Hume AI

Read our latest blog post about voice AI data and building realistic speech modelsThe world's most realistic & expressive voice AIVoice AI models powered by emotional intelligence for creators, developers, and enterprises. Create audio books, podcasts, conversational agents and more.Text-to-SpeechVoice DesignVoice CloningConversational AIText-to-speechI'm your host, Alex Tran, and today—strap in—we’re talking about... robot chickens. Yeah, you heard me. Robot. Chickens. Now before your mind runs off to late-night cartoons or mad science labs, let me paint a picture. Imagine a fluffy, feathery bird—but under that? A network of sensors, tiny actuators, and maybe—just maybe—a Wi-Fi module. Welcome to the poultry revolution.Create a Podcast376/500LanguageEnglishVoicePodcast HostPlay audioGet started with the most expressive text-to-speechSign up for freeTrusted by teams atProductsBuild with emotional intelligence modelsOctaveText-to-speech with emotional intelligence. Generate expressive, natural speech.Text to SpeechEmpathic Voice InterfaceEmpathic voice interface for conversations. Build AI that listens and responds with care.Speech to SpeechExpression MeasurementAnalyze emotions from face and voice. Understand how people truly feel at scale.MultimodalCapabilitiesVoice AI that handles the hard partsVoice CreationDesign voices with wordsDescribe the voice you want in natural language and our AI creates it. No voice actors needed—just your imagination.“The speaker has an expressive, totally disgusted Valley Girl voice, with a heavy Californian accent, delivering each word with maximum disdain, like a lifestyle influencer reacting to a truly tragic fashion faux pas.”Play“The speaker is a high-energy hype man with the contagious enthusiasm of a sports announcer, the rhythmic cadence of a seasoned rapper, and the irresistible charisma of a famous motivational speaker.”Play“The speaker has a boisterous, gravelly voice, like a grizzled old sea captain with a thick stereotypical pirate accent, perfect for recounting tales of daring raids and buried treasure, and is intensely charismatic.”PlayVoice CloningClone any voice instantlyCreate a natural-sounding voice clone from just a few seconds of audio. Play the original, then hear the AI clone.ItoAlfredAudreyOriginalCloneCross LingualOne voice, any languageMaintain consistent voice identity across 100+ languages. The same voice can speak English, Mandarin, Spanish, and more with native-level pronunciation.DeutschEspañolItalianoPortuguês한국어日本語العربيةEnglishActing InstructionsDirect the performanceAdd stage directions to guide delivery. Whisper, shout, pause for effect—your voice does exactly what you tell it to.“With warm enthusiasm”Play“Speak slowly and in a whisper”Play“Speak with a sarcastic tone”PlayTGTurtle GuruATAunt TeaSGSitcom GuyUse CasesGenerate life-like AI audio for your content creation needsView customersAudiobooksCreate high-quality multi-character audiobooks. Upload your PDF, select your characters, direct delivery and publish.Video voiceoversChoose the perfect voice for your video or clone your own voice. Then generate high-quality voiceovers for ads, shorts, or feature-length films. PodcastsCreate multi-speaker podcasts that sound like real, studio quality dialogue. Select your voices, generate audio, and download.'TwasthenightbeforeChristmas,whenallthroughthehouse.Notacreaturewasstirring,notevenamouse.Thestockingswerehungbythechimneywithcare.InhopesthatStNicholassoonwouldbethere.PlayI'm,like,stunned.It'stotallyinsanetome,honestly.Like,can'ttheyjustwaitoneday?PlaySohe,uh,hegoesdownthiscrazyrathole.Imean,pictureit.It's1954.He'sstandingbarefootonthebackofallamaandheyellsatthetopofhislungs-PlayAudiobooksCreate high-quality multi-character audiobooks. Upload your PDF, select your characters, direct delivery and publish.'TwasthenightbeforeChristmas,whenallthroughthehouse.Notacreaturewasstirring,notevenamouse.Thestockingswerehungbythechimneywithcare.InhopesthatStNicholassoonwouldbethere.PlayVideo voiceoversChoose the perfect voice for your video or clone your own voice. Then generate high-quality voiceovers for ads, shorts, or feature-length films. I'm,like,stunned.It'stotallyinsanetome,honestly.Like,can'ttheyjustwaitoneday?PlayPodcastsCreate multi-speaker podcasts that sound like real, studio quality dialogue. Select your voices, generate audio, and download.Sohe,uh,hegoesdownthiscrazyrathole.Imean,pictureit.It's1954.He'sstandingbarefootonthebackofallamaandheyellsatthetopofhislungs-PlayBuilt on ScienceDecades of research, one platformDynamic ReactionExplore dataset#1in naturalness and expressivity600+tagsof emotions and voice characteristics detected250msspeech LLM latencyFor DevelopersBuild in minutes, scale foreverTypeScriptPython.NETSwiftindex.tsCopyimport { HumeClient } from 'hume' const client = new HumeClient({ apiKey: 'YOUR_API_KEY' }) await client.tts.synthesizeFileStreaming({ utterances: [ { text: 'Dogs became domesticated between 23,000 and 30,000 years ago.', voice: { name: 'Male English Actor', provider: 'HUME_AI' }, }, ], })DocumentationComprehensive guides, tutorials, and API references to get you building fast.Open SourceSDKs, examples, and tools—all open source on GitHub.From the BlogLatest updatesView all postsProduct UpdatesBuilding Voice Models Is No Longer a Modeling ProblemWhat’s changed isn’t just where voice is used, but what it represents. Voice is no longer a feature layered on top of an intelligent system. It’s becoming a foundational modality through which models reason, interact, and are judged by users.Product UpdatesOctave 2: next-generation multilingual voice AIToday we’re launching Octave 2, the second generation of our frontier voice AI model for text-to-speech. We just made a preview of Octave 2 available on our platform and through our API.Product UpdatesIntroducing EVI 3: the world’s most realistic and instructible speech-to-speech foundation modelAt Hume, we promised ourselves that before the end of 2025, we’d achieve a voice AI experience that can be fully personalized. We believe this is an essential step toward voice being the primary way people want to interact with AI.View all postsReady to build with empathy?Start building AI that understands human emotion. Free to get started, with usage-based pricing as you scale.Get started freeView pricingStay in the loopGet the latest on empathic AI research, product updates, and company news.SubscribeJoin the communityConnect with other developers, share projects, and get help from the team.Join our Discord --- PricingPricing that works at any scaleChoose an affordable plan that’s packed with the best features for engaging your audience, creating customer loyalty, and driving sales.contact salessign up for freePlanFree$0 / monthSign upStarter$3 / monthSign up1st month 50% offCreator$7 / month$14 / monthSign upPro$70 / monthSign upScale$200 / monthSign upBusiness$500 / monthSign upEnterpriseCustomContact usText-to-speechOctave 1Octave 2Monthly included characters10,000(~10 minutes)30,000(~30 minutes)140,000(~140 minutes)1,000,000(~1,000 minutes)3,300,000(~3,300 minutes)10,000,000(~10,000 minutes)As much as you needAdditional characters cost (Usage-based)——$0.15/1,000$0.12/1,000$0.10/1,000$0.05/1,000CustomRPM (requests per minute)15157575150225CustomProjects—201,0003,00010,00020,000As much as you needVoice conversionCommercial license——Speech-to-speechEVI 3EVI 4 miniMonthly EVI usage included5 minutes40 minutes($0.07/minute)200 minutes($0.07/minute)1,200 minutes($0.06/minute)5,000 minutes($0.05/minute)12,500 minutes($0.04/minute)As much as you needAdditional EVI 3 cost (Usage-based)$0.06/minute$0.05/minute$0.04/minuteCustomExternal LLMsConcurrent connections155102030As much as you needVoicesVoice cloningCreate onlyCreate onlyUnlimited (create and use)Unlimited (create and use)Unlimited (create and use)Unlimited (create and use)Unlimited (create, use and access via API)Team collaborationOrganization————Team seats————35UnlimitedSupport & ComplianceSupportDiscordDiscordDiscordDiscordDiscordDiscordCompliance——————Free$0 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters10,000(~10 minutes)Additional characters cost (Usage-based)—RPM (requests per minute)15Projects—Voice conversionCommercial license—Speech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage5 minutesAdditional EVI 3 cost (Usage-based)External LLMsConcurrent connections1VoicesVoice cloningCreate onlyTeam collaborationOrganization—Team seats—Support & ComplianceSupportDiscordCompliance—Starter$3 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters30,000(~30 minutes)Additional characters cost (Usage-based)—RPM (requests per minute)15Projects20Voice conversionCommercial license—Speech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage40 minutes($0.07/minute)Additional EVI 3 cost (Usage-based)External LLMsConcurrent connections5VoicesVoice cloningCreate onlyTeam collaborationOrganization—Team seats—Support & ComplianceSupportDiscordCompliance—1st month 50% offCreator$7 / month$14 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters140,000(~140 minutes)Additional characters cost (Usage-based)$0.15/1,000RPM (requests per minute)75Projects1,000Voice conversionCommercial licenseSpeech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage200 minutes($0.07/minute)Additional EVI 3 cost (Usage-based)External LLMsConcurrent connections5VoicesVoice cloningUnlimited (create and use)Team collaborationOrganization—Team seats—Support & ComplianceSupportDiscordCompliance—Pro$70 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters1,000,000(~1,000 minutes)Additional characters cost (Usage-based)$0.12/1,000RPM (requests per minute)75Projects3,000Voice conversionCommercial licenseSpeech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage1,200 minutes($0.06/minute)Additional EVI 3 cost (Usage-based)$0.06/minuteExternal LLMsConcurrent connections10VoicesVoice cloningUnlimited (create and use)Team collaborationOrganization—Team seats—Support & ComplianceSupportDiscordCompliance—Scale$200 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters3,300,000(~3,300 minutes)Additional characters cost (Usage-based)$0.10/1,000RPM (requests per minute)150Projects10,000Voice conversionCommercial licenseSpeech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage5,000 minutes($0.05/minute)Additional EVI 3 cost (Usage-based)$0.05/minuteExternal LLMsConcurrent connections20VoicesVoice cloningUnlimited (create and use)Team collaborationOrganizationTeam seats3Support & ComplianceSupportDiscordCompliance—Business$500 / monthSign upText-to-speechOctave 1Octave 2Monthly included characters10,000,000(~10,000 minutes)Additional characters cost (Usage-based)$0.05/1,000RPM (requests per minute)225Projects20,000Voice conversionCommercial licenseSpeech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usage12,500 minutes($0.04/minute)Additional EVI 3 cost (Usage-based)$0.04/minuteExternal LLMsConcurrent connections30VoicesVoice cloningUnlimited (create and use)Team collaborationOrganizationTeam seats5Support & ComplianceSupportDiscordCompliance—EnterpriseCustomContact usText-to-speechOctave 1Octave 2Monthly included charactersAs much as you needAdditional characters cost (Usage-based)CustomRPM (requests per minute)CustomProjectsAs much as you needVoice conversionCommercial licenseSpeech-to-speechEVI 3EVI 4 miniMonthly EVI usage includedThese are additional EVI minutes included with your plan on top of the specified amount of TTS usageAs much as you needAdditional EVI 3 cost (Usage-based)CustomExternal LLMsConcurrent connectionsAs much as you needVoicesVoice cloningUnlimited (create, use and access via API)Team collaborationOrganizationTeam seatsUnlimitedSupport & ComplianceSupportComplianceExpression Measurement pricingFeaturePay as you goEnterpriseVideo with audioFacial expression, Speech prosody, Vocal burst, Emotional language, Facemesh, Transcription$0.0828 / minVolume discountsAudio onlySpeech prosody, Vocal burst, Emotional language, Transcription$0.0639 / minVolume discountsVideo onlyFacial expression, Facemesh$0.045 / minVolume discountsImagesFacial expression, Facemesh$0.00204 / imageVolume discountsText onlyEmotional language$0.00024 / wordVolume discountsPay as you goVideo with audioFacial expression, Speech prosody, Vocal burst, Emotional language, Facemesh, Transcription$0.0828 / minAudio onlySpeech prosody, Vocal burst, Emotional language, Transcription$0.0639 / minVideo onlyFacial expression, Facemesh$0.045 / minImagesFacial expression, Facemesh$0.00204 / imageText onlyEmotional language$0.00024 / wordEnterpriseVideo with audioFacial expression, Speech prosody, Vocal burst, Emotional language, Facemesh, TranscriptionVolume discountsAudio onlySpeech prosody, Vocal burst, Emotional language, TranscriptionVolume discountsVideo onlyFacial expression, FacemeshVolume discountsImagesFacial expression, FacemeshVolume discountsText onlyEmotional languageVolume discountsStay in the loopGet the latest on empathic AI research, product updates, and company news.SubscribeJoin the communityConnect with other developers, share projects, and get help from the team.Join our Discord --- Hume AI is pioneering the development of artificial intelligence that understands and responds to human emotions.Our VisionWe envision a future where AI enhances human connection rather than replacing it—where technology understands not just what we say, but how we feel.Today's AI systems are remarkably capable at processing language and generating content, but they remain fundamentally disconnected from the emotional dimension of human experience.We're changing that. By combining cutting-edge machine learning with decades of scientific research on human expression, we're creating AI that can perceive, understand, and respond to the full spectrum of human emotion.Watch videoOur ValuesBeneficenceAI should be deployed only if its benefits substantially outweigh its costs.EmpathyAI privy to cues of our emotions should serve our emotional well-being.Scientific LegitimacyApplications of AI should be supported by collaborative, rigorous, inclusive science.Emotional PrimacyAI should be prevented from treating human emotion as a means to an end.InclusivityThe benefits of AI should be shared by people from diverse backgrounds.TransparencyPeople affected by AI should have enough data to make decisions about its use.ConsentAI should be deployed only with the informed consent of the people whom it affects.Learn more at The Hume InitiativeOur ResearchWe are addressing the AI alignment problem by building AI systems that measure and optimize for human emotional well-being.Our ProductsVoice AI models powered by emotional intelligence for creators, developers, and enterprises.Our Academic OriginsA history of emotion scienceWe're continuing the legacy of emotion science and bringing it into the next era with AI.1739 | David HumeHume argues that emotions drive choice and well-beingAt Hume AI, we take this as a guiding principle behind ethical AI: in order to serve our preferences, algorithms should be guided by our emotions.Recognizing the need to map out the emotions that animate thought and action, Hume also proposed a taxonomy of over 16 emotional states, but lacked scientific evidence.1872 | Charles DarwinDarwin surveys human emotionCharles Darwin described similarities and differences in over 20 facial, bodily, and vocal expressions across species, cultures, and stages of life. The Expression of the Emotions in Man and Animals was his third major work.He lacked statistical methods to test his hypotheses about human emotion. But 150 years later, studies are confirming many of Darwin's observations.1969 | The Basic 6Ekman documents six facial expressionsPaul Ekman traveled the world to find that six expressions are universally recognized. By focusing on a narrow set of behaviors, Ekman was able to use the statistical methods available to him to confirm some of Darwin's ideas.However, the focus on just six emotions also introduced what we call the 30% problem: the focus of scientists for 50 years on only 30% of the full range of emotions people experience.1969 | Valence And ArousalScientists try to reduce human emotionWhile many scientists focus on six emotions, others attempt to derive taxonomy of emotion from data. However, due to statistical limitations, these results lead to even more reductive theories of emotion.Some scientists endorse "core affect": the notion that emotions are largely captured by how pleasant or unpleasant and calm or aroused an experience or expression seems.Today | Our ApproachThe full spectrum of emotionHume's scientists are revolutionizing emotion study through data-driven methods, employing computational techniques, and analyzing vast datasets to understand the spectrum of human emotions.They've gathered millions of reactions across videos, music, and art, studied brain mechanisms of emotion, explored ancient sculptures' expressions, and applied deep learning to global video expressions. Their research uncovers over 30 emotion dimensions.1739 | David HumeHume argues that emotions drive choice and well-beingAt Hume AI, we take this as a guiding principle behind ethical AI: in order to serve our preferences, algorithms should be guided by our emotions.Recognizing the need to map out the emotions that animate thought and action, Hume also proposed a taxonomy of over 16 emotional states, but lacked scientific evidence.1872 | Charles DarwinDarwin surveys human emotionCharles Darwin described similarities and differences in over 20 facial, bodily, and vocal expressions across species, cultures, and stages of life. The Expression of the Emotions in Man and Animals was his third major work.He lacked statistical methods to test his hypotheses about human emotion. But 150 years later, studies are confirming many of Darwin's observations.1969 | The Basic 6Ekman documents six facial expressionsPaul Ekman traveled the world to find that six expressions are universally recognized. By focusing on a narrow set of behaviors, Ekman was able to use the statistical methods available to him to confirm some of Darwin's ideas.However, the focus on just six emotions also introduced what we call the 30% problem: the focus of scientists for 50 years on only 30% of the full range of emotions people experience.1969 | Valence And ArousalScientists try to reduce human emotionWhile many scientists focus on six emotions, others attempt to derive taxonomy of emotion from data. However, due to statistical limitations, these results lead to even more reductive theories of emotion.Some scientists endorse "core affect": the notion that emotions are largely captured by how pleasant or unpleasant and calm or aroused an experience or expression seems.Today | Our ApproachThe full spectrum of emotionHume's scientists are revolutionizing emotion study through data-driven methods, employing computational techniques, and analyzing vast datasets to understand the spectrum of human emotions.They've gathered millions of reactions across videos, music, and art, studied brain mechanisms of emotion, explored ancient sculptures' expressions, and applied deep learning to global video expressions. Their research uncovers over 30 emotion dimensions.Join our missionWe're always looking for talented people who share our vision of building AI that truly understands humanity.View open rolesStay in the loopGet the latest on empathic AI research, product updates, and company news.SubscribeJoin the communityConnect with other developers, share projects, and get help from the team.Join our Discord --- ResearchThe science of emotionExplore our publications, models, and datasets pushing the boundaries of empathic AI.#1in naturalness and expressivity600+tagsof emotions and voice characteristics detected250msspeech LLM latencyState-of-the-art performance across all benchmarksPerformanceState-of-the-art resultsOur models consistently achieve top performance across industry benchmarks.NaturalnessMost natural voice conversationsIn blind comparisons, users consistently rate Hume voices as more natural and human-like than alternatives.•Authentic speech rhythms and pauses•Natural intonation patterns•Human-like breathing and cadenceNaturalness Score (higher is better)EmpathySuperior emotional understandingHume's empathic AI demonstrates significantly higher emotional awareness and appropriate responses in conversations.•Recognizes frustration and responds with patience•Detects excitement and matches energy•Senses uncertainty and offers reassuranceEmpathy Score (higher is better)ExpressivenessMost expressive voice AIHume voices convey a wider range of emotions and nuanced expressions compared to other voice AI providers.•Warm enthusiasm for good news•Gentle concern when discussing problems•Playful humor in casual momentsExpressiveness Score (higher is better)Hard InputsBest pronunciation of challenging contentOur TTS excels at pronouncing difficult content like phone numbers and mathematical expressions that trip up other systems.The local mycologist explained that consuming just one fourth plus one fourth equals one half ounce of the misidentified death caps could prove fatal within forty eight hours.Most businesses close in the late afternoon from between two until four thirty or five o'clock when it can get hot.On december fifteenth two thousand seven, Dennis Kucinich raised one hundred thirty one thousand four hundred dollars from approximately one thousand six hundred donors.Pass Rate by Input Type (higher is better)Tested on 2,167 samplesEmotion RecognitionMost accurate emotion identificationWhen listeners rate how well they can identify the intended emotion, Hume voices consistently outperform competitors.•Joy, sadness, anger, fear, surprise•Subtle cues like hesitation or relief•Complex emotions like bittersweet nostalgiaDistressed1 / 8Identification score (higher is better)Instruction FollowingPrecisely follows your vocal directionsWhen you ask for a specific vocal style, emotion, or character, Hume delivers exactly what you requested.•"Speak with a whisper, like sharing a secret"•"Sound excited and out of breath"•"Use a sarcastic, know-it-all tone"Instruction Following (higher is better)Tested across 32 vocal instructionsPerformanceState-of-the-art resultsOur models consistently achieve top performance across industry benchmarks. Every claim is backed by rigorous evaluation and reproducible methodology.NaturalnessMost natural voice conversationsIn blind comparisons, users consistently rate Hume voices as more natural and human-like than alternatives.•Authentic speech rhythms and pauses•Natural intonation patterns•Human-like breathing and cadenceEmpathySuperior emotional understandingHume's empathic AI demonstrates significantly higher emotional awareness and appropriate responses in conversations.•Recognizes frustration and responds with patience•Detects excitement and matches energy•Senses uncertainty and offers reassuranceExpressivenessMost expressive voice AIHume voices convey a wider range of emotions and nuanced expressions compared to other voice AI providers.•Warm enthusiasm for good news•Gentle concern when discussing problems•Playful humor in casual momentsHard InputsBest pronunciation of challenging contentOur TTS excels at pronouncing difficult content like phone numbers and mathematical expressions that trip up other systems.The local mycologist explained that consuming just one fourth plus one fourth equals one half ounce of the misidentified death caps could prove fatal within forty eight hours.Most businesses close in the late afternoon from between two until four thirty or five o'clock when it can get hot.On december fifteenth two thousand seven, Dennis Kucinich raised one hundred thirty one thousand four hundred dollars from approximately one thousand six hundred donors.Emotion RecognitionMost accurate emotion identificationWhen listeners rate how well they can identify the intended emotion, Hume voices consistently outperform competitors.•Joy, sadness, anger, fear, surprise•Subtle cues like hesitation or relief•Complex emotions like bittersweet nostalgiaInstruction FollowingPrecisely follows your vocal directionsWhen you ask for a specific vocal style, emotion, or character, Hume delivers exactly what you requested.•"Speak with a whisper, like sharing a secret"•"Sound excited and out of breath"•"Use a sarcastic, know-it-all tone"Recent PublicationsPeer-reviewed insightsView allarXiv·Feb 2026TADA: A Generative Framework for Speech Modeling via Text-Acoustic Dual Alignment (Under Review)TDSRAG+6Trung Dang, Sharath Rao, Ananya Gupta and 6 moreModern Text-to-Speech (TTS) systems increasingly leverage Large Language Model (LLM) architectures to achieve scalable, high-fidelity, zero-shot generation. However, these systems typically rely on fixed-frame-rate acoustic tokenization, resulting in speech sequences that are significantly longer than, and asynchronous with their corresponding text. Beyond computational inefficiency, this sequence length disparity often triggers hallucinations in TTS and amplifies the modality gap in spoken language modeling (SLM). In this paper, we propose a novel tokenization scheme that establishes one-to-one synchronization between continuous acoustic features and text tokens, enabling unified, single-stream modeling within an LLM. We demonstrate that these synchronous tokens maintain high-fidelity audio reconstruction and can be effectively modeled in a latent space by a large language model with a flow matching head. Moreover, the ability to seamlessly toggle speech modality within the context enables text-only guidance--a technique that blends logits from text-only and text-speech modes to flexibly bridge the gap toward text-only LLM intelligence. Experimental results indicate that our approach achieves performance competitive with state-of-the-art TTS and SLM systems while virtually eliminating content hallucinations and preserving linguistic integrity, all at a significantly reduced inference cost.View paperDownload PDFFrontiers in Psychology·May 2024How emotion is experienced and expressed in multiple cultures: a large-scale experiment across North America, Europe, and JapanGP+13Alan Cowen, Jeffrey Brooks, Gautam Prasad and 13 moreCore to understanding emotion are subjective experiences and their expression in facial behavior. Past studies have largely focused on six emotions and prototypical facial poses, reflecting limitations in scale and narrow assumptions about the variety of emotions and their patterns of expression. View paperDownload PDFiScience·Feb 2024Deep learning reveals what facial expressions mean to people in different culturesLKMO+10Jeffrey Brooks, Lauren Kim, Michael Opara and 10 moreCross-cultural studies of the meaning of facial expressions have largely focused on judgments of small sets of stereotypical images by small numbers of people. Here, we used large-scale data collection and machine learning to map what facial expressions convey in six countries. View paperDownload PDFEverything your model needsWhy Our DatasetsWorld-class data for pre-training and fine-tuning your emotion AI models, backed by years of scientific research.Contact usEthically SourcedAll data collected with informed consent and rigorous privacy protections.Globally DiverseRepresentative samples across cultures, ages, genders, and demographics.Expert AnnotatedLabeled by trained researchers using validated scientific frameworks.Research ReadyClean, structured formats optimized for modern ML pipelines.Research AreasWhere Hume enables researchFrom fundamental affective computing to applied behavioral research, our tools power studies across the full spectrum of emotion science.Affective ComputingStudy how AI systems can recognize, interpret, and respond to human emotions across modalities.Human-AI InteractionResearch the dynamics of emotional exchange between humans and AI systems.Psychology & BehaviorUse emotion recognition to study human behavior, mental health, and psychological phenomena.Speech & LanguageAnalyze prosodic features, sentiment, and emotional expression in human communication.Multimodal LearningExplore how emotion manifests simultaneously across face, voice, and language.Ethics & AI SafetyStudy the ethical implications of emotionally-aware AI systems and develop guidelines.From the BlogLatest research updatesView allResearchOpensourcing TADA: Fast, Reliable Speech Generation Through Text-Acoustic SynchronizationTADA (Text-Acoustic Dual Alignment) is Hume AI's open-source speech-language model that synchronizes text and audio one-to-one. Mar 10, 2026ResearchIntroducing OCTAVE (Omni-Capable Text and Voice Engine)A frontier speech-language model with new emergent capabilities, like on-the-fly AI voice and personality creation.Dec 23, 2024ResearchHow can emotionally intelligent voice AI support our mental health?Recent advances in voice-to-voice AI, like EVI 2, offer emotionally intelligent interactions, picking up on vocal cues related to mental and physical health, which could enhance both clinical care and daily well-being.Oct 22, 2024Partner with us on researchAccess world-class datasets and collaborate with our team on advancing emotion AI.Contact researchView publicationsStay in the loopGet the latest on empathic AI research, product updates, and company news.SubscribeJoin the communityConnect with other developers, share projects, and get help from the team.Join our Discord