Contribute | SceneMakerAI

pyannote-audio: Fixed an issue where the pipeline failed with a KeyError because the default URI was missing from the input dict containing file objects

📄️Fixed an issue where uppercase fields in nested dataclasses were omitted in case-insensitive mode

Fixed an issue where capitalized fields in nested pydantic dataclasses were not populated by env and dotenv when `pydantic-settings: case_sensitive=False` was set.

📄️Issue where misaligned segments are missing from SRT/VTT output

whisperX: Reported an issue where segments that `align()` failed to align remain in TXT and TSV files but are omitted from SRT and VTT files without any warning

📄️Fixed an issue where `SSDictCursor` returned a tuple instead of a dict

asyncmy: Since it's a Cython `cdef` method, overrides don't work. Fixed the issue where `SSDictCursor` was returning a tuple by changing it to `cpdef`.

📄️Fixed an issue where `collect_env` would either read packages from a different environment or crash in a UV environment

https://github.com/vllm-project/vllm/pull/60191

📄️Issue where `save_filename` is ignored in `COPY_SKIP_CACHE`, causing the label encoder not to load

speechbrain: Reported an issue where the VoxLingua107 label encoder silently fails because the COPY_SKIP_CACHE strategy returns paths without `fetch()`

📄️Audio.crop pad mode returns incorrect lengths and actual audio from sections outside the specified range

pyannote-audio: Report of an issue where the "pad" mode in Audio.crop returns incorrect results for segments that fall completely outside the file range

📄️Fixed an issue where the inference progress hook reported a higher number of completions than the total

pyannote-audio: Fixed an issue where the number of completions passed to the progress hook was greater than the total when the last batch was small

📄️Fixed a missing escape character in the VARCHAR primary key when querying or deleting by ID

pymilvus: Fixed an issue where MilvusClient's get/query/delete(ids=) method would cause errors or delete the wrong rows because it did not escape VARCHAR keys containing quotes or backslashes.

📄️The issue where `transcribe(chunk_length)` permanently changes the window size for subsequent calls

faster-whisper: Reported a state leak where a call with a specified `chunk_length` changes the window size for subsequent calls to the same model

📄️Fixed an issue where `detect_language` crashed when `vad_parameters` was in dictionary format

Fixed an issue where `faster-whisper: detect_language` would fail with an `AttributeError` when receiving `vad_parameters` in the form of a dictionary as described in the documentation.

📄️A ValueError occurs during SSLContext creation if only `ssl_verify_identity` is enabled

PyMySQL: I've seen reports of an issue where, if you set only `ssl_verify_identity=True` without `ssl_verify_cert`, `CERT_NONE` and `check_hostname` conflict, causing a `ValueError` before the connection is established.

📄️Add an "Allow" header to StaticFiles 405 responses

Starlette: Fixed an issue where the 405 response from StaticFiles was missing the "Allow" header required by RFC 9110

📄️Fixed an issue where `dotenv run` rejected .env files in FIFO format

python-dotenv: Fixed an issue where `dotenv run` would reject a .env file in named-pipe format with a "does not exist" error; now handles it the same way as the library's `list/get` methods.

📄️If I set UPLOAD_DIR to a string and enable UPLOAD_KEEP_FILENAME, I get a TypeError. Fix

python-multipart: A TypeError occurs when UPLOADDIR is set to a string and UPLOADKEEP_FILENAME is enabled

📄️--Issue where all h11 and wsproto responses fail due to a space after the colon in the header value

uvicorn: Reported an issue where adding a space after the colon (e.g., --header "Name: Value") causes all responses and handshakes to fail in h11·wsproto (Potential Issue discussion)

📄️Issue where Korean and English words become joined due to missing spaces when merging long audio chunks

Qwen3-ASR: Missing Spaces When Merging Long Audio Chunks (Korean and English Words Stuck Together)

📄️Issue where sorting query/search results in place causes vectors and dynamic fields to end up in different rows

pymilvus: Reported an issue where, when sorting, reversing, or removing elements from query/search results in place, delayed-decoding fields end up attached to different rows.

📄️Fixed an issue where the `sf.info() duration` value was displayed incorrectly in exponential notation for files longer than 10 minutes

python-soundfile: Issue where the duration returned by sf.info() in files longer than 10 minutes is displayed in exponential notation

📄️Fixed an issue where speech was missing after a forced split when max_speech_duration_s was set

silero-vad: Fixed an issue where speech retained by `neg_threshold` was discarded when forced to split at `max_speech_duration_s`.

📄️Fixed an issue where `COPY_SKIP_CACHE` returned paths without `fetch()`

speechbrain: Pull request to fix Issue #3091. Handles the name change that previously depended on `force_filename` in `COPY_SKIP_CACHE` directly, ensuring the label encoder loads correctly.

📄️Fixed an issue where including "16k" in the path caused the default ONNX model to be mistakenly identified as 16 kHz-only

silero-vad: Fixed an issue where OnnxWrapper would search the entire path for "16k," and if the parent directory name contained "16k," the default model would reject 8 kHz input.

📄️Issue where the stt_svc process crashes due to a Rust panic while extracting text from web search results

agent-stt: Record of the issue where a Rust html2text panic inside ddgs.extract() causes the process to crash, and the corrective actions taken

📄️Fixed an issue where HTTPSRedirectMiddleware removed the square brackets from IPv6 hostnames, resulting in an incorrect Location header

starlette: Fixed an issue where the square brackets were missing from the redirect URL when the host was an IPv6 address followed by the default port, such as [::1]:80 (Discussion #3637)

📄️Modified to encode audio paths and bytes even without torchcodec (audiofolder, from_dict)

huggingface/datasets: Modified to allow encoding of audio paths and bytes without torchcodec

📄️Fixed an issue where unsorted segments were left in SRT/VTT

whisperX: Fix PR for Issue #1490. Since `align()` fails to align, `words` outputs empty segments to the queue even in SRT and VTT.

📄️Issue where the JSON Schema for [TypedDict] kwargs in `Unpack` is generated differently from how validation works

pydantic: Reported an issue where the JSON Schema generated for the function's **kwargs: Unpack[TypedDict] is inconsistent with the actual validation in both directions (fix prepared)

📄️The issue of punctuation within words splitting them in `word_timestamps`

openai/whisper: When using `word_timestamps`, an issue was raised (in a Q&A discussion) where punctuation within a word—such as the 3.5% or the particle following a quotation mark—splits the word into two parts, with the punctuation attached to the front of the second part.

📄️Fixed an issue where TextEncoder ignored the `n_most_common` and `min_count` arguments

speechbrain: Fixed an issue where `TextEncoder.limited_labelset_from_iterable` passed a fixed value instead of an argument, causing the label set to remain unrestricted.

📄️If you use `del` on a non-existent attribute in `State`, a `KeyError` is raised instead of an `AttributeError`.

starlette: Raises a KeyError directly for attributes that do not have `State.__delattr__`, raising concerns that this is a different type of exception than `getattr` (Potential Issue discussion)

📄️Fixed an issue where video was missing in the fallback because `hosted_vllm` modified the caller's video file part

litellm: Issue where hostedvllm in-place replaces the caller's video file part with videourl

0.7 regression where ContentDetector discards cuts that fall short of `min_scene_len` at 29.97/59.94 fps

PySceneDetect: In a 29.97/59.94 fps video, ContentDetector discards cuts that are separated by a number of frames equal to minscenelen.

📄️Fixed an issue where ImageMagick-format EXIF data was missing from derived images after calling `exif_transpose()`

Pillow: Issue where ImageMagick-formatted EXIF data disappears entirely after calling exif_transpose()

📄️Issue where videos without frame numbers (WebM/MKV streams, raw H.264) are loaded with 1 frame or negative metadata

vLLM: Issue where videos without recorded frame counts (streaming WebM/MKV, raw H.264) are truncated to 1 frame or fail in the sampler

📄️An issue where re-exporting a value declared in a previous ESM block causes parsing to fail and the error to point to the wrong line

micromark-extension-mdxjs-esm Issue Draft: If a value declared in a previous ESM block is re-exported in the next block, an error occurs in an unexpected location

📄️Issue where token ID prompts are rejected as "Unmapped prompt format" in hosted_vllm text completion

litellm: Issue with token ID prompts being rejected in hosted_vllm text completion (Issue)

📄️Fixed a regression where Exif.tobytes() and saving failed with a struct.error after reading IFD1 (thumbnail IFD)

Pillow: Fixed an issue where Exif.tobytes() failed with a struct.error after reading an IFD1 file

📄️Fixed an issue where run-batch did not import the tool/reasoning parser plugin

vLLM: Parser plugin not loaded in run-batch

📄️Fixed an issue where setting `top_logprobs=-1` in `/v1/responses` returned an empty list

vLLM: Issue where toplogprobs returns an empty result when toplogprobs=-1 is requested in the Responses API

📄️Fixed an issue where voice recognition and translation requests bypassed the server's default presence/frequency penalty

vLLM: The Issue Where Speech Recognition and Translation Requests Bypass the Server’s Default Presence/Frequency Penalty

📄️Fixed an issue where `get_all_files` excluded all files following those matched by `exclude_and`

speechbrain: Fixed an issue where getallfiles removed all files following those matched by exclude_and