Svara · design exploration · 2026 · round 4

Sixteen logged-out aesthetics for the same model card.

Same product. Same content. Sixteen fundamentally different page architectures. The product underneath is consistent: svara-global-v1, a 780M-parameter open-weights TTS model that holds a single voice across 50+ languages in one API call, with inline emotion / prosody tags and Apache 2.0 licensing. V1V5 were round 1, V6V10 were round 2 grounded in real provider playbooks, V11V15 are round 3 — variations on the V6 shape pitched at content teams, creators, and business buyers (Murf, Hume, Lovo, ElevenLabs verticals, Cartesia realtime). V16 is the Mistral / Voxtral-style model launch — audio-first hero, gradient-warm research-marketing tone, win-rate chart, cascaded translation diagram.

V1 / LIGHT / EDITORIAL
Vinyl Sleeve
Voices as a tracklist. Single-column print catalog.
Stones Throw / Apartamento. The model is a record, the voices are tracks, the specs are liner notes.
V2 / DARK / SIGNAGE
Departure Board
Solari split-flap. 50 voices as departing flights.
Industrial signage. Latency cells flip every couple of seconds. Click a row to "board" — its status flips through to PLAYING.
V3 / LIGHT / DEMO-FIRST
Try-It-First / IMAX
The fold is the playground.
No marketing preamble. Type, generate, hear. Narrative scroll lives below the fold.
type anything · हिन्दी · தமிழ்
GENERATE
V4 / DARK / PLANETARY
Voice Atlas
A rotating 3D globe. Every voice is a pin.
Cinematic edge-network feel. Drag to rotate, hover for tooltip, click a pin to fly the camera and play that voice.
V5 / LIGHT / SPATIAL
Workspace
A fake desktop OS. Draggable windows, dock, menubar.
Voices.app, NowPlaying, README, Terminal, Pricing.numbers — the model lives across windows you can rearrange.
Svara · File · Edit
Get started →
V6 / LIGHT / TRI-FORK
Studio · API · Enterprise
ElevenLabs / Murf shape. Three audiences, one page.
Hero with the 780M / 50+ langs trust strip, then three product tiles — creators get a Studio demo, developers get a code snippet, enterprises get a feature checklist.
One small voice model.
Every language. Every voice.
Studioplay voices in browser
API$ pip install svara
Enterpriseself-host, SOC 2
V7 / DARK / PLAYGROUND
Hero Playground
Deepgram Aura shape. The fold is the prompt.
Single dark playground card with inline [emotion] chips, voice list, and a vertical-tagged voice library below for agents / support / media / IVR.
"Welcome back. [warm] Glad to see you. [lang=es] Hola, ¿cómo estás?"
allagentssupportmediaivr
GENERATE
V8 / CREAM / DSL
Prompt is the interface
Cartesia / Fish shape. Inline tags as the whole story.
A single transcript card where every [lang=…] / [emotion] / [laugh] chip is clickable. One prompt → one stream → multilingual code-switched audio.
"[lang=en] Welcome back, [warm] glad to see you. [lang=hi] क्या मैं मदद कर सकती हूँ? [laugh] [pause=400]"
V9 / LIGHT / ROSTER
Voice Talent Gallery
WellSaid shape. Voices as portraits.
Sticky filter sidebar, gallery of avatar cards in 12 deterministic palettes. Each card carries language pill, style tag, latency, sample languages, and a play meter.
V10 / PAPER / RESEARCH
Model Card / Research Page
Hume · Inworld · HuggingFace shape. Trust through receipts.
Reads like an arXiv preprint. Abstract, capabilities table, 3 benchmark tables (MOS / latency / RTF), languages × emotions matrix, BibTeX citation. Subdued type, no marketing chrome.
svara-global-v1: a 780M-parameter multilingual TTS model
ModelParamsMOSELO
svara-global-v1780M4.421271
ElevenLabs Multi v2~3B4.391248
Cartesia Sonic~2B4.361235
V11 / LIGHT / USE-CASE SWITCHER
Built for your team
Murf · Play.ht shape. Switch the audience, switch the page.
Hero with a rotating tab — Marketing / E-learning / Audiobooks / Agents / Games / IVR. The big visual swaps to a real-feeling product mock per tab. Voice carousel, customer logos, end CTA.
Built for marketing teams
Sounds like your brand
MarketingE-learningAudiobooksAgents
Aurora Coffee — Spring Drop · 0:14
V12 / WARM / EMPATHY
A voice that listens back
Hume EVI · Sesame · Inworld shape. Conversation-first.
Big italic serif headline, animated pulse-mic, a real-time conversation transcript with emotion + language tags. Pitched at companions, agents, characters — feels human, not like stats.
Aria · 187ms
Stressed about tomorrow.
[warm] Take a breath. Walk me through it.
[lang=hi] मेरे आँकड़े कमज़ोर हैं
V13 / LIGHT / MARKETPLACE
Meet your cast
Lovo Genny · Speechify Studio shape. Voice talent agency.
Big italic serif "Meet your cast" hero with a stack of tilted persona cards. Filter by mood / use case / accent. Pricing tiers Free / Creator / Team / Business pitched at agencies, not devs.
Meet your cast.
52 voices.
V14 / DARK / VERTICALS
Industry long-scroll
ElevenLabs · Murf vertical pages shape. One section per industry.
Long scroll with six dedicated industry sections — Marketing / Audiobooks / Education / Agents / Games / IVR — each with its own visual treatment, copy, customer story, and CTA. Sticky filter rail.
One model.
Six industries.
MktAudioEduAgtGameIVR
V15 / DARK / REALTIME
Real-time voice, anywhere
Cartesia · Sesame · Resemble realtime shape. Latency as the headline.
Live ticking widget — pulsing mic, scope, P50 latency counter, regional latency map. Three product columns (WS / HTTP / SIP). Three playable agent transcripts. Two-line code panel.
Real-time voice.
Anywhere on Earth.
V16 / LIGHT / MODEL LAUNCH
Audio is the new UX.
Mistral · Voxtral shape. Research-marketing model launch with audio-first hero + win-rate chart.
Announcement pill, gradient-keyword display headline, three interactive sample cards as the hero, numbered features, voice emulation gallery, win-rate bar chart vs competitors, cascaded speech-to-speech diagram, three-stage architecture, dark footer.
Audio is the
new UX.
AshaHindi · Happy
AriaEnglish · Stern
MeeraTamil · Surprise