Sesame CSM-1B
Free.ai (self-hosted)
·
tts
·
~500 token kasta clip
·
4.3
ka 3 user of this category
Sesame CSM-1B waa a Codka qoraalka-ugu-hadalka dhisay Sesame. Ugu xoog badan ee Real-time voice agents, conversational AI, streaming TTS where time-to-first-audio matters more than studio-perfect reads.. Is-hoosaysiinta Free.ai GPUs - bilaash ayaa ka socda ishaada maalinlaha ah (500 tokens oo ah fiidiyow kasta). Soo baxay hoos Apache 2.0 — isticmaalka ganacsi ee la oggol yahay on Free.ai.
isticmaalka API
curl -X POST https://api.free.ai/v1/tts/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"csm-1b","text":"hello world"}'Su'aalaha badanaa la isweydiiyo
Sesame CSM-1B waxay taageertaa noocyo kala duwan oo afaf ah. Liiska saxda ah wuxuu ku xiran yahay mashiinka; foomka boggani wuxuu aqbalaa qoraal kasta oo mashiinka wuxuu ku soo bandhigi doonaa luqadaha lagu taageerayo. Eeg /voice/ si aad u hesho mashiinka multi-engine ee buuxa haddii aad u baahan tahay luqad gaar ah.
Injinnada badankoodu waxay ku sameeyaan Ingiriisiga dhexdhexaadka ah ee Mareykanka oo aan la dooran iyo afka gobolka ee ku habboon luqadaha aan Ingiriisiga ahayn. Injiinnada Premium waxay muujin karaan noocyo kala duwan oo afka ah - ku dheji tusaale si aad u barbar dhigto.
SSML taageero kala duwan yahay by engine. Pause, prosody, iyo xusuusin tags waxaa lagu maamuusay on engines premium badan iyo qaar ka mid ah kuwa self-hosted.
Streaming TTS waxaa laga heli karaa on premium engines via the /v1/tts/ API endpoint la stream = run. Web UI on page this soo celin doonaa clip buuxa marka rendering dhamaystiran.
Sesame CSM-1B ku socda GPUs our gaar ah. Generation ka soo qaadaa maalin kasta free pool hore. Marka la soo afjaro, lacagta lagu bixiyo tokens bilaabaan $ 5 → 200,000 tokens. Ku dhawaad ~ 5 tokens per character, ugu yaraan 100 per clip.
Inta u dhaxaysa 5,000 xaraf oo codsi kasta ah ee web UI. Bilood dheer (audiobooks, qaybaha buuxa), isticmaal / codka / audiobook / oo chunks iyo stitches otomaatig ah, ama u yeedha API in loop.
Haa — POST liiska strings in /v1/tts/batch /, ama isticmaalka workspace UI at /workspace / in ay silsilad TTS in pipeline dheer (tusaale ahaan, beddel → hadal → sudhan).
Haa — POST qoraalka /v1/tts / la model="Sesame CSM-1B" (ama slug on page this). Ku soo laabtay WAV ama MP3. Eeg /api/ u soo jeedinta buuxda + SDK snippets.
Boggan waa qoraal-u-hadalka, ma aha codka isku-dhafan — codka waa nidaamka ugu horeeya ee mashiinka. Codka isku-dhafan (u soo dejinta codka soo jeedinta), eeg /voice/clone/, taas oo kuu baahan inaad ama leedahay xuquuqda codka ama aad leedahay oggolaansho qoraal ah.
Injiilooyinka is-hoosaysiinta ah waxay ku socdaan GPUs Free.ai-oo leh; waxba kama tagaan server-yadayaga. Injiilooyinka Premium waxay ku gudbiyaan qoraalka si ay u bixiyaan adeegyada qaabka hoos yimaada DPA-yadayaga.
Haa — Free.ai waxay siisaa isticmaal ganacsi oo ah codka la abuuray. Liisankaaga mashiinka (Apache 2.0, MIT, ama shuruudaha iibiyaha) waxaa lagu muujiyaa kor iyo bogga tusaalaha; dhaqan ahaan, tani waxay ka dhigan tahay in codka, xayeysiiska, podcasts, iyo barnaamijyada dhammaantood ay ku jiraan.
Haa — shaqo ku fashilantay auto-soo celinta in ay asalka (dhaqameed maalin kasta ama lacagta la siiyay). Haddii lacag la siiyo ma muujiyaan maalintii, email contact@free.ai.