stream : fix build

stream : add sliding window mode

whisper : fix UB when reading buffer of length 0 bytes (#265)

ggml : fix indentation

ggml : make compatible with c99 (#262)

talk : improve prompting

release : v1.0.3

Update README.md

release : v1.0.2

Update README.md

Update README.md

Node.js package (#260)

talk : make compatible with c++11 (part 2)

talk : make compatible with c++11

cmake : require c++11 instead of c++20

Remove C++20 requirement (#257)

Add newline per segment for text output (#254)

bench : more concise representation of the results (#89)

minor : fix .gitignore to not ignore examples

bench.wasm : same as "bench" but runs in the browser (#89)

Update README.md

talk : fix build for MSVC

talk : talk with AI in the terminal

fix potential bug reading model data into a small size optimized string which could lead to memory corruption. In an SSO string, you can't write data to &str[0] and expect it to work well.

whisper : minor improvemnt in decoding strategy (#244)

ggml : add alternative cblas_sgemm call

make : indentation + .gitignore

Flag for Position Independent Code

twitch.sh : various fixes and polishing

Allow for Twitch.tv live transcription

Fix paths echoed after the download

fix compilation on haiku

yt-wsp.sh : improve usage instructions

yt-wsp.sh : fix usage instruction + comment

Update README.md

livestream.sh : remove obsolete comment

ggml : fix typo in previous commit

ggml : use macros to inline FP16 <-> FP32 conversions

ggml : add F16C CPU flag check

add fp16/fp32 convert intrinsics

models : add the new "large" model release by OpenAI

bench : add commit hash to bench-all.sh results

Try to improve the token sampling strategy (#193)

tests : adding transcription tests

ggml : remove inline specifier from fp16 <-> fp32 converters

livestream : handle ffmpeg errors gracefully and stabilize transcript

livestream : minor changes

livestream : fix losing words across audio chunk (#195)

Fix Darwin flags - was incorrectly always using the Linux else clause

whisper : add mechanism for aborting the whisper_full() computation