Lexical Diversity.
- ImprovedTwo views per metric panel — Every panel now offers an overall trend and progress by group.
- ImprovedA fuller can-do checklist — The proficiency checklist covers the complete set of interpersonal statements Talio grades at B2.
- ImprovedPronunciation scored on the turn it belongs to — Each turn's pronunciation score is computed from the audio of that turn.
- ImprovedPronunciation assessment improvements across languages.
- FixedSpeaking Speed is measured over a pooled window, so a very short turn cannot produce an implausible rate.
- FixedTone feedback is produced for Mandarin and Cantonese conversations.
UX
- NewTranscript audio playback — Play a whole conversation back from its transcript, or jump straight to any AI turn, with a scrubber showing elapsed and total time.
- NewRetry Scenario — Relaunch the exact scenario behind a past conversation from My Transcripts.
- ImprovedWider scenario choice for a can-do statement — Browsing scenarios for a can-do statement includes the levels either side of it, so there is more to practice with.
Other
- NewSpeaking time on a course card — Students enrolled in a course see how long they have spoken in it.
- ImprovedAll Time, This Week and Weekly views — A course's practice summary and its class panel each offer the three ranges.
- FixedConversations start only in languages Talio is set up to serve end to end.
Fourteen more languages get conversation partners written for them.
Conversational AI
- NewConversation partners in fourteen more languages — Each with a regional character of their own, a speaking voice, and feedback written for that language.
- FixedConversations start in Yoruba, Bangla and Tajik.
Analytics
- ImprovedReference audio for German, Spanish and Cantonese — Pronunciation feedback in these languages compares a user's speech against a native recording.
- ImprovedOne speaking-rate calculation — Speaking Speed is computed the same way on every surface that shows it.
- ImprovedCantonese reference audio matches Traditional-character Cantonese.
UX
- ImprovedThe teaching code stays hidden until asked for — A teacher reveals their course's teaching code from their own course card, so it stays off screen during a screen share.
Other
- NewA free 14-day trial, agreed up front — New users read the free-trial terms and agree to them before reaching the app.
- ImprovedSign-in improvements — closing the sign-in window is treated as a cancellation.
- FixedA brand-new account can start a free trial straight after signing up.
Courses arrive — students join by code, teachers see their class — alongside ten more languages.
Analytics
- ImprovedSpeaking-time measurement improvements.
- FixedEvery spoken turn appears in the combined transcript.
- FixedEnd-of-conversation results are saved as a conversation closes.
UX
- NewMy Courses — Students and teachers see every course they belong to, and their role in each.
- NewJoin a course with a code — A course code enrolls a student, with the disclosure of what is shared, and with whom, shown before they agree to join.
- NewCreate a section — A teacher creates a section of their course and receives a join code to hand out.
- NewThe class list — A teacher sees who is enrolled in a section.
- NewPractice time by student — A teacher sees how long each student in a section has practiced, per language.
- NewA join-a-course step at sign-up — New users can enter a course code while getting set up, or skip it.
Other
- NewTen more languages — Yoruba, Bangla, Hindi, Swahili, Urdu, Tajik and Ukrainian, plus Egyptian, Tunisian and Levantine Arabic — all in Alpha.
- ImprovedA server-managed language list — Which languages are offered, and whether each is Alpha or Beta, are now served by Talio, so the list can change without an app release.
- ImprovedCourse-aware language picker — For a student enrolled in a course, the picker names the course and offers the languages it covers.
- FixedEgyptian, Tunisian and Levantine Arabic conversations run in the dialect chosen, and pronunciation is scored against it.
- FixedAccess codes are checked against the course term and the access window they carry.
Analytics
- FixedTone scores use one scale wherever they appear.
- FixedCantonese tone scoring is measured against Cantonese references.
Other
- ImprovedSecurity and privacy improvements — Ongoing work across Talio's app and web surfaces as part of our regular security review.
- FixedAn access code entered while getting set up grants access, and names the plan and access window it covers.
Analytics
- ImprovedFaster results after a conversation — Analysis and feedback arrive sooner once a conversation ends.
- FixedA saved word keeps its heard and used counts when it is saved again.
UX
- ImprovedTranscript proficiency card now carries an explanatory summary.
Other
- ImprovedSecurity and privacy improvements — Ongoing work across Talio's web surfaces as part of our regular security review.
A significant accuracy correction to speaking rate, plus per-language goals and faster analytics.
Conversational AI
- ImprovedPlain-language feedback — Explanations use everyday wording.
- ImprovedFewer, better suggestions — A shorter, prioritized set of suggestions, with examples drawn from users' own speech.
Analytics
- NewA target level for each language — Users can set a different proficiency goal per language they study. Conversations, suggestions, and recommendations all follow the right one.
- NewWord choice as its own score — Word choice now appears as a scored band alongside pronunciation, clarity, and speaking rate.
- ImprovedFaster analytics — Vocabulary and analytics views load substantially faster, and results after a conversation arrive sooner.
- ImprovedSpeaking-rate accuracy improvements.
- ImprovedVocabulary counting improvements.
- ImprovedLanguage handling improvements in analysis.
Other
- ImprovedConversation start reliability improvements.
- ImprovedImproved reliability when saving a user's target speaking level.
Conversational AI
- NewEnd-of-scenario prompt — When a user has accomplished what a scenario set out, Talio offers to wrap up.
Analytics
- NewGrammar insights — A new Grammar view charting how users' word forms and tenses develop over time, with coaching cards naming specific habits — like leaning on the present tense — and a jump straight into scenarios that practice them.
- NewConfidence bands on analytics — Charts now show how much data a number is based on, so users can tell a firm trend from an early one.
- ImprovedWeekly changes are shown honestly — A week where mastery went down is displayed as a negative, not hidden. Metrics with no data say so explicitly rather than quietly disappearing.
UX
- Improved"Speaking level" replaces "ACTFL" — Plain-language naming everywhere the concept appears.
Other
- NewSlovenian — Added to the language picker in Alpha.
- NewA rewritten homepage — Talio's public site now leads with the measurements behind the practice, and a single self-serve way to try it.
- ImprovedSign-in reliability improvements on Firefox.
- ImprovedLanguage preference improvements.
- FixedSpeaking-level goals are set per language.
Conversational AI
- NewOne scenario, many conversations — Many scenarios let users choose the setting, role, or topic before starting — so the same scenario yields a different conversation each time. Left blank, Talio picks.
- NewNamed characters — Every conversation partner is a named character with their own voice and regional background.
- ImprovedMore natural conversations — Additional tuning, with particular attention to advanced scenarios.
- ImprovedChinese word-choice improvements.
- FixedWord-choice feedback extended to Traditional-character writing.
Analytics
- NewScenario suggestions — Recommended scenarios arrive in three flavors — master a skill, try something new at the user's level, or reinforce what they already know — each naming the skill it targets.
- FixedFiltered out conversations that contain no user dialogue.
UX
- NewConversation context before starting — Users see who they are talking to, where they are, and their conversation objectives before starting.
- NewReport bad data — Users can provide feedback on transcripts and suggestions.
- NewStarred scenarios — Users keep the ones they return to in a section of their own.
- ImprovedCorrections land on the words they refer to — Word-choice suggestions are marked inline in the transcript rather than listed separately.
Other
- FixedMaintenance and code cleanup.
The largest release in this window — a redesign, a new way to track progress, and eight more languages.
Conversational AI
- NewScenario objectives — Every scenario states its objectives, conversations are graded against them, and completing them advances the scenario's difficulty ladder.
- NewUser-written scenarios — Users describe what they want to practice and get calibrated options back, with a name, summary, roles, and adjustable objectives.
Analytics
- NewProficiency tracking — Every conversation is graded against real-world "I can…" skills, which progress as users demonstrate them consistently over time.
- ImprovedCoaching instead of raw numbers — Proficiency measures are shown as below, at, or above the user's level, each with one plain sentence on what to do about it.
- ImprovedWord forms grouped by lemma — Inflected forms — run, ran, running — are grouped under a single lemma in My Vocabulary rather than counted separately.
- ImprovedChinese pronunciation detail improvements.
- ImprovedPronunciation result improvements.
- ImprovedChart rendering improvements.
- FixedClarity scores shown on every conversation.
UX
- NewTranscripts as full pages — Transcripts open as their own linkable page with an analytics pane, objectives, and stacked characters with pinyin.
- NewVocabulary definitions — Tap any word in a transcript for a definition.
- NewFlashcards — Saved words can be studied as flashcards, with examples drawn from users' own correct usage.
- ImprovedA full visual redesign — New color and type systems across Analysis, Vocabulary, Help, Settings, Scenario Creator, and Transcripts.
- ImprovedLive processing progress — Users watch their conversation being analyzed step by step, and jump straight to the finished transcript.
- ImprovedMobile web layout — Hints stack below the conversation on narrow screens, charts collapse to one column, overlays are capped to the viewport, and tap targets are larger.
- ImprovedDecorative sparklines are hidden from screen readers rather than announced as meaningless chart nodes.
Other
- NewEight more languages — French, German, Spanish, Portuguese, Cantonese, Turkish, Vietnamese, and Polish opened as target languages. Newer additions are marked Alpha or Beta.
- NewWeekly speaking goal — Users set a weekly minutes target and track progress against it.
- NewSelf-serve account deletion — Available directly in the app.
- ImprovedImprovements to saved language preferences.
- ImprovedEmail-link sign-in reliability improvements.
- ImprovedAI Tutor Settings improvements.
Conversational AI
- NewScenario Creator — Users build their own conversation scenarios and launch them like any other. My Scenarios lets them edit, relaunch, and delete.
- ImprovedUpgraded voice engine — Voice conversations moved to a newer speech model, with cleaner handling when a session cannot start.
Analytics
- NewVocabulary analytics — A full analytics surface above the word list — totals, an HSK-level breakdown, a first-heard trend, and panels for the words a user has actually started using.
- ImprovedVocabulary variety updated — The calculation covers short conversations and reflects how much users repeat the same words across sessions.
- ImprovedVocabulary count improvements.
- ImprovedMetric coverage improvements.
- FixedFull pinyin shown on multi-character words.
UX
- NewOne unified Talio — Talio collapsed into a single app whose tabs adapt to the user, rather than separate apps to choose between.
- NewBrowse by level — The scenario catalog is organized by level and theme, with filters for category, difficulty, and skill.
- NewPractice home — A suggested next session, a weekly-progress indicator, and a row of scenarios chosen for the user.
- FixedPersonal Analytics scrolls on all screen sizes.
Other
- NewSign in with an email link — Alongside Google sign-in.
- ImprovedNo more silent Chinese default — Vocabulary, analytics, and scenario authoring all resolve the language a user is actually studying.
- ImprovedGoogle sign-in reliability improvements.
- ImprovedTranscript audio playback improvements.
Conversational AI
- ImprovedHints give a scaffold, not the answer — A hint offers a partial sentence for the user to complete rather than a finished one, shown in both languages.
- ImprovedConversational AI improvements for scenarios — Scenarios open in character, and each one carries its own tone.
- ImprovedConversational AI improvements for beginners — The tutor's language is better matched to lower speaking levels.
- ImprovedFeedback calibration improvements.
Analytics
- NewClarity score — A measure of how understandable a user's speech was, shown on each message and on the conversation as a whole.
- ImprovedVocabulary variety improvements.
- ImprovedPronunciation trend chart improvements.
- FixedProficiency estimates require recorded speech.
UX
- NewPlayback of a user's own recording — Users play back what they said from any turn in a transcript.
- NewSpeaking level selection — Users set a speaking level for the tutor, and every new conversation adapts the tutor's language to it.
- ImprovedForm validation errors are shown as text, not signaled by color alone.
Other
- NewCantonese — Cantonese conversations with Jyutping romanization alongside Traditional characters, and Cantonese-specific feedback.
- NewSpanish and German — Each with its own grammar feedback, spoken audio, and starter scenario.
- ImprovedOnboarding improvements — New learners start at a level that matches them.
- ImprovedMicrophone detection improvements.
- FixedTranscript history spans every language a user studies.
Conversational AI
- NewTask-completion feedback — Users see whether they accomplished what the scenario set.
- NewIn-conversation hints — Users open a hint panel mid-conversation whenever they want a nudge.
Analytics
- ImprovedCharts show a reference range — So users can tell whether a number is good, not just what it is.
- ImprovedPronunciation result recovery improvements.
- FixedVocabulary variety plotted on the analysis charts.
UX
- NewRomanization for every language — Transcripts show romanized script alongside the native script, not only for Mandarin.
- NewMicrophone check — Talio confirms the microphone is working before a conversation starts, and reports any problem.
- ImprovedCleaner transcripts — Messages are grouped by speaker and short fragments combined, so a transcript reads like a conversation.
- ImprovedStar ratings are announced correctly to screen readers, menus carry proper labels, and tappable words in transcripts have larger targets.
- ImprovedProcessing notice improvements.
- FixedLong feedback explanations display in full.
Other
- NewPronunciation audio playback — Playback of the correct pronunciation for mispronounced words.
Conversational AI
- NewPre-conversation briefing — Users see the scenario, the task, and what is expected before starting.
Analytics
- NewMy Vocabulary — Save words from conversations to a personal vocabulary list, with active and passive usage ratios and vocabulary progress over time. Includes pinyin for Chinese.
- ImprovedOne pronunciation scale — Every pronunciation display uses the same 1–5 scale and the same colors, so a score means the same thing wherever it appears.
UX
- NewSettings page — Customize AI tutor preferences from within the app.
- NewSubmit feedback — Send feedback directly to the Talio team, with validation and guidance built in.