Improve render.py robustness/performance, add editing helpers#63
Open
eaikarensantos-gif wants to merge 4 commits into
Open
Improve render.py robustness/performance, add editing helpers#63eaikarensantos-gif wants to merge 4 commits into
eaikarensantos-gif wants to merge 4 commits into
Conversation
- Surface ffmpeg stderr on failure instead of swallowing it - Match output fps to the source instead of forcing 24fps - Probe each source once (HDR + orientation + fps) with a cache - Encode segments in parallel (up to 4 ffmpeg processes) - Detect libass support by probing -filters instead of hardcoding paths - Refuse lossless concat when segment resolutions mismatch - Escape subtitle paths safely for the filtergraph (brackets, quotes) - Use -map 0:a? so audioless sources don't fail the composite Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Reuse one Anthropic client instead of per-request instantiation - Cache TechCrunch headlines for 5 minutes - Update model to claude-opus-4-8 - Disable Flask debug mode by default (opt in via FLASK_DEBUG=1) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
18 issues found across 12 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name=".gitignore">
<violation number="1" location=".gitignore:75">
P2: Root-only .gitignore patterns are not anchored with a leading '/', causing them to match directories at any depth despite the explicit "repo root" comment.</violation>
</file>
<file name="helpers/export_formats.py">
<violation number="1" location="helpers/export_formats.py:48">
P2: `get_video_info` does not handle ffprobe failures or malformed output, leading to unhandled `JSONDecodeError`, `KeyError`, or `StopIteration` on invalid/non-video inputs.</violation>
<violation number="2" location="helpers/export_formats.py:70">
P2: ffmpeg global options `-y` and `-loglevel` are placed after the output path, so they may not take effect and non-interactive runs can hang on overwrite prompts</violation>
</file>
<file name="helpers/burn_subs_pil.py">
<violation number="1" location="helpers/burn_subs_pil.py:181">
P1: Silent fallback from yuva420p to yuv420p removes alpha channel, breaking transparent subtitle overlay</violation>
</file>
<file name="helpers/correct_transcript.py">
<violation number="1" location="helpers/correct_transcript.py:94">
P1: Invalid `ensure_ascii=False` passed to built-in `open()`, causing a runtime TypeError before the corrected JSON is written.</violation>
</file>
<file name="helpers/batch.py">
<violation number="1" location="helpers/batch.py:89">
P2: Using only `Path(video_path).stem` for generated artifact paths causes collisions when inputs share the same base name with different extensions (e.g., `clip.mp4` and `clip.mov`). The second input reuses or overwrites the first's transcript, EDL, and final output, silently corrupting results.</violation>
</file>
<file name="helpers/thumbnail.py">
<violation number="1" location="helpers/thumbnail.py:35">
P2: Uncaught `ValueError` when ffprobe returns non-numeric duration (e.g., `N/A`) crashes the script instead of exiting gracefully.</violation>
<violation number="2" location="helpers/thumbnail.py:51">
P1: Missing validation for non-positive `--interval` causes infinite loop in frame extraction</violation>
<violation number="3" location="helpers/thumbnail.py:193">
P1: `saved.append` runs unconditionally even when `cv2.imwrite` was skipped, causing `make_contact_sheet` to crash on `FileNotFoundError` when it tries to open missing thumbnail files.</violation>
</file>
<file name="helpers/detect_breaths.py">
<violation number="1" location="helpers/detect_breaths.py:47">
P2: Audio extraction reads WAV data with a hardcoded 44-byte header offset, which can mis-parse valid WAV files that contain extra chunks (e.g., LIST, fact, metadata) before the data chunk. Non-audio bytes would then be interpreted as samples, corrupting RMS calculations and breath detection results.</violation>
<violation number="2" location="helpers/detect_breaths.py:57">
P1: Missing edge-case guards can crash on invalid `--window-ms` or very short audio: `window` may be 0 causing `ZeroDivisionError`, and empty RMS causes `np.percentile` to crash on empty array.</violation>
<violation number="3" location="helpers/detect_breaths.py:77">
P2: Minimum breath duration filtering uses `int()` (floor) for `min_windows`, allowing regions shorter than the user-configured `--min-ms`.</violation>
</file>
<file name="helpers/denoise.py">
<violation number="1" location="helpers/denoise.py:57">
P2: curl model download lacks `--fail`, allowing HTTP error pages to be silently saved as the model file, breaking `arnndn` denoise mode persistently.</violation>
</file>
<file name="news_server.py">
<violation number="1" location="news_server.py:102">
P2: request.json can raise BadRequest/UnsupportedMediaType before the route's try/except, causing the API to return framework error responses instead of its own JSON error contract.</violation>
</file>
<file name="helpers/reframe.py">
<violation number="1" location="helpers/reframe.py:128">
P2: Missing guard for empty sampled positions can crash `reframe.py` with `ValueError` when processing a corrupt or zero-frame video that opens but yields no frames.</violation>
<violation number="2" location="helpers/reframe.py:187">
P2: ffmpeg global options `-y` and `-loglevel` placed after output file may be ignored</violation>
</file>
<file name="helpers/check_gaps.py">
<violation number="1" location="helpers/check_gaps.py:27">
P2: Using `Path.glob()` with raw user/data-derived `source_name` can produce incorrect matches if the source name contains glob metacharacters (`*`, `?`, `[]`). Furthermore, selecting `candidates[0]` without sorting makes the file choice nondeterministic when multiple files match.</violation>
<violation number="2" location="helpers/check_gaps.py:28">
P2: Missing transcript files are silently ignored, leading to false "no gaps found" reports when segments were never actually checked.</violation>
</file>
Tip: instead of fixing issues one by one fix them all with cubic
Re-trigger cubic
| result = subprocess.run(cmd, capture_output=True) | ||
| if result.returncode != 0: | ||
| # Fallback: yuv420p se yuva420p não suportado | ||
| cmd[cmd.index("yuva420p")] = "yuv420p" |
There was a problem hiding this comment.
P1: Silent fallback from yuva420p to yuv420p removes alpha channel, breaking transparent subtitle overlay
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/burn_subs_pil.py, line 181:
<comment>Silent fallback from yuva420p to yuv420p removes alpha channel, breaking transparent subtitle overlay</comment>
<file context>
@@ -0,0 +1,195 @@
+ result = subprocess.run(cmd, capture_output=True)
+ if result.returncode != 0:
+ # Fallback: yuv420p se yuva420p não suportado
+ cmd[cmd.index("yuva420p")] = "yuv420p"
+ result2 = subprocess.run(cmd, capture_output=True)
+ if result2.returncode != 0:
</file context>
| data["words"] = correct_words(data["words"], api_key=args.api_key) | ||
|
|
||
| out_path = args.output or str(Path(args.transcript).with_suffix(".corrected.json")) | ||
| with open(out_path, "w", ensure_ascii=False) as f: |
There was a problem hiding this comment.
P1: Invalid ensure_ascii=False passed to built-in open(), causing a runtime TypeError before the corrected JSON is written.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/correct_transcript.py, line 94:
<comment>Invalid `ensure_ascii=False` passed to built-in `open()`, causing a runtime TypeError before the corrected JSON is written.</comment>
<file context>
@@ -0,0 +1,101 @@
+ data["words"] = correct_words(data["words"], api_key=args.api_key)
+
+ out_path = args.output or str(Path(args.transcript).with_suffix(".corrected.json"))
+ with open(out_path, "w", ensure_ascii=False) as f:
+ json.dump(data, f, ensure_ascii=False, indent=2)
+
</file context>
Comment on lines
+193
to
+195
| saved.append((ts, dst, sc, box)) | ||
| print(f" #{rank} t={ts:.1f}s score={sc:.1f} → {dst}") | ||
|
|
There was a problem hiding this comment.
P1: saved.append runs unconditionally even when cv2.imwrite was skipped, causing make_contact_sheet to crash on FileNotFoundError when it tries to open missing thumbnail files.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/thumbnail.py, line 193:
<comment>`saved.append` runs unconditionally even when `cv2.imwrite` was skipped, causing `make_contact_sheet` to crash on `FileNotFoundError` when it tries to open missing thumbnail files.</comment>
<file context>
@@ -0,0 +1,206 @@
+ img = cv2.imread(img_path)
+ if img is not None:
+ cv2.imwrite(dst, img, [cv2.IMWRITE_JPEG_QUALITY, 95])
+ saved.append((ts, dst, sc, box))
+ print(f" #{rank} t={ts:.1f}s score={sc:.1f} → {dst}")
+
</file context>
Suggested change
| saved.append((ts, dst, sc, box)) | |
| print(f" #{rank} t={ts:.1f}s score={sc:.1f} → {dst}") | |
| if img is not None: | |
| cv2.imwrite(dst, img, [cv2.IMWRITE_JPEG_QUALITY, 95]) | |
| saved.append((ts, dst, sc, box)) | |
| print(f" #{rank} t={ts:.1f}s score={sc:.1f} → {dst}") |
| ], check=False) | ||
| if os.path.exists(out_path): | ||
| frames.append((t, out_path)) | ||
| t = round(t + interval_s, 3) |
There was a problem hiding this comment.
P1: Missing validation for non-positive --interval causes infinite loop in frame extraction
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/thumbnail.py, line 51:
<comment>Missing validation for non-positive `--interval` causes infinite loop in frame extraction</comment>
<file context>
@@ -0,0 +1,206 @@
+ ], check=False)
+ if os.path.exists(out_path):
+ frames.append((t, out_path))
+ t = round(t + interval_s, 3)
+ return frames
+
</file context>
|
|
||
| def compute_rms(samples, sr, window_ms=50): | ||
| """Calcula RMS por janelas de window_ms milissegundos.""" | ||
| window = int(sr * window_ms / 1000) |
There was a problem hiding this comment.
P1: Missing edge-case guards can crash on invalid --window-ms or very short audio: window may be 0 causing ZeroDivisionError, and empty RMS causes np.percentile to crash on empty array.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/detect_breaths.py, line 57:
<comment>Missing edge-case guards can crash on invalid `--window-ms` or very short audio: `window` may be 0 causing `ZeroDivisionError`, and empty RMS causes `np.percentile` to crash on empty array.</comment>
<file context>
@@ -0,0 +1,227 @@
+
+def compute_rms(samples, sr, window_ms=50):
+ """Calcula RMS por janelas de window_ms milissegundos."""
+ window = int(sr * window_ms / 1000)
+ n_windows = len(samples) // window
+ rms = np.array([
</file context>
|
|
||
| @app.route("/api/generate", methods=["POST"]) | ||
| def generate(): | ||
| data = request.json or {} |
There was a problem hiding this comment.
P2: request.json can raise BadRequest/UnsupportedMediaType before the route's try/except, causing the API to return framework error responses instead of its own JSON error contract.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At news_server.py, line 102:
<comment>request.json can raise BadRequest/UnsupportedMediaType before the route's try/except, causing the API to return framework error responses instead of its own JSON error contract.</comment>
<file context>
@@ -0,0 +1,133 @@
+
+@app.route("/api/generate", methods=["POST"])
+def generate():
+ data = request.json or {}
+ headline = data.get("headline", "").strip()
+ summary = data.get("summary", "").strip()
</file context>
| crop_h = int(src_w * target_h / target_w) | ||
|
|
||
| # posição X do centro de interesse (mediana suavizada) | ||
| median_cx = float(np.median([cx for _, cx in smooth_pos])) |
There was a problem hiding this comment.
P2: Missing guard for empty sampled positions can crash reframe.py with ValueError when processing a corrupt or zero-frame video that opens but yields no frames.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/reframe.py, line 128:
<comment>Missing guard for empty sampled positions can crash `reframe.py` with `ValueError` when processing a corrupt or zero-frame video that opens but yields no frames.</comment>
<file context>
@@ -0,0 +1,198 @@
+ crop_h = int(src_w * target_h / target_w)
+
+ # posição X do centro de interesse (mediana suavizada)
+ median_cx = float(np.median([cx for _, cx in smooth_pos]))
+ center_x = int(median_cx * src_w)
+
</file context>
Comment on lines
+187
to
+195
| out_path, "-y", "-loglevel", "error" | ||
| ] | ||
| result = subprocess.run(cmd) | ||
| if result.returncode == 0: | ||
| size_mb = os.path.getsize(out_path) / 1024 / 1024 | ||
| print(f"\ndone: {out_path} ({size_mb:.1f} MB)") | ||
| else: | ||
| sys.exit("Erro no reframe.") | ||
|
|
There was a problem hiding this comment.
P2: ffmpeg global options -y and -loglevel placed after output file may be ignored
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/reframe.py, line 187:
<comment>ffmpeg global options `-y` and `-loglevel` placed after output file may be ignored</comment>
<file context>
@@ -0,0 +1,198 @@
+ "-vf", vf,
+ "-c:v", "libx264", "-crf", "20", "-preset", "fast",
+ "-c:a", "copy",
+ out_path, "-y", "-loglevel", "error"
+ ]
+ result = subprocess.run(cmd)
</file context>
Suggested change
| out_path, "-y", "-loglevel", "error" | |
| ] | |
| result = subprocess.run(cmd) | |
| if result.returncode == 0: | |
| size_mb = os.path.getsize(out_path) / 1024 / 1024 | |
| print(f"\ndone: {out_path} ({size_mb:.1f} MB)") | |
| else: | |
| sys.exit("Erro no reframe.") | |
| cmd = [ | |
| "ffmpeg", "-y", "-loglevel", "error", | |
| "-i", video_path, | |
| "-vf", vf, | |
| "-c:v", "libx264", "-crf", "20", "-preset", "fast", | |
| "-c:a", "copy", | |
| out_path, | |
| ] |
| # index transcripts by source name | ||
| transcripts = {} | ||
| for source_name in edl["sources"]: | ||
| candidates = list(Path(transcripts_dir).glob(f"{source_name}.json")) |
There was a problem hiding this comment.
P2: Using Path.glob() with raw user/data-derived source_name can produce incorrect matches if the source name contains glob metacharacters (*, ?, []). Furthermore, selecting candidates[0] without sorting makes the file choice nondeterministic when multiple files match.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/check_gaps.py, line 27:
<comment>Using `Path.glob()` with raw user/data-derived `source_name` can produce incorrect matches if the source name contains glob metacharacters (`*`, `?`, `[]`). Furthermore, selecting `candidates[0]` without sorting makes the file choice nondeterministic when multiple files match.</comment>
<file context>
@@ -0,0 +1,87 @@
+ # index transcripts by source name
+ transcripts = {}
+ for source_name in edl["sources"]:
+ candidates = list(Path(transcripts_dir).glob(f"{source_name}.json"))
+ if candidates:
+ transcripts[source_name] = load_words(str(candidates[0]))
</file context>
| transcripts = {} | ||
| for source_name in edl["sources"]: | ||
| candidates = list(Path(transcripts_dir).glob(f"{source_name}.json")) | ||
| if candidates: |
There was a problem hiding this comment.
P2: Missing transcript files are silently ignored, leading to false "no gaps found" reports when segments were never actually checked.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At helpers/check_gaps.py, line 28:
<comment>Missing transcript files are silently ignored, leading to false "no gaps found" reports when segments were never actually checked.</comment>
<file context>
@@ -0,0 +1,87 @@
+ transcripts = {}
+ for source_name in edl["sources"]:
+ candidates = list(Path(transcripts_dir).glob(f"{source_name}.json"))
+ if candidates:
+ transcripts[source_name] = load_words(str(candidates[0]))
+
</file context>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
-filtersinstead of hardcoding Homebrew paths; refuse lossless concat when segment resolutions mismatch (previously produced a corrupt file); escape subtitle paths safely in the filtergraph;-map 0:a?so audioless sources don't fail the composite.claude-opus-4-8, Flask debug off by default.Test plan
python3 -m py_compileon all edited files🤖 Generated with Claude Code
Summary by cubic
Improve
helpers/render.pyreliability and speed by matching source FPS, parallelizing segment encodes, surfacingffmpegerrors, and safer subtitle/libass handling. Add a compact toolkit of editing helpers plus a smallnews_server.py, and ignore large media in.gitignore.New Features
batch.py,denoise.py,reframe.py,detect_breaths.py,burn_subs_pil.py,correct_transcript.py,check_gaps.py,export_formats.py,thumbnail.py.news_server.py: shared Anthropic client, 5‑min RSS cache, modelclaude-opus-4-8, debug off by default..gitignore: ignore common media binaries and working folders at repo root.Bug Fixes
-map 0:a?to handle audioless sources; auto-pick a libass-capableffmpegwhen burning subs.ffprobeper source, encode segments in parallel (up to 4), and surfaceffmpegstderr on failure.Written for commit f44ac89. Summary will update on new commits.