Models

Browse and download the latest ai-coustics inference artifacts. Pick the model that matches your version, language and use case.

Last updated 2026-06-10 14:22 UTC

Find compatible models

Language

Version

Show guide

Quail Voice Focus

Quail Voice Focus is optimized for near-field, single-user voice interactions, prioritizing speech close to the microphone while suppressing background noise and distant speech to improve downstream Speech-to-Text (STT) performance. It features an intelligent warm-up period that listens for a moment to confidently identify the primary speaker before full suppression kicks in, prioritizing accuracy over speed.

Model	Version	Download	SHA-256
quail-vf-l-16khz	v1	quail_vf_l_16khz_jc5pk1aa_v17.aicmodel	`1509e36bd30ca296204515eb777dd3663c46a6d4531136599d9e8494e6e4fc27`
quail-vf-1.1-l-16khz	v1	quail_vf_1_1_l_16khz_d00ghjzn_v15.aicmodel	`e0337ec3388fd7be4cd0ab3758b62dd7afbd7817a15b406b17d0250b18d6340d`
quail-vf-2.0-l-16khz	v2	quail_vf_2_0_l_16khz_d42jls1e_v18.aicmodel	`c33a73442e2598acfd2fdc88ca127d1e8ecea0941dc93e4d3e1169246941de6e`
quail-vf-2.1-l-16khz	v3	quail_vf_2_1_l_16khz_8xope536_v11.aicmodel	`04c1ef610feccf2e6f24af608c6cf3f73490ffd34c31a8000a898b24d70998a8`
quail-vf-2.1-l-16khz	v4	quail_vf_2_1_l_16khz_8xope536_v11.aicmodel	`fa1908c2a9557d019b16044cfcfb2cc63bea82811c598a2ebbc30411dff77fdd`
quail-vf-2.1-l-16khz	v5	quail_vf_2_1_l_16khz_8xope536_v11.aicmodel	`e3f6cd3fda62767b601faf50d12dfb75cf1782960ec5a4df5cff32a00f331944`
quail-vf-2.1-s-16khz	v3	quail_vf_2_1_s_16khz_5i8jb8of_v12.aicmodel	`4f46d47179a4505e0220d1c7e2ccc3041a59724dff0afdb9218507274b4b6616`
quail-vf-2.1-s-16khz	v4	quail_vf_2_1_s_16khz_5i8jb8of_v12.aicmodel	`cf251ecad62b043a9fbfe274fe25b8876c5ac1921bc9fb8a8c73c8cb9deada1d`
quail-vf-2.1-s-16khz	v5	quail_vf_2_1_s_16khz_5i8jb8of_v12.aicmodel	`7d0cc51114b77510601adfa0272ebd7918bf02fa36e7e32bdbf11c83b7af43d5`

Quail VAD

Standalone voice activity detection. Robust to noise, runs without a separate enhancement model.

Model	Version	Download	SHA-256
quail-vad-2.0-xxs-16khz	v4	quail_vad_2_0_xxs_16khz_g6609la0_v21.aicmodel	`0caaa4b29864cf6b329e6fc92b3d626c93b79aff52dd4d45aa28d4862fecc2be`
quail-vad-2.0-xxs-16khz	v5	quail_vad_2_0_xxs_16khz_g6609la0_v21.aicmodel	`852005789c9b982f0786c896e6ea1f9ac72c8640f4ea456dceb58e5f54738929`

Quail

Quail is purpose-built for far-field and multi-speaker environments, tuning its processing specifically to enhance downstream Speech-to-Text (STT) engine performance. It does not suppress distant-sounding speech, making it better suited for speakerphone setups, meeting rooms, or situations with multiple participants spread across a space.

Model	Version	Download	SHA-256
quail-l-16khz	v1	quail_l_16khz_dtv5nvgu_v20.aicmodel	`b3db85b6091949cf84d9d6cde1f150cbd74746252a0ae61241d739536b5ea362`
quail-l-16khz	v2	quail_l_16khz_dtv5nvgu_v20.aicmodel	`3327953548d078d575394181f7df265fe5d32383a55e4fcdb8a08f2688c93549`
quail-l-16khz	v3	quail_l_16khz_dtv5nvgu_v20.aicmodel	`8839706f3462adcef003181e70919cc5ca41fa62096634d005ba1fbc89f2f0f4`
quail-l-16khz	v4	quail_l_16khz_dtv5nvgu_v20.aicmodel	`6572795dd68d41940b3bc78df4ccee243468e69db390343075d6574d229dc8f9`
quail-l-16khz	v5	quail_l_16khz_dtv5nvgu_v20.aicmodel	`6f1aeeb545cef59055be1cb5f825889e84c8a16572c17c869acebd87fe1b1a3f`
quail-l-8khz	v1	quail_l_8khz_mp51agn0_v13.aicmodel	`9b5e3d094c32f417d7477b4acaa4ef26510dabbb20d6009a1795940dfee28a74`
quail-l-8khz	v2	quail_l_8khz_mp51agn0_v13.aicmodel	`fe26e46f03451791f592ff69972a7b845a9092ebc305062ed46a713571743738`
quail-l-8khz	v3	quail_l_8khz_mp51agn0_v13.aicmodel	`689eeea2ddd54c1081e6c6bb5287060b631ea4d704fc285ef40af48d296cc445`
quail-l-8khz	v4	quail_l_8khz_mp51agn0_v13.aicmodel	`cdb0602d02e8c329fddd57eacc89fc2d823b33c1c2d477295f5d2afb460be85e`
quail-l-8khz	v5	quail_l_8khz_mp51agn0_v13.aicmodel	`6e7501e989c91e35d906112ffa9456d486515396c712675e3d79919594916bc5`
quail-s-16khz	v1	quail_s_16khz_ilmexgyt_v8.aicmodel	`d613cdd917a47296accb47c45bcc268994f032e1df636c504201099d61c51bf7`
quail-s-16khz	v2	quail_s_16khz_ilmexgyt_v8.aicmodel	`027242aa1a5174a1eec7a77fdd60c0895c934bd24020cb57cd76117ab5c0f197`
quail-s-16khz	v3	quail_s_16khz_ilmexgyt_v8.aicmodel	`bbdcb6b0f43957ddc7b1ef150e31b906fd03e5732f41082bebfe81c52f690022`
quail-s-16khz	v4	quail_s_16khz_ilmexgyt_v8.aicmodel	`e6a6a873dadaa5e4d5df6135fd4e94be52b1c0d12a27fa24841a6f32addf651d`
quail-s-16khz	v5	quail_s_16khz_ilmexgyt_v8.aicmodel	`2f7da1c7d1e1e9b53220908b84d2fb25b0c34d1a3c8538db70080f4967a48c17`
quail-s-8khz	v1	quail_s_8khz_9q57ojki_v20.aicmodel	`093103c9f17261d29d45de91e3506f412eb66c47ed5777eecdd8378a557d6746`
quail-s-8khz	v2	quail_s_8khz_9q57ojki_v20.aicmodel	`80581fe9d63f1170357db24c2edd214e03701450d57b3befa5d92e59acdef2fb`
quail-s-8khz	v3	quail_s_8khz_9q57ojki_v20.aicmodel	`6e1f83c0ad0c341f253d77e5dda08088379e0937c349d9c8ccfbbf880a2e496e`
quail-s-8khz	v4	quail_s_8khz_9q57ojki_v20.aicmodel	`de55b61a97e4febb7b9e5da7f472727be5ce98077962035a4a3c1fa745f95133`
quail-s-8khz	v5	quail_s_8khz_9q57ojki_v20.aicmodel	`e6096081912241879b5e97a5755b598faa0bbad1e18cfd7acb7507cf66515ca0`

Rook

Speech enhancement for human-to-human communication. Reduces noise and reverberation while keeping speech natural.

Model	Version	Download	SHA-256
rook-l-16khz	v2	rook_l_16khz_dtv5nvgu_v20.aicmodel	`e72c5a7d6cc6f396daeff83ffb25be41318cc87a962e22a607963bebf3cfdcfc`
rook-l-16khz	v1	rook_l_16khz_dtv5nvgu_v20.aicmodel	`ec8550a4f41f58beac13d2a576fdb36f62a89410192ce0f52ed43b8289b7970c`
rook-l-16khz	v3	rook_l_16khz_dtv5nvgu_v20.aicmodel	`b90223799cac9f92bc92a1bb87c27b3b072a5c34897680b9953180fdca0610df`
rook-l-16khz	v4	rook_l_16khz_dtv5nvgu_v20.aicmodel	`2e25022720e41c15f3ec39b46a10dc117b4bf7803320fbc26bad1d4e8790c215`
rook-l-16khz	v5	rook_l_16khz_dtv5nvgu_v20.aicmodel	`b4df0e31be9e0bf1e8e29172be2ca1bbf7429abc42f8ce56cf61c1797fd0df64`
rook-l-48khz	v2	rook_l_48khz_7n1r44ri_v23.aicmodel	`ff8838cef6c6fa13681f2354f6f6d637f330a107c845854c09a4f2ed9837b7ac`
rook-l-48khz	v1	rook_l_48khz_7n1r44ri_v23.aicmodel	`04a63bf298821874ebef7ac558e66be7964c21baf651fb9638d557766353bbcb`
rook-l-48khz	v3	rook_l_48khz_7n1r44ri_v23.aicmodel	`51a41edd010809035deddaf3914abc6382fd2bb2592994ca8a514cc58de5deab`
rook-l-48khz	v4	rook_l_48khz_7n1r44ri_v23.aicmodel	`3ca83524c1499fc98399df5471dc209a3650cbf2977a34097683f7e10785d84d`
rook-l-48khz	v5	rook_l_48khz_7n1r44ri_v23.aicmodel	`cb261875281ec668635c6ca7779a13dffde113c8f1033cdfce9155b0b55959dd`
rook-l-8khz	v2	rook_l_8khz_mp51agn0_v13.aicmodel	`fe2b1c02e60b886d571ce40c05e3d215184450c6c7220acb2c3b4ded1b3f5e95`
rook-l-8khz	v1	rook_l_8khz_mp51agn0_v13.aicmodel	`e916757e356dd299bde6842a31c2e50ea11d81bdca0434a09c9d6e25f2ab5038`
rook-l-8khz	v3	rook_l_8khz_mp51agn0_v13.aicmodel	`9a335d6ad7f456256e244d2594101b26f06c4cafd9a62f3d85e211782785515c`
rook-l-8khz	v4	rook_l_8khz_mp51agn0_v13.aicmodel	`111fecf8eaa5dea2aac2d41cd481b38edbe4df87c96633f915ee74a726396516`
rook-l-8khz	v5	rook_l_8khz_mp51agn0_v13.aicmodel	`7cb2db7178c683fb720e9579f2bc6c88d683c8d864b8b7368d1e4c88888d60bb`
rook-s-16khz	v2	rook_s_16khz_ilmexgyt_v8.aicmodel	`3eff7b5c6874f06a3ced70844fe31705c35c646ff198aa3590ece6007eb7bc25`
rook-s-16khz	v1	rook_s_16khz_ilmexgyt_v8.aicmodel	`43ef287f103839330e71f508f130339c5753b50ac1254436368168ab759faf2d`
rook-s-16khz	v3	rook_s_16khz_ilmexgyt_v8.aicmodel	`1952e587be74c2a409ab1795af1bb9e892f46ffcb254525a8109debae31f989e`
rook-s-16khz	v4	rook_s_16khz_ilmexgyt_v8.aicmodel	`cdcb5e5b264301a341e6cdcf836112e8a02919106653e0ecd2da350e51c5e323`
rook-s-16khz	v5	rook_s_16khz_ilmexgyt_v8.aicmodel	`b20438883d3c21e97b73010da650f7faab104b7807773fa19b12eb29ee999cc4`
rook-s-48khz	v2	rook_s_48khz_kg8kp9qm_v52.aicmodel	`0328e6f8f45a07c6b5a957a571a0565370622b9bf98c18b7ade45cbeb7cc24a1`
rook-s-48khz	v1	rook_s_48khz_kg8kp9qm_v52.aicmodel	`291ca14c21b66e1acc3b6ca352e9452c5a04da2730f433ab9d49edf2528229ac`
rook-s-48khz	v3	rook_s_48khz_kg8kp9qm_v52.aicmodel	`35529310b22b132d3a30443a0d66a280cec407f3bb406a79dcb1d2dab7f975bb`
rook-s-48khz	v4	rook_s_48khz_kg8kp9qm_v52.aicmodel	`bb60832a0954b09a7ddc9324b813005adf2794cc839684894bd74153cc48615c`
rook-s-48khz	v5	rook_s_48khz_kg8kp9qm_v52.aicmodel	`06ad93e8c6cfbb96819ec6b1de65e308d543f117d19f11b8a2e41c6f86e7168e`
rook-s-8khz	v2	rook_s_8khz_9q57ojki_v20.aicmodel	`ccb1a220f1d1bc26426e03848e6bcb79183ab9f1022e9ae07e9ab531347809e2`
rook-s-8khz	v1	rook_s_8khz_9q57ojki_v20.aicmodel	`a97d221f192730532109125effdbb22da64cb07a66c473cf4e2f9a9559d94c68`
rook-s-8khz	v3	rook_s_8khz_9q57ojki_v20.aicmodel	`a8212ae3b9addffa06b56629c31dda4ad30fd9a30fbf7dece1f41319f3736d14`
rook-s-8khz	v4	rook_s_8khz_9q57ojki_v20.aicmodel	`9cca1885bb2d8db024ef4f77fdf3bea6032f313a5f1eea380c49657ca9af3421`
rook-s-8khz	v5	rook_s_8khz_9q57ojki_v20.aicmodel	`abf4f50acb00a1c05d2a3be7848fb0152e8841379a16e77bbdd126d286ad4b5a`
rook-xs-48khz	v1	rook_xs_48khz_g4qwc6lv_v21.aicmodel	`63bffca7273040fc82b28a53eedfb69fdd4d9beacab5c3c13fa4d8a44fd65a1b`
rook-xxs-48khz	v1	rook_xxs_48khz_wsur2zkw_v7.aicmodel	`914b8deb45a49aa32e73828c2cc54472fda211df276a206d6691e7c428eb36aa`

Tyto

Tyto is a lightweight model that runs on your audio stream and predicts whether the audio reaching your agent will cause downstream failures. It outputs a single score plus a breakdown across seven dimensions: noise, reverb, dropouts, codec, and more.

Model	Version	Download	SHA-256
tyto-l-16khz	v5	tyto_l_16khz_yhlek4hc_v43.aicmodel	`8fb7cb977fb6f81b75fb27398eb1e4a5c619f7f95c252e17df75ef7c27faf336`

Other

Additional models that don't fit the categories above.

Model	Version	Download	SHA-256
bypass	v1	bypass.aicmodel	`883fbd2d86e951cc4c4fac4af547c41dc61da385beed01084021c15f2c9f0c1e`
bypass	v2	bypass.aicmodel	`5f6808ec14d53534d78738d8e05bd2fa27953eb58b960240971307a19dfc9ecf`
bypass	v3	bypass.aicmodel	`bcaffed11b4eaa6bf9ad656089b1192628c68df2ad56880d3aa8b1982e9f8d4a`
bypass	v4	bypass.aicmodel	`f1041d8173dcbd7715bc668ae6baab6c43598bb6b95d63ddbbd6c19f2bd5f89b`
bypass	v5	bypass.aicmodel	`3239206fdbac28e89bae144a4f2c5a87528e6f4f912d470bb011c94f4b650e7a`

Learn more

Documentation for picking the right model and getting the most out of it.

Model guide

Detailed information on model capabilities, supported sample rates, latency characteristics, and how to choose the right model for your use case.

Read the model guide

Speech enhancement for ASR

How to use our models to improve downstream Speech-to-Text (STT) accuracy, including integration patterns and tuning tips.

Improve STT performance