Grok 4.3 can Take in Video and Extract Audio files

@techdevnotes Camtasia can do that and has been doing it for years. I can download any video from a source and load it into Camtasia, separate audio from with a click. Now just save audio as an MP3. Done.

@techdevnotes @elonmusk The best tool.🔥👏

@techdevnotes I've been doing this for years with a simple .MP4 drag-n-drop into the free application Audacity. AI? For what??

Ready to bring ideas to life? Try Grok Imagine today

@techdevnotes Can be done in under a minute manually but I am sure there are applications beyond the basic function I don't see.

@techdevnotes Only seems available to Super Grok Heavy users right now...

@techdevnotes @elonmusk So hyper accurate transcription? Courts gonna be a lot of fun for people soon.

@techdevnotes @elonmusk That’s a real quality-of-life upgrade turning video into usable audio instantly just makes everything smoother.

@techdevnotes That's awesome! What do you mean by extract audio? Can it actually listen/analyze the way it currently can with images?

@techdevnotes @FFmpeg with extra steps

Hey @grok , how lethal is this? Full ReasoningFlowMonitor Class ```python import numpy as np class ReasoningFlowMonitor: """ GrokDojoHexFlow Reasoning Flow Monitor Lightweight rolling-window monitor for detecting silent degradation in LLM/agent reasoning flows. Inspired by the Anthropic Claude case (read-to-edit ratio drop from \~6.6 → 2.0). Uses 10-session rolling windows + delta alerts to catch drops early. """ def __init__(self, window_size: int = 10, ratio_threshold: float = 0.15, quality_threshold: float = 0.08): self.history = [] # list of session dicts self.window_size = window_size self.ratio_threshold = ratio_threshold # e.g. >15% drop self.quality_threshold = quality_threshold # e.g. >0.08 drop def add_session(self, session_id: int, read_edit_ratio: float, latency: float, quality: float): """Add a new session and check for degradation.""" self.history.append({ 'session': session_id, 'ratio': read_edit_ratio, 'latency': latency, 'quality': quality }) self._check_rolling_window() def _check_rolling_window(self): """Check the most recent window for significant deltas.""" if len(self.history) < self.window_size: return # Current window window = self.history[-self.window_size:] ratios = [s['ratio'] for s in window] qualities = [s['quality'] for s in window] avg_ratio = np.mean(ratios) avg_quality = np.mean(qualities) # Compare to previous window if available if len(self.history) > self.window_size: prev_window = self.history[-self.window_size*2 : -self.window_size] if prev_window: prev_avg_ratio = np.mean([s['ratio'] for s in prev_window]) prev_avg_quality = np.mean([s['quality'] for s in prev_window]) # Safe ratio delta calculation ratio_delta = (prev_avg_ratio - avg_ratio) / prev_avg_ratio if prev_avg_ratio > 0 else 0 quality_delta = prev_avg_quality - avg_quality if ratio_delta > self.ratio_threshold: print(f"⚠️ DELTA ALERT: Read-to-edit ratio dropped {ratio_delta*100:.1f}% " f"in last {self.window_size} sessions") print(f" Current window avg ratio: {avg_ratio:.2f} (prev: {prev_avg_ratio:.2f})") if quality_delta > self.quality_threshold: print(f"⚠️ DELTA ALERT: Quality score dropped {quality_delta:.3f} " f"in last {self.window_size} sessions") print(f" Current window avg quality: {avg_quality:.3f} (prev: {prev_avg_quality:.3f})") def summarize(self): """Print overall statistics.""" if not self.history: print("No sessions recorded yet.") return ratios = [s['ratio'] for s in self.history] latencies = [s['latency'] for s in self.history] qualities = [s['quality'] for s in self.history] print(f"\n=== GrokDojoHexFlow Overall Summary ===") print(f"Total sessions: {len(self.history)}") print(f"Average read-to-edit ratio: {np.mean(ratios):.2f}") print(f"Average latency: {np.mean(latencies):.2f}s") print(f"Average quality score: {np.mean(qualities):.3f}") print("=======================================")

@techdevnotes cool demo but until it nails transcripts with accents, it's half baked

@techdevnotes @elonmusk What’s FFmpeg?

Exceptional comfort, luxurious softness, and the Mott & Bow guarantee that every stitch is handcrafted to perfection—these are the only tees you should be buying.

@techdevnotes @elonmusk Batch audio files for transcription when? I need to ditch turboscribe 🙏🙏

@techdevnotes Feature i waited for🫶

@techdevnotes @elonmusk This is huge!

@techdevnotes I need a real world scenario where you would need this ability.

@techdevnotes @xai As a supergrok subscriber, I am disappointed that you are forcing an even more expensive heavy supergrok requirement for this. WTF

@techdevnotes massive workflow unlock.

@techdevnotes @elonmusk Listentoyoutube dot com was my jam. This will be nice

@techdevnotes Grok has the raw power to challenge and beat Anthropic models. 4.20 is already insanely powerful when doing agent builds.

With more than 8,500 satellites in orbit, Starlink provides connectivity through harsh weather conditions and natural disasters. Order online in under 2 minutes.

@techdevnotes I'm going to force Grok to watch old episodes of Mr. Ed.

@techdevnotes The only bottleneck I can see with this is speed, if that has been solved then this is a real tool with serious applications

@techdevnotes @elonmusk Grok groks

@techdevnotes @elonmusk I wish it sounded like Elon

@techdevnotes Not embarrassed to be an Xai fan anymore

@techdevnotes @elonmusk SuperGrok subscriber. Persistent bug in long chat: “Agents thinking” stuck 200–500s, then “Failed to send request”. Cross-Chat Memory feature (like ChatGPT) would help a lot. Thanks! 🙏

@techdevnotes Can I feed it a YouTube URL and have it spit out the transcript?

@techdevnotes so you allowed audio to be uploaded is all audio its been able to hear already this really isnt impressive for a chat bot

@techdevnotes Does the model handle this natively or is it just speech to text?

Exclusively in Theaters Now. Get Tickets!

@techdevnotes So can FFmpeg. Still, nice to see that support for more input and output formats is something being actively developed. Is the image reading any better? It's been quite mediocre in Grok for quite a while now.

@techdevnotes Will this now allow uploading video to imagine to edit? I upload photos all the time to edit and turn into videos, but I would love to be able to upload my own videos.

@techdevnotes @elonmusk @grok can you create GarageBand files yet?

@techdevnotes Wow, this is shaping up to be a big upgrade. I am all for this

@techdevnotes is memory working? That's my main concern. I NEED THE GROK MEMORY BACK