Echo 2.0
A real-looking signer, not an animated character. Echo 2.0 turns English into ASL video at 720p, from a vocabulary of 2,475 signs, with the smoothest motion of the systems we compared.
Our speech-to-sign model. Type or speak in English, and Echo signs it in ASL as a photorealistic person.
A real-looking signer, not an animated character. Echo 2.0 turns English into ASL video at 720p, from a vocabulary of 2,475 signs, with the smoothest motion of the systems we compared.
Our fastest speech-to-sign model, with streaming output for any avatar.
Join the waitlistOur first speech-to-sign model: spoken or written English, signed as photorealistic ASL video.
Read the postEcho 2.0 powers speech-to-sign and text-to-sign in Talksign products today, and it’s available to organisations through the Talksign API.
It signs American Sign Language (ASL) from English, from a vocabulary of 2,475 signs; other words are fingerspelled or replaced with a synonym. You choose the signer, Arthur or Amara today with Elena and Daniel coming soon, and every video comes with a matte, so the background can be any colour.
Once English has been put into ASL word order, typically in under 1.5 seconds, the signs are joined into one continuous video with smooth transitions between them.
To get started, request API access or email dev@talksign.co.
Echo turns English into ASL wherever Deaf people need to see it, across sectors and settings:
Sign what clinicians say at reception, in appointments and at the pharmacy counter, so patients follow every step.
Publish ASL versions of lessons, courses and school announcements, signed by the same familiar presenter.
Give signed answers at help desks, kiosks and in apps, any time, without waiting for an interpreter.
Sign public announcements, forms and guidance at council offices, transport hubs and service centres.
Add ASL to websites, videos, social posts and campaigns, with a signer and background that match your brand.
Speech-to-sign in Talksign apps, so Deaf and hearing people can talk face to face.
Echo 2.0 moves more smoothly from frame to frame than the published human-animation systems we compared, and its hands come close to a real signer’s.
Frame continuity against research systems
| Echo 2.0Talksign | SignMimicResearch | Animate AnyoneResearch | MimicMotionResearch | Magic PoseResearch | |
|---|---|---|---|---|---|
| SSIM-TShigher is better | 0.966 | 0.963 | 0.934 | 0.904 | 0.898 |
| PSNR-TShigher is better | 33.0 | 28.90 | 27.49 | 23.75 | 25.24 |
Identity similarity against avatar systems
| Echo 2.0Talksign | HeyGenAvatar system | KlingAvatar 2.0Avatar system | OmniHuman-1.5Avatar system | |
|---|---|---|---|---|
| IDChigher is better | 0.64 | 0.79 | 0.77 | 0.71 |
Hand accuracy and speed
Where Echo is still growing. Unlike avatar tools that only have to hold a face steady while it talks, Echo has to hold it steady while the hands and body move fast. Echo also follows ASL word order but makes limited use of signing space, classifiers and role shift, and its facial expressions come from each sign rather than from the meaning of the whole sentence.
Echo 2.0 figures come from 20 ASL signs held out from training (979 frames). Other systems’ values are from their papers: frame continuity from SignMimic (ASL 50K), identity from JoyStreamer. They were measured on different test sets from ours, so they show the level each system reaches, not a head-to-head.
Echo 2.0’s signing can look natural and still sign the wrong meaning, so products using it should show that the signing is machine-generated. Avatars are only created from photos of people who’ve given consent, and Echo 2.0 isn’t a replacement for a qualified interpreter in legal, medical or emergency settings. We’re preparing a blind study where fluent Deaf ASL users rate how understandable, natural and grammatical the signing is, and we’ll publish the results, including where Echo falls short.
Use Echo whenever English needs to be seen in ASL: conversations in Talksign apps, and signed versions of websites, videos, courses and announcements. For legal, medical or emergency settings, use a qualified interpreter.
English is put into ASL word order, typically in under 1.5 seconds, and the signs are joined into one continuous video.
Arthur or Amara today, with Elena and Daniel coming soon, on any background colour.
American Sign Language (ASL) today, from a vocabulary of 2,475 signs. Other sign languages are in development with Deaf communities.