Jump to content

Recommended Posts

Posted

Thought the following two quoted posts would make for a good start on how far AI has come and the many uses of AI.

On 6/14/2025 at 7:57 AM, Yessongs said:

Sunset off the balcony....wish I could get rid of the power lines once and for all LOL! 

EZ9A9265.JPG

 

21 hours ago, forcebwithu said:

Not a once and for all removal, but this is the result from Chat GPT when I asked the AI to remove the power lines. You could use this Photoroom tool for a final touch up by removing the remaining lines and aerial antenna (link).

image.png

  • Thumbs Up 4
Posted

I've found DeepSeek to be even better than Chat GBT for helping with snippets of computer code.

One project I'm working on is to determine if a location for an upcoming hash run is close to, or inside a known restricted area.

Over the years we've accumulated a number of areas where we've been asked not to set trail through or seek permission before using the area. Each of those areas are defined by a polygon of gps coordinates that I've mapped on Google Earth, exported as an XML file, and then imported into Google Sheets. I then use that data of coordinates as source data for Google Apps Script code.

I asked DeepSeek to write the code for me with this prompt "Using google apps script, write code that given a set of gps coordinates defining a boundary can tell if a gps point is outside or inside that boundary".

Within seconds DeepSeek came up with code that would have taken me hours to work through how to do. And unlike Chat GBT, the code produced by DeepSeek works out of the box.

Without boring you with the actual code produced, here's the comments and reasoning behind the code that DeepSeek produced.

/**
 * Determines if a point is outside a polygon boundary using the ray-casting algorithm.
 * 
 * @param {Object} point - The point to check {lat: number, lng: number} or [lat, lng]
 * @param {Array} polygon - Array of polygon vertices as [lat, lng] arrays
 * @return {boolean} True if the point is outside the polygon, false if inside or on the edge
 */

How It Works:

  1. The function uses the ray-casting algorithm, which draws a horizontal line to the right from the point and counts how many times it intersects with the polygon edges.

  2. If the number of intersections is odd, the point is inside the polygon. If even, it's outside.

  3. The function also handles edge cases where the point is exactly on a vertex or edge of the polygon (considered inside).

Important Notes:

  • The polygon vertices must be ordered either clockwise or counter-clockwise.

  • The first and last points don't need to be the same (the polygon is automatically closed).

  • This works for both simple and complex polygons (including those with holes, as long as the winding order is consistent).

 

Cool stuff and quite the time saver.

  • Like 3
Posted
1 hour ago, forcebwithu said:

I've found DeepSeek to be even better than Chat GBT for helping with snippets of computer code.

One project I'm working on is to determine if a location for an upcoming hash run is close to, or inside a known restricted area.

Over the years we've accumulated a number of areas where we've been asked not to set trail through or seek permission before using the area. Each of those areas are defined by a polygon of gps coordinates that I've mapped on Google Earth, exported as an XML file, and then imported into Google Sheets. I then use that data of coordinates as source data for Google Apps Script code.

I asked DeepSeek to write the code for me with this prompt "Using google apps script, write code that given a set of gps coordinates defining a boundary can tell if a gps point is outside or inside that boundary".

Within seconds DeepSeek came up with code that would have taken me hours to work through how to do. And unlike Chat GBT, the code produced by DeepSeek works out of the box.

Without boring you with the actual code produced, here's the comments and reasoning behind the code that DeepSeek produced.

/**
 * Determines if a point is outside a polygon boundary using the ray-casting algorithm.
 * 
 * @param {Object} point - The point to check {lat: number, lng: number} or [lat, lng]
 * @param {Array} polygon - Array of polygon vertices as [lat, lng] arrays
 * @return {boolean} True if the point is outside the polygon, false if inside or on the edge
 */

How It Works:

  1. The function uses the ray-casting algorithm, which draws a horizontal line to the right from the point and counts how many times it intersects with the polygon edges.

  2. If the number of intersections is odd, the point is inside the polygon. If even, it's outside.

  3. The function also handles edge cases where the point is exactly on a vertex or edge of the polygon (considered inside).

Important Notes:

  • The polygon vertices must be ordered either clockwise or counter-clockwise.

  • The first and last points don't need to be the same (the polygon is automatically closed).

  • This works for both simple and complex polygons (including those with holes, as long as the winding order is consistent).

 

Cool stuff and quite the time saver.

I always thought that method was one of Thaddeus Vincenty's contributions to the human race. He also has formula for finding ellipsoidal distance and bearing if you're interested.

 

  • Great Info 1
Posted

^^^

image.png

Jackson Pollock's "Blue Poles," owned by the National Gallery of Australia, is currently valued at $500 million. This makes it one of the most valuable paintings in the world and an unquestioned masterpiece despite its initial controversy. 

 

  • Surprised 1
Posted

This guy is a Sydney based Brit who trades options & futures and has produced a weekly Youtube commentaty on the markets for almost 15 years. His approach includes identifying professional and amatuer traders by the number of contracts in trading positions and has developed various indicators on the Tradestation platform.

https://www.youtube.com/@EminiWatch

.................................

I Let ChatGPT Design My Macro Portfolio (And You Might Too)

Barry Taylor from Emini-Watch.com
 

Last week, I paid $0 for macro analysis that hedge funds charge $250,000 a year to access. How?

I asked ChatGPT to design a portfolio for the next 40 years.

What came back was brutal, somewhat brilliant, and more concise than any fund manager report I’ve seen.

We covered (almost) everything: why rates can’t normalize, which assets survive financial repression, and the exact ETF or futures trades to execute. This AI even identified the 10x opportunity that most funds are missing. Your move, Ray Dalio.

WARNING: 100% AI-Generated Content Ahead

Everything below was created by ChatGPT, an AI that’s smarter than your fund manager but has zero liability for your losses. This is NOT investment advice – it’s what happens when you ask a computer to think about money. The AI recommended leveraged positions that could either make you rich or blow up your account. It doesn’t care which. Consult real advisors who have licenses to lose. Then use this to make them earn their fees. Don’t be the idiot who loses their house because “the AI said uranium was going to 10x”. You’ve been warned.

[ And this Disclaimer was written by AI too! ]

Here’s the complete conversation, the prompts I used, and the ETF and futures portfolio ChatGPT designed.


AI Prompt #1: From 1980 to 2020, interest rates in the US and around the world declined steadily. Why?

Interest rates didn’t fall continuously from 1980 to 2020, but they did trend dramatically downward over that 40-year period. Here’s why:

Starting Point: The Volcker Shock (1979-1982)

    Fed funds rate peaked at 20% in 1981
    10-year Treasury yields hit 15.8%
    Paul Volcker crushed 1970s inflation with extreme monetary tightening

The Great Moderation (1980s-2007) Key drivers of declining rates:

    Inflation conquered – From 13.5% (1980) to 2-3% target range
    Globalization – Cheap labor from China/emerging markets = deflationary pressure
    Technology – Productivity gains kept prices down
    Demographics – Aging populations save more, spend less
    Debt accumulation – Higher debt levels made economies more rate-sensitive

Financial Crisis Acceleration (2008-2020)

    Central banks went to zero (ZIRP)
    Quantitative easing pushed long-term rates down
    Europe/Japan went negative
    “Secular stagnation” – structurally lower growth potential

The Feedback Loop: Lower rates enabled more debt -> More debt required lower rates -> Central banks couldn’t raise rates without crushing the economy. By 2020, the 10-year Treasury hit 0.5% (pandemic low). The 40-year bond bull market was the defining macro trade of multiple generations. The trend reversed violently in 2021-2022 with inflation’s return, but that’s outside your timeframe.

AI Prompt #2: During this “defining macro trade”, what macro assets were the best and worst to hold?

The Winners (1980-2020):

1. Long-Duration Bonds – The purest play

    30-year Treasury total return: ~11% annualized
    Starting yield 10%+ -> ending yield <2%
    Price gains + coupon = generational wealth creator

2. US Equities (especially growth/tech)

    S&P 500: ~11.5% annualized
    NASDAQ (1985-2020): ~12.5% annualized
    Lower rates = higher multiples, especially for long-duration growth

3. Real Estate

    REITs: ~12% annualized
    Direct property ownership leveraged at 70-80% LTV printed money
    Cap rates compressed from 10%+ to 4-6%

4. Private Equity

    Top quartile funds: 15-20% net IRRs
    Cheap leverage + multiple expansion = the golden age

AI-BARRY PODCAST: Death of the 60/40 Portfolio and the Next 40-Year Macro Trade

Article too long? Let AI-Barry explain why your portfolio is dying and what to own for the next 40 years. Yes, that’s an AI voice reading an AI-generated global macro strategy – the future is now, deal with it.

https://emini-watch.com/wp-content/uploads/2025/06/ai-macro-portfolio-14Jun25.mp3

 

The Losers:

1. Cash – Death by a thousand cuts

    Average yield fell from 10%+ to 0%
    Lost ~2-3% annually to inflation
    Opportunity cost: catastrophic

2. Commodities

    Gold: Only ~5% annualized (saved by 2000s bull run)
    Oil: Negative real returns
    Agricultural: Crushed by technology/productivity

3. Japanese Assets (1990-2020)

    Nikkei: -50% from peak, still not recovered
    JGBs: Positive but minimal returns despite ZIRP

4. Value Stocks

    Massively underperformed growth
    Banks/utilities/industrials left behind

The Leverage Multiplier: Any leveraged long position (real estate, PE, levered equities) crushed unlevered positions. The cost of capital collapse was THE trade. The hard truth is that if you held cash or “stayed defensive” waiting for rates to normalize, you got destroyed. The Fed put was real, and fighting it was financial suicide.
AI Prompt #3: The defining feature of current developed economies is that governments and individuals have built up extreme debt. Does this mean that interest rates will never rise to the levels seen in 1980? What other macro overlay will define the next 40 years?

You’re right – we can’t go back to 1980s rates. The math doesn’t work.

The Debt Trap Reality:

    US debt/GDP: 50% (1980) -> 130% (2024)
    Corporate debt: $3T -> $13T
    Household debt: $1.4T -> $17T
    At 10% rates, US interest expense alone = 50% of tax revenue

Rate ceiling is probably 5-7% before something breaks. We saw this in 2022-2023:

    UK pension crisis at 4.5% gilt yields
    Regional bank failures at 5% Fed funds
    Commercial real estate implosion beginning

The Next 40-Year Overlay: Financial Repression 2.0

1. Negative Real Rates Forever

    Inflation allowed to run at 3-4%
    Nominal rates kept at 2-3%
    Stealth debt debasement: -1-2% real yields

2. Yield Curve Control (YCC)

    Japan model goes global
    Central banks cap long-term rates
    Financial markets become policy tools, not price discovery

3. Currency Debasement Race

    All developed markets debase together
    Real assets (property, commodities, Bitcoin) become monetary refuges
    Wealth inequality explodes further

4. Geopolitical Fragmentation

    Dollar hegemony challenged
    Multiple currency/trading blocs
    Higher structural inflation from deglobalization

5. Demographics Doom Loop

    Aging populations = more entitlements
    Fewer workers = higher wages = inflation
    Immigration becomes economic necessity

The Trade for Next 40 Years:

    Long: Real assets, emerging markets, commodity producers
    Short: Currency, bonds at par, pension promises
    Variable: Equities (winners/losers diverge massively)

Bottom Line: We’re Japan with worse demographics and no social cohesion. Rates can spike temporarily (like 2022), but the debt math forces them back down. The next crisis won’t be inflation – it’ll be currency/sovereign debt confidence. Prepare for financial repression, not rate normalization.


AI Prompt #4: Under this scenario, what macro asset classes will outperform? What asset classes should I hold long and short?

Core Longs (Leverage These):

1. Residential Real Estate – 70-80% LTV

    Governments will protect this at all costs (votes)
    Fixed-rate mortgages = free inflation hedge
    Supply constrained in tier-1 cities
    Target: 15-20% IRR with leverage

2. Commodity Producers – 1.5-2x leveraged ETFs

    Energy (XOM, CVX) not just oil futures
    Copper/Lithium miners for electrification
    Fertilizer (NTR, MOS) – food inflation inevitable
    Own the equity, not the commodity

3. Emerging Market Local Currency Debt – 2-3x via swaps

    Brazil, Mexico, India
    8-12% yields while USD debases
    EM central banks can actually raise rates

Tactical Longs (No Leverage):

1. Bitcoin – Unlevered only

    Sovereign/currency crisis hedge
    Digital gold thesis plays out
    5-10% portfolio max

2. Infrastructure/Utilities

    Government-guaranteed returns
    Inflation pass-throughs built in
    Boring = beautiful in chaos

3. Defense Contractors

    Geopolitical fragmentation = spending boom
    LMT, RTX, NOC print regardless

Core Shorts (Use Options/Structured Products):

1. Long-Duration Sovereign Bonds

    German Bunds, JGBs, USTs >10yr
    Asymmetric risk/reward at these yields
    Use put spreads, not outright shorts

2. European Banks

    Negative rates killed their model
    Deutsche, Credit Suisse analogs
    Demographic/regulatory death spiral

3. Traditional Retail REITs

    Structural decline + leverage = death
    Malls, office (excluding prime/converted)

Currency Pairs:

    Long: NOK, AUD, CAD (commodity currencies)
    Short: EUR, JPY (demographic disasters)
    Use 3-5x forex leverage max

The Barbell Structure:

    40% Real assets (leveraged)
    30% EM/Commodity equity
    20% Cash-flowing defensive
    10% Convex hedges (puts, bitcoin)

Critical Execution Rules:

    Never leverage rates-sensitive assets
    Roll fixed-rate debt eternally – never pay it off
    Own cash flow, not promises
    Geographic diversification mandatory

What Dies:

    60/40 portfolios
    Investment-grade bonds
    European equities
    Pension funds (short pension-heavy cos)

The Uncomfortable Truth: You’re not investing anymore – you’re front-running policy. The best trade isn’t finding value; it’s positioning where governments MUST inject liquidity. Act accordingly.
AI Prompt #5: For the above recommendations, tell me the exact ETF or leveraged ETF to hold.

LEVERAGED LONGS:

Real Estate

    XLRE – Real Estate Select Sector (1x base)
    URE – ProShares Ultra Real Estate (2x)
    DRN – Direxion Daily Real Estate Bull 3x

Commodity Producers

    XLE – Energy Select Sector (1x base)
    ERX – Direxion Daily Energy Bull 2x
    GUSH – Direxion Daily S&P Oil & Gas 2x
    NUGT – Direxion Daily Gold Miners 2x
    COPX – Global X Copper Miners

Emerging Markets

    EEM – iShares MSCI Emerging Markets (1x base)
    EDC – Direxion Daily Emerging Markets Bull 3x
    EMLC – VanEck EM Local Currency Bond

UNLEVERAGED LONGS:

Bitcoin

    BITO – ProShares Bitcoin Strategy
    GBTC – Grayscale Bitcoin Trust
    Or just buy actual Bitcoin

Infrastructure/Utilities

    IGLB – iShares Global Infrastructure
    XLU – Utilities Select Sector
    TOLL – Direxion Global Infrastructure

Defense

    ITA – iShares US Aerospace & Defense
    PPA – Invesco Aerospace & Defense

SHORT POSITIONS:

Long-Duration Bonds

    TBT – ProShares UltraShort 20+ Year Treasury (2x inverse)
    TTT – ProShares UltraPro Short 20+ Year (3x inverse)
    TBF – ProShares Short 20+ Year Treasury (1x inverse)

European Banks

    EUFN – iShares MSCI Europe Financials (short this)
    EPV – ProShares UltraShort FTSE Europe (2x inverse)

Retail REITs

    SRS – ProShares UltraShort Real Estate (2x inverse)
    Or short SPG, MAC, SKT directly

CURRENCY PLAYS:

    UUP – Dollar Bull (short this for dollar weakness)
    FXA – Australian Dollar Trust (long)
    FXC – Canadian Dollar Trust (long)

    FXE – Euro Trust (short)
    FXY – Japanese Yen Trust (short)

Portfolio Construction:

    40% URE/DRN + owned real estate
    20% ERX/GUSH/COPX
    15% EDC/EMLC
    10% IGLB/ITA/XLU
    10% TBT/TTT
    5% Bitcoin

Warning: Leveraged ETFs decay over time due to daily rebalancing. Use them for 3-12 month trades, not buy-and-hold. For core positions, use 1x ETFs with margin or options for leverage. The play is to own things governments can’t print. Short things they must debase. Use their leverage against them.


AI Prompt #6: Which asset class or ETF play has the greatest upside potential?

Greatest Upside Potential: Uranium

Here’s why uranium crushes everything else:

Supply/Demand Violence:

    200M lbs demand vs 140M lbs production
    60M lb structural deficit
    No substitutes, reactors MUST buy

Price Dynamics:

    Spot: $100/lb (was $20 in 2020)
    Utilities need $150+ to incentivize new supply
    Last cycle hit $140 (2007)

The Kicker: China building 150 reactors, US/Europe reversing nuclear policy, data centers need baseload power for AI

The Plays:

    URNM – Sprott Uranium Miners ETF (1x)
    URA – Global X Uranium ETF (1x)
    CCJ – Cameco (largest producer)
    SRUUF – Sprott Physical Uranium Trust (owns physical)

Asymmetric Risk/Reward Ranking:

    Uranium – 5-10x potential over 5 years
    Bitcoin miners – 3-5x if BTC hits $200k
    EM local currency debt – 3x with carry
    Turkish/Brazilian equities – 3-4x
    US residential real estate (leveraged) – 2-3x

The Contrarian Take: Everyone’s worried about rates/recession. Nobody’s positioned for an energy supercycle driven by:

    AI power demand (each ChatGPT query = 10x Google search energy)
    Electrification mandates
    Deglobalization (energy security)

Maximum Violence Portfolio:

    30% URNM
    20% MARA/RIOT
    20% ERX
    20% EDC
    10% TBT

Reality Check: Leveraged ETFs are weapons, not investments. If you can’t handle 50% drawdowns, stick to the barbell approach I outlined earlier. This either makes you rich or you blow up spectacularly. No middle ground.
AI Prompt #7: For the above recommendations, tell me the exact futures market trades to make.

LONG POSITIONS:

Energy/Commodities (Instead of XLE, ERX, GUSH):

    CL – WTI Crude Oil: $10/barrel move = $10,000 per contract
    NG – Natural Gas: $0.10 move = $1,000 per contract
    RB – RBOB Gasoline: $0.01 move = $420 per contract
    Position size: 2-3x your normal futures allocation (matching leveraged ETF exposure)

Precious Metals (Instead of NUGT):

    GC – Gold: $10/oz move = $1,000 per contract
    SI – Silver: $0.50/oz move = $2,500 per contract
    Use 2:1 ratio (2 gold to 1 silver)

Industrial Metals (Instead of COPX):

    HG – Copper: $0.05/lb move = $1,250 per contract
    This is pure copper exposure vs mining equity beta

Emerging Markets (Instead of EEM, EDC):

    EMD – MSCI Emerging Markets Mini: $5/point move = $250 per contract
    More liquid than full-size EM futures
    Match 3x ETF exposure with position sizing

Bitcoin (Instead of BITU, miners):

    BTC – CME Bitcoin Futures: $5/point = $25 per point move
    MBT – Micro Bitcoin: $5/point = $5 per point move
    Use MBT for precision sizing

SHORT POSITIONS:

Bonds (Instead of TBT, TTT):

    ZN – 10-Year Treasury Note: SHORT 2-3 contracts per $100k
    ZB – 30-Year Treasury Bond: SHORT 1-2 contracts per $100k
    ZT – 2-Year Treasury Note: LONG as spread hedge
    Classic steepener: Long ZT, Short ZB

Currencies (Direct plays): LONGS:

    6A – Australian Dollar: $100k AUD per contract
    6C – Canadian Dollar: $100k CAD per contract
    6N – NZ Dollar: $100k NZD per contract

SHORTS:

    6E – Euro: $125k EUR per contract
    6J – Japanese Yen: ¥12.5M per contract
    DX – Dollar Index: SHORT for broad USD weakness

FUTURES-ONLY MACRO PORTFOLIO:

Core Positions (40% margin):

    LONG 2 CL (crude oil)
    LONG 1 GC (gold)
    LONG 1 HG (copper)
    LONG 2 MBT (micro Bitcoin)
    SHORT 2 ZN (10-year Treasury)

Satellite Trades (20% margin):

    LONG 1 6A (Aussie dollar)
    SHORT 1 6E (Euro)
    LONG 3 EMD (emerging markets)

Spread Trades (Lower Risk):

    Calendar spreads on CL (long far month, short near)
    Intermarket: Long HG/Short GC ratio
    Crack spread: Long 3 RB + HO, Short 2 CL

CRITICAL EXECUTION NOTES:

Margin Management:

    Initial margin: ~$5-15k per contract
    Keep 50% excess margin for volatility
    Never exceed 40% margin utilization

Roll Strategy:

    Roll 2 weeks before expiry
    Use calendar spreads to reduce cost
    Never hold into delivery month

The Uranium Problem: No direct uranium futures exist. Your options:

    Trade uranium miners via equity futures (ES/NQ exposure)
    Use UX spot market (requires special broker)
    Stick with URNM/CCJ in equity account

Maximum Leverage Setup:

    30% Crude oil (3 CL)
    20% Precious metals (2 GC, 1 SI)
    20% Bitcoin (4 MBT)
    20% Bond shorts (3 ZN short)
    10% FX plays

Warning: Futures give you 10-20x leverage built-in. One CL contract controls $70,000 of oil with $7,000 margin. This is a double-edged sword. Size accordingly or you’ll blow up. Most of these futures markets trade 23 hours a day. No ETF decay. No management fees. Pure, direct exposure. But also no forgiveness for bad risk management.

The Brutal Truth

AI is your new macro analyst. No ego. No biases. No need to justify last quarter’s bad calls. Just cold, hard analysis of debt dynamics, currency debasement, and the trades that matter. While your financial advisor is still explaining why bonds are “safe”, AI just mapped out why uranium could 10x and residential real estate with leverage beats everything.

The game has changed. The gatekeepers are dead. And if you’re still paying someone to tell you to “stay diversified” while they collect management fees on your melting purchasing power, you deserve what’s coming. The future belongs to those who adapt – your move.
 

 

 

https://emini-watch.com/global-macro/ai-macro-portfolio/?utm_source=Emini-Watch+Email+Subscribers&utm_campaign=02f250b726-EMAIL_CAMPAIGN_2025_06_14_09_33&utm_medium=email&utm_term=0_-02f250b726-73845237

 

About the Author

Barry Taylor is a full-time trader, founder of Emini-Watch.com and developer of the ‘Better’ Trading Indicators. The ‘Better’ Indicators are a unique set of 3 non-correlated indicators that will give you an edge, whether you’re a day trader, swing trader or investor. With over 14 years of full-time trading and traveling, Barry splits his time between Byron Bay, Biarritz and Kauai.

  • Great Info 1
Posted

DSCF9255-Edit-2-Edit.jpg

DSCF9292-Edit.jpg

DSCF9295-Edit.jpg

Lightroom and Photoshop have had recent updates.  The above backgrounds all created without any creativity on my part.  Initial images were taken at the Terminal 21 Car on Light Event.

DSCF9298-Edit.jpg

Posted

I've got an old movie from the 1980's (Gene Hackman's version of 'Misunderstood'), but there are no subtitles available anywhere that I've been able to find. Transcribing the film's English audio track and then converting it into a time-stamped SubRip (.srt) file would be a perfect job for AI, it's a really horrible, fiddly, tedious task to try and do manually. There are indeed plenty of tools out there that offer to do this for a (not insubstantial) fee, but nothing that's free as far as I can see. Does anyone have any experience of creating SubRip files directly from an audio track via AI, or know of a free tool that will do so for you? 

Posted
3 minutes ago, Toy Boy said:

I've got an old movie from the 1980's (Gene Hackman's version of 'Misunderstood'), but there are no subtitles available anywhere that I've been able to find. Transcribing the film's English audio track and then converting it into a time-stamped SubRip (.srt) file would be a perfect job for AI, it's a really horrible, fiddly, tedious task to try and do manually. There are indeed plenty of tools out there that offer to do this for a (not insubstantial) fee, but nothing that's free as far as I can see. Does anyone have any experience of creating SubRip files directly from an audio track via AI, or know of a free tool that will do so for you? 

A great question to ask ChatGPT.

Is there an AI that can create a SubRip (.srt) file directly from an audio track?
ChatGPT said:

Yes, there are AI tools that can create SubRip (.srt) subtitle files directly from an audio track. These tools use automatic speech recognition (ASR) and often include speaker diarization, timestamp alignment, and optional translation. Here are some reliable options:


🔹 1. Whisper by OpenAI (Open-source, highly accurate)

  • Type: Local or cloud-based (via command line or GUI wrappers)

  • Input: Audio file (.mp3, .wav, etc.)

  • Output: Can directly generate .srt, .vtt, and .txt files

  • Pros: High accuracy, supports multiple languages

  • Cons: Needs setup (unless using GUI tools)

  • How to use:

     
    bash
    whisper your_audio.mp3 --model medium --output_format srt
  • GitHub repo


🔹 2. Whisper Web GUIs / Wrappers

If you want to avoid code:


🔹 3. Subtitle Edit (Windows GUI with AI support)

  • Free software with Whisper AI and Google STT support

  • Can generate and edit .srt files

  • Includes auto-alignment and translation tools

  • Download Subtitle Edit


🔹 4. Online Tools (No install required)

a. Veed.io

  • Upload audio → Auto-transcribe → Export .srt

  • Freemium (limits on length/export)

b. Happy Scribe

  • More professional-grade; higher accuracy

  • Charges per minute after free trial

c. Kapwing

  • Quick online subtitle generation with .srt export


Summary Table

Tool Accuracy Free SRT Export Local/Online
Whisper (CLI) Local
Subtitle Edit Local
Veed.io Limited Online
Happy Scribe Online

Would you like step-by-step help setting up Whisper locally or using one of these online tools?

  • Great Info 1
  • Thumbs Up 1
Posted

I asked ChatGPT a follow up question.

Which of the AI tools are free?
ChatGPT said:

Here's a breakdown of AI tools that are free (or have free versions) for generating .srt subtitle files from audio:


Completely Free Tools

🔹 1. [Whisper by OpenAI (CLI version)]

  • Cost: Free and open-source

  • What you need: Python installed + command line usage

  • Pros: Very accurate, supports .srt directly, multilingual

  • How to use:

     
    bash
    whisper audio.mp3 --model medium --output_format srt
  • Download: GitHub – openai/whisper

🔹 2. Whisper.cpp (Fast local version)

  • Cost: Free and open-source

  • What you need: Compiled binary or run with Docker

  • Pros: Lighter, faster on CPU than Whisper Python

  • GUI: Optional GUIs available (like whisper-webui)

  • Download: GitHub – ggerganov/whisper.cpp

🔹 3. [Subtitle Edit (Windows only)]


🟡 Free With Limits (Freemium Tools)

🔹 4. Veed.io

  • Free Tier: Limited video/audio length, watermarks may apply

  • Export: .srt, .txt, embedded subtitles

  • Usage: Upload audio → Auto-subtitle → Export

🔹 5. Kapwing

  • Free Tier: Limited exports per month, file duration limits

  • Export: .srt, hardcoded subtitles

  • Usage: Very easy via browser


Paid Tools Only (or Trial-limited)

🔹 6. Happy Scribe / Sonix / Trint

  • Free trial only (5–10 minutes)

  • After that: Pay-per-minute

  • Professional use: Great for teams, high-quality outputs


Summary Table

Tool Cost SRT Export Notes
Whisper (CLI) Free Best for accuracy, CLI-based
Whisper.cpp Free Lightweight, fast
Subtitle Edit Free Windows GUI, powerful editing
Veed.io ⚠️ Freemium Browser, limited exports
Kapwing ⚠️ Freemium Easy, some limitations
Happy Scribe Paid Free trial only

If you prefer something completely free and offline, go with:

  • Whisper CLI (OpenAI) or

  • Subtitle Edit with Whisper model

Would you like a quick guide for setting up Whisper or using Subtitle Edit with Whisper?

  • Great Info 1
  • Thumbs Up 1
Posted (edited)

Sure... lots of AI girly pics.

Some of them titillating mild SFW, some of them for advertising, most of it NSFW porn (including: illegal deep fakes, revenge porn & kitty porn 😻).

Scroll down for a brief literature review.

. . .

506473793_122111125946894438_8575246255036755893_n.jpg

. . .

The Present and Future of Adult Entertainment: A Content Analysis of AI-Generated Pornography Websites

https://pubmed.ncbi.nlm.nih.gov/40032709/

"Fueled by advances in artificial intelligence (AI), the adult entertainment industry is undergoing a significant transformation. AI-generated pornography-or AI porn-is reshaping how people create and consume sexually explicit content, progressively offering rapid, mass access to large quantities of interactive and highly customizable experiences. 

...Carrying significant social and ethical implications, these findings point to a gradual evolution toward an AI-driven porn landscape where individuals can create and interact with sexual content tailored to their preferences and fantasies..."

. . .

Generative AI pornography

https://en.wikipedia.org/wiki/Generative_AI_pornography

"Generative AI pornography or simply AI pornography is a digitally created pornography produced through generative artificial intelligence (AI) technologies. Unlike traditional pornography, which involves real actors and cameras, this content is synthesized entirely by AI algorithms. These algorithms, including Generative adversarial network (GANs) and text-to-image models, generate lifelike images, videos, or animations from textual descriptions or datasets..."

. . .

Best AI Porn Sites of 2025
What was once the realm of speculative fiction is now a tangible, rapidly evolving industry. What does that mean for the future of intimacy?

https://chicagoreader.com/adult/ai-porn-sites/

"...Rather than replacing human connection, AI generated intimacy has the potential to complement it, serving as a tool for self-discovery, social confidence, and personal well-being. By engaging with these technologies responsibly and thoughtfully, we can harness their benefits without losing sight of the irreplaceable depth of real human relationships.

As technology continues to evolve, so too must our conversations around its responsible use. This article will explore the complex landscape of AI generators and the adult content they produce, its potential as a harmless escape, its dangers as a slippery slope, and the ethical quagmire it presents..."

. . .

Porn and AI Are the Real Sex Educators

https://www.psychologytoday.com/us/blog/the-future-of-intimacy/202506/porn-and-ai-are-the-real-sex-educators

"...The average age of first exposure to pornography is now 12, with 15% of kids seeing it before age 10. According to a 2022 Common Sense Media survey, more than half of kids reported seeing adult content accidentally—often by clicking on a link they didn’t realize would lead to porn.

Unintentional exposure to pornography seems to be a common experience for teens, as 63% of those who said they have seen pornography only accidentally reported that they had been exposed to pornography in the past week. By age 17, three out of four teens have viewed pornography, 41% have seen it during the school day, and 45% felt that online pornography gave helpful information about sex..."

. . .

AI and the future of sex
The rise of AI porn could change our expectations of relationships.

https://www.technologyreview.com/2024/08/26/1096526/ai-sex-relationships-porn/

"...Anti-porn crusades have been at the heart of the US culture wars for generations, but by the start of the 2000s, the issue had lost its hold. Smartphones made porn too easy to spread and hard to muzzle. Porn became a politically sticky issue, too entangled with free speech and evolving tech. An uneasy truce was made: As long as the imagery was created by consenting adults and stayed on the other side of paywalls and age verification systems, it was to be left alone. 

But today, as AI porn infiltrates dinner tables, PTA meetings, and courtrooms, that truce may not endure much longer. The issue is already making its way back into the national discourse; Project 2025, the Heritage Foundation–backed policy plan for a future Republican administration, proposes the criminalization of porn and the arrest of its creators...

...The rise of AI-generated porn may be a symptom of a new synthetic sexuality, not the cause. In the near future, we may find this porn arousing because of its artificiality, not in spite of it."

. . .

SF shuts down 10 of the world's most-visited websites using AI to generate explicit content

https://abc7news.com/post/deepfake-porn-san-francisco-shuts-down-10-worlds-most-visited-websites-using-ai-generate-explicit-content/16638231/

"...San Francisco is known as the AI capital of the world and now the city's attorney wants to make sure this home is protecting adults and minors against AI generated deepfake images used for explicit purposes.

"The investigation and this work has brought into the darkest corners of the internet we are talking about the use of artificial intelligence to generative non-consensual pornography," said Chiu.

In the lawsuit, City Attorney Chiu goes after 16 website owners and operators who he says violate state and federal laws prohibiting deepfake pornography, revenge pornography, and child pornography. 10 of those sites are now gone..."

 

Edited by lazarus
....
  • Great Info 1
Posted

@forcebwithu, thanks! I'd already looked at some of those sites and the ones I tried all had limitations that would mean paying to extract the subs from a full movie. I'd never heard of Whisper, though, and as Subtitle Edit is my default editing tool and I know it well that app with Whisper sounds perfect. 

I followed the Google AI instructions, accepting the suggested defaults along the way. In what must be about the final step you have to select the Whisper model:

image.png

I tried to download three of them but all it does is to DL a 1 KB bin file to a Whisper subfolder in the Subtitle Edit app.

image.png

 

When I try opening that with Subtitle Edit it tells me I need a username and password which I don't have. 

image.png

However, Google AI tells me that I don't need these, while the app itself says that I do, which leaves me at a bit of an impasse! lol 

image.png

  • Like 1
  • Great Info 1
  • Thumbs Up 1
Posted

I realised that my version of Subtitle Edit might be out of date I downloaded the latest version and that one installs the models with no issues. It did fling up an exception, though, which means zilch to me. I tried two different models and they both crashed during post-processing, saying that I already have an exception set. WTF??? 

So I edited the log and removed the exception which is apparently added by the program to the Whisper command line. I then tried yet another model and it seemed to work, at least it didn't throw up any errors and this is what it's doing right now....

image.png

"Time remaining 900 minutes".... I thought AI was supposed to do difficult stuff quickly and painlessly, I could manually transcribe a fair chunk of the audio in 15 hours. I'm also thinking that this process is just going to lead to a non-time-stamped transcription, when easily the most difficult part of creating subtitles is getting the timings correct. 

 

  • Thumbs Up 2
Posted
44 minutes ago, Toy Boy said:

I realised that my version of Subtitle Edit might be out of date I downloaded the latest version and that one installs the models with no issues. It did fling up an exception, though, which means zilch to me. I tried two different models and they both crashed during post-processing, saying that I already have an exception set. WTF??? 

So I edited the log and removed the exception which is apparently added by the program to the Whisper command line. I then tried yet another model and it seemed to work, at least it didn't throw up any errors and this is what it's doing right now....

image.png

"Time remaining 900 minutes".... I thought AI was supposed to do difficult stuff quickly and painlessly, I could manually transcribe a fair chunk of the audio in 15 hours. I'm also thinking that this process is just going to lead to a non-time-stamped transcription, when easily the most difficult part of creating subtitles is getting the timings correct. 

 

Is there an option to select a small, sample portion for the subtitle generation? You could then check to see if it works before enduring a 900 hour run.

  • Thumbs Up 1
Posted
16 hours ago, forcebwithu said:

Is there an option to select a small, sample portion for the subtitle generation? You could then check to see if it works before enduring a 900 hour run.

I wasn't going to leave it running overnight so I pressed the cancel button and it allows you to save what has been done so far. It is indeed adding the time stamps, but in 90 minutes on my PC (quite a powerful one as I do a lot of video editing and other CPU-intensive tasks) it had managed to transcribe just 29 lines covering the first 4 1/2 minutes or so of the film. 

image.png

I've now fired it up again and restarted the process, the question is will it have begun again at the beginning or is it smart enough to realise that it's already done some work and simply continue from where it left off? I'm hopeful it will be the latter as everything else seems to have worked OK and the subs look like they're good quality. Whatever, it's an interesting experiment and I'm enjoying it as I'm a nerd about this stuff, I've been messing with subtitles for almost two decades. 

If it works it could be a boon. For example, I'm a fan of the British TV series "Coast". Years ago, I spent a lot of time editing and appending various SubRip files to fit the version of the three "Coast Australia" series that I had, and that worked great in the end. I've also got "Coast New Zealand" but no subs. This might allow me to extract subtitles for the series from the video files, I'm hoping so, anyway. 

  • Like 1
  • Thumbs Up 1
Posted

^^^
That gave me another idea for a question for Chat GPT.

What desktop computer systems are optimized for hosting AI apps?

When choosing a desktop computer system optimized for hosting AI apps—especially local AI models, inference tasks, or development—you’ll want to focus on systems that emphasize GPU performance, RAM, and storage speed. Here are the top options and recommendations:


🧠 Key Specs for AI Hosting

Component Recommendation
GPU (most important) NVIDIA RTX 4090 / 4080 / 3090 (or A100, H100 for extreme use); CUDA support is key
CPU AMD Ryzen 9 7950X / Intel i9-13900K or better
RAM 64 GB minimum (128 GB preferred for large models)
Storage NVMe SSD (1 TB or more) for fast data access
Cooling Strong cooling system (air or liquid) for sustained high loads
PSU 850W+ to support high-end GPU and CPU

💻 Best Prebuilt AI Workstation Desktops (2024–2025)

1. Lambda Tensor Workstation

  • Target: AI researchers, ML engineers

  • Specs: Up to 4x NVIDIA RTX or A-series GPUs

  • OS: Ubuntu pre-installed and pre-configured with deep learning libraries

  • Why: Specifically designed for AI training/inference workloads

2. Puget Systems Workstations

  • Target: Customizable professional AI setups

  • Specs: Pick your own high-end NVIDIA GPUs, AMD/Intel CPUs

  • Why: Highly customizable, built-to-order, great support

3. BOXX APEXX Systems

  • Target: AI, VFX, and engineering professionals

  • Specs: Powerful CPU and up to 3 GPUs

  • Why: Used in high-end creative and AI environments

4. HP Z by HP Workstations (Z8 G5, Z6)

  • Target: Enterprise AI and data science

  • Specs: ECC RAM, Xeon CPUs, NVIDIA RTX / A-series

  • Why: Stable, ISV-certified, often used in corporate settings


🛠️ Custom-Built AI Desktop (DIY)

If you're technically inclined, building your own system is cost-effective and more flexible:

  • Example Build (AI Inference):

    • CPU: AMD Ryzen 9 7950X or Intel Core i9-14900K

    • GPU: NVIDIA RTX 4090 (24 GB VRAM is great for large models like LLaMA, SDXL)

    • RAM: 128 GB DDR5

    • Storage: 2 TB NVMe Gen 4 SSD

    • OS: Ubuntu 22.04 LTS or Windows with WSL2 + Docker

This is excellent for running large language models locally (e.g., LLaMA 3, Mistral, or even fine-tuned GPT-J), Stable Diffusion, or AI API services.


🧰 Optimizations for AI Hosting

  • Use Docker to manage AI environments cleanly

  • Install CUDA + cuDNN for NVIDIA GPU acceleration

  • Use WSL2 (on Windows) to run Linux-native AI tools

  • Run models with frameworks like TensorFlow, PyTorch, ONNX Runtime, or Ollama for LLMs


Summary

Use Case Best Option
Out-of-the-box AI system Lambda or Puget Systems
DIY power user Custom build with RTX 4090 + Ryzen 9/i9
Enterprise or research HP Z Workstations or BOXX
Budget AI testing RTX 4070/4060 build with 32–64 GB RAM

Would you like recommendations for a specific budget range or use case (e.g., image generation, LLMs, chatbots)?

  • Like 1
Posted
2 hours ago, forcebwithu said:

you’ll want to focus on systems that emphasize GPU performance, RAM, and storage speed

That might explain why it's so slow, I've got 16 GB of RAM and the fastest processor that was available when the PC was built for me, but as I'm not interested in games I went for a standard graphics card. 

Actually, it shows how little I know about AI, but I thought that the Nvidia GPU's were only essential for training the systems, not for running the end-result? I mean, you can run small AI models on your phone and that's hardly got a high-end GPU in it.  

  • Like 1
Test: whoReadTheTopic output below
×
×
  • Create New...