Skip to content

Herk Edit Kit by Nate Herk: edit videos, reels, and YouTube Shorts with Codex or Claude Code. 15 skills, transcript-driven cuts, storytelling, 406 motion-graphics cards, and HyperFrames templates.

Document service setup, API credits, and transcription alternatives

Nate Herkb1afdb1dcbca
diff --git a/.agents/skills/edit-video/SKILL.md b/.agents/skills/edit-video/SKILL.md
index 79cdfcf..9a7e96f 100644
--- a/.agents/skills/edit-video/SKILL.md
+++ b/.agents/skills/edit-video/SKILL.md
@@ -11,10 +11,12 @@ missing, propose a style from `style-library/registry.json` based on the footage
1. Inspect duration, streams, and source resolution with ffprobe. Copy the source
into the project's assets or reference an explicitly provided local path.
-2. Reuse a matching word transcript. Otherwise use
- `node scripts/transcribe-elevenlabs.mjs <source>`. Transcription sends audio to
- ElevenLabs and uses the user's API credits. Existing authorization carries
- forward; make this concrete before requesting any missing authorization.
+2. Read `docs/TOOLS-AND-API-KEYS.md` before service setup. Reuse a matching word
+ transcript. Otherwise use the student's chosen provider; Nate defaults to
+ `node scripts/transcribe-elevenlabs.mjs <source>`. OpenAI Whisper and local
+ Whisper are supported workflow alternatives after transcript normalization.
+ Check available credentials or local dependencies. Explain uploads and costs
+ before any service call whose authorization is still missing.
3. Read `../cut-silences/SKILL.md`. Produce an EDL and a retimed transcript, and
render the silence pass when editing is authorized.
4. Read `../cut-mistakes/SKILL.md`. Inspect candidates in context and preserve
diff --git a/.agents/skills/short-form-edit/SKILL.md b/.agents/skills/short-form-edit/SKILL.md
index cdd71a7..a584e64 100644
--- a/.agents/skills/short-form-edit/SKILL.md
+++ b/.agents/skills/short-form-edit/SKILL.md
@@ -27,7 +27,9 @@ this test. Treat the answer as editorial judgment until audience data exists.
over aesthetic defaults. Keep all artifacts in `video-projects/<slug>/`.
2. Probe the supplied source and reference: duration, dimensions, frame rate,
audio, and existing transcript identity. Preserve originals. Transcribe with
- ElevenLabs Scribe using the workspace script. Never substitute an older take.
+ the student's chosen provider; ElevenLabs Scribe is Nate's default. Read
+ `docs/TOOLS-AND-API-KEYS.md` for Whisper alternatives, word-timestamp
+ normalization, credentials, and service costs. Never substitute an older take.
3. When a reference is supplied, inspect the entire reel using contact sheets,
individual full-size frames, and contiguous frame strips at representative
transitions. Review its sound when an audio-capable review surface is available.
@@ -232,7 +234,9 @@ human reactions where they support the story.
## Generate and assemble
-1. Use Kie text-to-video for simple moving inserts when authorized. Verify the
+1. Kie.ai is Nate's optional provider for generated moving inserts and images.
+ Use the student's chosen provider or supplied assets; check its account, credits,
+ and integration before generating. See `docs/TOOLS-AND-API-KEYS.md`. Verify the
current official API schema. Save model, prompt, task ID, selected output,
and generation state; reuse successful jobs rather than resubmitting blindly.
Download and inspect real video files, including full-size corners throughout
diff --git a/.agents/skills/short-form-video/SKILL.md b/.agents/skills/short-form-video/SKILL.md
index 9b61e04..216c732 100644
--- a/.agents/skills/short-form-video/SKILL.md
+++ b/.agents/skills/short-form-video/SKILL.md
@@ -17,7 +17,7 @@ Short-form = 1080x1920 vertical, 10–30s, talking-head face + motion-graphic sc
## The playbook (high-level)
1. **Audio is source of truth.** Edit audio FIRST (cut retakes, pauses). Save as `<name>-edit.mp4`. Measure exact duration with `ffprobe` — this is the composition's `data-duration`.
-2. **Transcribe the edited audio** with the root `scripts/transcribe-elevenlabs.mjs` helper (ElevenLabs Scribe), or if retiming an existing build with a `shift()` function in captions, keep the existing captions and just shift scene starts.
+2. **Transcribe the edited audio** with the root `scripts/transcribe-elevenlabs.mjs` helper (ElevenLabs Scribe), a student-selected alternative normalized per `docs/TOOLS-AND-API-KEYS.md`, or if retiming an existing build with a `shift()` function in captions, keep the existing captions and just shift scene starts.
3. **Author scene boundaries in edited-time** — NEVER mix original-time and edited-time anchors in the same file. See "Audio-sync protocol" below.
4. **Build the composition scaffold** (4 layers: ambient-bg, seam-treatment, captions, face) — see "Composition scaffold" below.
5. **Author scenes with LOCAL offsets** relative to each scene's `data-start`. Each scene is its own sub-composition under `compositions/scene<N>-<label>.html`.
diff --git a/.claude/skills/edit-video/SKILL.md b/.claude/skills/edit-video/SKILL.md
index 79cdfcf..9a7e96f 100644
--- a/.claude/skills/edit-video/SKILL.md
+++ b/.claude/skills/edit-video/SKILL.md
@@ -11,10 +11,12 @@ missing, propose a style from `style-library/registry.json` based on the footage
1. Inspect duration, streams, and source resolution with ffprobe. Copy the source
into the project's assets or reference an explicitly provided local path.
-2. Reuse a matching word transcript. Otherwise use
- `node scripts/transcribe-elevenlabs.mjs <source>`. Transcription sends audio to
- ElevenLabs and uses the user's API credits. Existing authorization carries
- forward; make this concrete before requesting any missing authorization.
+2. Read `docs/TOOLS-AND-API-KEYS.md` before service setup. Reuse a matching word
+ transcript. Otherwise use the student's chosen provider; Nate defaults to
+ `node scripts/transcribe-elevenlabs.mjs <source>`. OpenAI Whisper and local
+ Whisper are supported workflow alternatives after transcript normalization.
+ Check available credentials or local dependencies. Explain uploads and costs
+ before any service call whose authorization is still missing.
3. Read `../cut-silences/SKILL.md`. Produce an EDL and a retimed transcript, and
render the silence pass when editing is authorized.
4. Read `../cut-mistakes/SKILL.md`. Inspect candidates in context and preserve
diff --git a/.claude/skills/short-form-edit/SKILL.md b/.claude/skills/short-form-edit/SKILL.md
index 3f84baa..acda767 100644
--- a/.claude/skills/short-form-edit/SKILL.md
+++ b/.claude/skills/short-form-edit/SKILL.md
@@ -27,7 +27,9 @@ this test. Treat the answer as editorial judgment until audience data exists.
over aesthetic defaults. Keep all artifacts in `video-projects/<slug>/`.
2. Probe the supplied source and reference: duration, dimensions, frame rate,
audio, and existing transcript identity. Preserve originals. Transcribe with
- ElevenLabs Scribe using the workspace script. Never substitute an older take.
+ the student's chosen provider; ElevenLabs Scribe is Nate's default. Read
+ `docs/TOOLS-AND-API-KEYS.md` for Whisper alternatives, word-timestamp
+ normalization, credentials, and service costs. Never substitute an older take.
3. When a reference is supplied, inspect the entire reel using contact sheets,
individual full-size frames, and contiguous frame strips at representative
transitions. Review its sound when an audio-capable review surface is available.
@@ -232,7 +234,9 @@ human reactions where they support the story.
## Generate and assemble
-1. Use Kie text-to-video for simple moving inserts when authorized. Verify the
+1. Kie.ai is Nate's optional provider for generated moving inserts and images.
+ Use the student's chosen provider or supplied assets; check its account, credits,
+ and integration before generating. See `docs/TOOLS-AND-API-KEYS.md`. Verify the
current official API schema. Save model, prompt, task ID, selected output,
and generation state; reuse successful jobs rather than resubmitting blindly.
Download and inspect real video files, including full-size corners throughout
diff --git a/.claude/skills/short-form-video/SKILL.md b/.claude/skills/short-form-video/SKILL.md
index 9b61e04..216c732 100644
--- a/.claude/skills/short-form-video/SKILL.md
+++ b/.claude/skills/short-form-video/SKILL.md
@@ -17,7 +17,7 @@ Short-form = 1080x1920 vertical, 10–30s, talking-head face + motion-graphic sc
## The playbook (high-level)
1. **Audio is source of truth.** Edit audio FIRST (cut retakes, pauses). Save as `<name>-edit.mp4`. Measure exact duration with `ffprobe` — this is the composition's `data-duration`.
-2. **Transcribe the edited audio** with the root `scripts/transcribe-elevenlabs.mjs` helper (ElevenLabs Scribe), or if retiming an existing build with a `shift()` function in captions, keep the existing captions and just shift scene starts.
+2. **Transcribe the edited audio** with the root `scripts/transcribe-elevenlabs.mjs` helper (ElevenLabs Scribe), a student-selected alternative normalized per `docs/TOOLS-AND-API-KEYS.md`, or if retiming an existing build with a `shift()` function in captions, keep the existing captions and just shift scene starts.
3. **Author scene boundaries in edited-time** — NEVER mix original-time and edited-time anchors in the same file. See "Audio-sync protocol" below.
4. **Build the composition scaffold** (4 layers: ambient-bg, seam-treatment, captions, face) — see "Composition scaffold" below.
5. **Author scenes with LOCAL offsets** relative to each scene's `data-start`. Each scene is its own sub-composition under `compositions/scene<N>-<label>.html`.
diff --git a/.env.example b/.env.example
index 6f5cced..eeca4c3 100644
--- a/.env.example
+++ b/.env.example
@@ -1,10 +1,23 @@
-# Optional: only needed to transcribe your own footage with ElevenLabs Scribe.
+# Fill only the keys needed for your chosen workflow.
+# See docs/TOOLS-AND-API-KEYS.md for alternatives and integration requirements.
+# Nate's default transcription service; the included helper loads this key.
# Get a key with speech_to_text access at https://elevenlabs.io/app/settings/api-keys
# Copy this file to .env locally. Never commit your key.
ELEVENLABS_API_KEY=
-# Optional integrations used by the older examples.
+# Only for older examples that actually call ClickUp; use your own API settings.
CLICKUP_API_KEY=
+# Alternative Whisper API transcription or optional OpenAI generation.
+# Configure the client separately: https://platform.openai.com/api-keys
OPENAI_API_KEY=
-# Optional generated footage; configure only when using this provider.
+# Nate's optional provider for generated video and image assets.
+# Requires your own credits and model-specific integration: https://docs.kie.ai/
KIE_API_KEY=
+
+# Optional Gemini descriptions in the website-to-video workflow.
+# https://aistudio.google.com/apikey
+GEMINI_API_KEY=
+
+# Optional Groq transcription mentioned in the framework reference.
+# Configure separately: https://console.groq.com/keys
+GROQ_API_KEY=
diff --git a/AGENTS.md b/AGENTS.md
index 6c162ae..8c7ce7c 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -57,10 +57,13 @@ new personal footage. Never print secrets or copy private media into library exa
## Editing and timing
-Use ElevenLabs Scribe for transcription via `scripts/transcribe-elevenlabs.mjs`;
-reuse existing word timestamps whenever possible. Audio is uploaded to ElevenLabs
-and consumes the user's credits. Use their own locally configured credentials.
-The synthetic demo requires no API key. Avoid implicit Whisper transcription.
+Before choosing a transcription, asset-generation, or voiceover service, read
+`docs/TOOLS-AND-API-KEYS.md` and check the user's provider choice and local setup.
+ElevenLabs Scribe is Nate's default; honor requests for OpenAI Whisper, local
+Whisper, or another provider. Normalize verified word timestamps for the cutting
+tools. The included transcription script is ElevenLabs-only. Kie.ai is optional
+and needs a configured integration and credits. Reuse existing transcripts and
+assets; make any unapproved uploads or paid calls concrete before asking.
Cut silences first; use its edited video AND retimed transcript for cut-mistakes.
Review each mistake in context. Intentional repetition is not a mistake. Record
diff --git a/CLAUDE.md b/CLAUDE.md
index 6c162ae..8c7ce7c 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -57,10 +57,13 @@ new personal footage. Never print secrets or copy private media into library exa
## Editing and timing
-Use ElevenLabs Scribe for transcription via `scripts/transcribe-elevenlabs.mjs`;
-reuse existing word timestamps whenever possible. Audio is uploaded to ElevenLabs
-and consumes the user's credits. Use their own locally configured credentials.
-The synthetic demo requires no API key. Avoid implicit Whisper transcription.
+Before choosing a transcription, asset-generation, or voiceover service, read
+`docs/TOOLS-AND-API-KEYS.md` and check the user's provider choice and local setup.
+ElevenLabs Scribe is Nate's default; honor requests for OpenAI Whisper, local
+Whisper, or another provider. Normalize verified word timestamps for the cutting
+tools. The included transcription script is ElevenLabs-only. Kie.ai is optional
+and needs a configured integration and credits. Reuse existing transcripts and
+assets; make any unapproved uploads or paid calls concrete before asking.
Cut silences first; use its edited video AND retimed transcript for cut-mistakes.
Review each mistake in context. Intentional repetition is not a mistake. Record
diff --git a/README.md b/README.md
index 36943ce..7c48778 100644
--- a/README.md
+++ b/README.md
@@ -31,6 +31,17 @@ https://github.com/user-attachments/assets/4ad48761-d14c-407a-9dec-1b76a9afb04a
- **12 existing teaching projects** preserved from the original student kit.
- A synthetic starter composition and editing fixture that need no footage or API key.
+## Tools and optional services
+
+**Nate uses ElevenLabs Scribe to transcribe and Kie.ai to generate videos and
+image assets.** Bring your own API keys and credits when using those services.
+You can ask the assistant to use OpenAI Whisper or local Whisper instead, or
+supply an existing word-level transcript. Generated assets are optional.
+
+The local starter needs no paid transcription or generation API. See the
+[tools, accounts, and API-key guide](docs/TOOLS-AND-API-KEYS.md) for required tools,
+optional services, setup details, and prompts you can copy.
+
## Install
Install Node.js **22 or newer**, Git, FFmpeg (including ffprobe), and Chrome or
@@ -45,10 +56,10 @@ npm run setup
npm test
```
-Setup checks the tools and creates `.env` only if it is absent. For your own
-recordings, add an ElevenLabs API key with `speech_to_text` access to that local
-file. Transcription uploads audio and uses your ElevenLabs credits. Everything
-else in the starter runs locally. [Setup and troubleshooting](docs/SETUP.md).
+Setup checks the tools and creates `.env` only if it is absent. Add only the keys
+for services you choose. The included transcription helper uses ElevenLabs;
+Whisper and Kie.ai integrations need the setup described in the
+[tools guide](docs/TOOLS-AND-API-KEYS.md). [Setup and troubleshooting](docs/SETUP.md).
## Render your first example
diff --git a/docs/SETUP.md b/docs/SETUP.md
index 439b530..c07e4ba 100644
--- a/docs/SETUP.md
+++ b/docs/SETUP.md
@@ -1,5 +1,8 @@
# Setup and troubleshooting
+Start with [tools, accounts, and API keys](TOOLS-AND-API-KEYS.md) for Nate's
+ElevenLabs and Kie.ai choices, Whisper alternatives, and optional services.
+
Use Node 22+ (required by the pinned HyperFrames CLI), FFmpeg with ffprobe, and
Chrome/Chromium. Run `npm ci` in the repository root before any project commands.
`npm run setup` checks tool availability; `npx hyperframes doctor` diagnoses the
diff --git a/docs/SHORT-FORM.md b/docs/SHORT-FORM.md
index 0de19c7..9742cd7 100644
--- a/docs/SHORT-FORM.md
+++ b/docs/SHORT-FORM.md
@@ -13,7 +13,7 @@ The starter is landscape; the skill must set the reel's composition dimensions,
layout, and metadata to 1080x1920 before authoring. Keep all source footage and
outputs inside the new project.
-1. Transcribe with ElevenLabs Scribe, or reuse verified word timestamps.
+1. Transcribe with ElevenLabs Scribe, your chosen alternative, or reuse verified word timestamps.
2. Review mistakes in context and establish one frame-aligned edit decision list.
3. Explore three opening directions and choose a truthful promise and payoff.
4. Plan scenes, exact captions, moving footage, and sound cues against retained words.
@@ -22,10 +22,12 @@ outputs inside the new project.
6. Validate, preview, render a draft, inspect frames, and listen to the full edit.
7. Resolve timing, audio, and composition problems before exporting the final video.
-Transcription sends audio to ElevenLabs and uses your credits. Generated footage
-is optional, requires your own provider account, and may cost money. Prepare the
-asset proposal and honor the user's existing authorization. No paid service is
-needed to run the bundled synthetic validation example.
+Nate uses ElevenLabs Scribe for transcription and Kie.ai for generated video and
+image assets. Students need their own keys and credits for those services. Whisper
+is a transcription alternative, and existing footage can replace generated assets.
+See [tools, accounts, and API keys](TOOLS-AND-API-KEYS.md) for setup and copyable
+prompts. Honor existing authorization when proposing paid assets. The synthetic
+validation example needs no paid service.
## Validate from the repository root
diff --git a/docs/TOOLS-AND-API-KEYS.md b/docs/TOOLS-AND-API-KEYS.md
new file mode 100644
index 0000000..c8f6e62
--- /dev/null
+++ b/docs/TOOLS-AND-API-KEYS.md
@@ -0,0 +1,134 @@
+# Tools, accounts, and API keys
+
+Nate uses **ElevenLabs Scribe for transcription** and **Kie.ai for generated video
+and image assets**. These are his preferred services, not requirements for every
+edit. Bring your own accounts and credits, and configure only what you use.
+
+## What you need
+
+| Tool or service | When you need it | Setup and cost |
+| --- | --- | --- |
+| Codex or Claude Code | Following the agent-guided editing skills | Your own assistant access. External service API credits are separate. |
+| Node.js 22+, npm, Git, FFmpeg with ffprobe, and Chrome/Chromium | Local editing, preview, and rendering | Install on your computer. See [setup](SETUP.md). |
+| HyperFrames and GSAP | Video compositions and animation | Installed by `npm ci`. Rendering runs locally. Some templates use online fonts or CDNs; localize selected assets for final rendering. |
+| ElevenLabs Scribe | Nate's default for speech transcription | Your own key with `speech_to_text` access and available usage. The included helper sends audio to ElevenLabs. |
+| OpenAI Whisper API | Alternative cloud transcription | Your own OpenAI API key and API billing. Ask the assistant to configure the call and normalize its word timestamps. |
+| Local Whisper or whisper.cpp | Alternative transcription on your computer | Separate runtime and model installation. No cloud transcription key; needs model downloads, disk space, and local compute. |
+| Kie.ai | Optional generated moving B-roll, images, and other supported assets | Your own account, key, and credits. Existing footage and supplied assets can be used instead. |
+| Your footage, logos, fonts, music, and sound effects | Building the actual edit | Supply assets you can use. The kit does not include a stock-media or music subscription. |
+
+The synthetic starter and checks need no paid transcription or generation API.
+Choose a provider only when the next stage needs it.
+
+## ElevenLabs: the included helper
+
+Create a key in [ElevenLabs API settings](https://elevenlabs.io/app/settings/api-keys)
+and put it in the root `.env` as `ELEVENLABS_API_KEY`. Then run:
+
+```sh
+node scripts/transcribe-elevenlabs.mjs video-projects/my-video/assets/raw.mp4
+```
+
+The helper writes word-level JSON beside the input. Reuse a verified transcript
+instead of paying to transcribe the same take again. See the official
+[Scribe API documentation](https://elevenlabs.io/docs/api-reference/speech-to-text/convert)
+for current model options and usage details.
+
+## Whisper and other transcription alternatives
+
+You can prompt the assistant to use another provider. The bundled
+`transcribe-elevenlabs.mjs` script itself only calls ElevenLabs; changing the key
+does not switch its provider.
+
+For the OpenAI API, use `whisper-1`, `response_format="verbose_json"`, and
+`timestamp_granularities=["word"]`. Store `OPENAI_API_KEY` locally. Check current
+upload limits before sending long recordings; split audio when needed and retain
+chunk offsets. See [OpenAI's guide](https://developers.openai.com/api/docs/guides/speech-to-text).
+
+For local transcription, ask the assistant to configure
+[OpenAI Whisper](https://github.com/openai/whisper) or
+[whisper.cpp](https://github.com/ggml-org/whisper.cpp). Python Whisper requires
+Python and model dependencies; whisper.cpp has its own installation. Neither is
+installed by `npm ci`.
+
+The cutting tools require this common format, with **word-level times in seconds**
+on the exact source recording's timeline:
+
+```json
+{
+ "text": "Hello world.",
+ "audio_duration_secs": 2.0,
+ "words": [
+ { "text": "Hello", "start": 0.2, "end": 0.6 },
+ { "text": "world.", "start": 0.7, "end": 1.2 }
+ ]
+}
+```
+
+Ask the assistant to map a provider's `word` field to `text` when necessary,
+preserve numeric start/end times, and get full source duration from ffprobe.
+Paragraphs, SRT subtitles, and segment-only transcripts do not supply the required
+word timing. Request word timestamps or perform alignment; do not invent timing.
+Later edits must retime words using their EDL. Short-form plans additionally track
+`sourceStart` and `sourceEnd`.
+
+Copy either prompt:
+
+> Use OpenAI Whisper instead of ElevenLabs for this recording. Use my locally
+> configured API key, request word timestamps, and normalize the result for this
+> kit before cutting. Preserve the source and verify timing against its audio.
+
+> Use local Whisper for transcription. Check its installation and model downloads
+> first, then normalize the word timestamps for this kit. Keep transcription local
+> and use my supplied footage and assets without paid generation.
+
+## Kie.ai: generated video and assets
+
+Nate uses [Kie.ai](https://kie.ai/) for generated video and image assets. Set
+`KIE_API_KEY` in `.env` if you choose it. Follow the
+[official API guide](https://docs.kie.ai/) for keys, model-specific requests,
+task status, and credits.
+
+The short-form skill describes a Kie workflow; the kit does **not** bundle a
+universal Kie generation script or connected account. Ask the assistant to configure
+your selected model using current documentation. Adding a key alone does not
+create an integration. Model inputs, formats, availability, and credit costs vary.
+
+Prepare the proposed assets and expected cost before paid work that you have not
+already authorized. Keep task IDs, reuse successful outputs, and save chosen
+results inside your private project's assets. Inspect their motion, duration,
+aspect ratio, and framing in the final edit.
+
+> Use Kie.ai for moving B-roll in the approved scenes. Check my local configuration
+> and current model documentation, then show me the assets you propose and expected
+> credit cost. Reuse existing footage wherever it fits.
+
+## Other optional tools referenced by bundled skills
+
+These are workflow options, not a claim that every example used them:
+
+- **Narration:** HyperFrames guides describe local Kokoro TTS with extra runtime
+ and model setup. ElevenLabs and HeyGen TTS are also mentioned; each needs your
+ service access and a configured API or connector.
+- **Website capture:** Gemini can optionally provide richer image descriptions.
+ Basic capture does not require that key.
+- **Other transcription:** the framework reference mentions Groq Whisper. It
+ needs its own key and the same transcript normalization checks.
+- **Older integration examples:** ClickUp or OpenAI access is needed only when
+ your selected project actually calls those services.
+- **Helper runtimes:** some optional scripts require Python. Playwright is
+ included for relevant browser workflows; browser binaries may need installation.
+
+A skill's mention of an MCP tool does not install that integration. Ask the
+assistant to check which connectors or API clients are available first.
+
+## Configure only what you use
+
+`npm run setup` creates `.env` from [`.env.example`](../.env.example) if absent.
+Leave unused entries blank. Existing `.env` files are preserved, so add newly
+needed entries yourself. The ElevenLabs helper loads the root `.env`; other
+providers need their own client or environment loading configured by the assistant.
+Keep credentials out of compositions, screenshots, and commits.
+
+Check current provider pricing and your balance before paid calls. The student
+kit does not include transcription, generation, voiceover, or assistant credits.
diff --git a/docs/WORKFLOW.md b/docs/WORKFLOW.md
index 7886c13..6721efd 100644
--- a/docs/WORKFLOW.md
+++ b/docs/WORKFLOW.md
@@ -9,7 +9,10 @@ at `video-projects/my-video/assets/raw.mp4`. Preserve this original.
node scripts/transcribe-elevenlabs.mjs video-projects/my-video/assets/raw.mp4
```
-This uploads audio to ElevenLabs. You can instead supply an existing word-level
+This is Nate's default and uploads audio to ElevenLabs using your key and credits.
+You may choose OpenAI Whisper, local Whisper, or another provider instead; see
+[provider setup and transcript normalization](TOOLS-AND-API-KEYS.md).
+You can also supply an existing word-level
JSON transcript with `words: [{text, start, end}]` and `audio_duration_secs`.
## 2. Remove dead air
diff --git a/scripts/setup.mjs b/scripts/setup.mjs
index 80471a9..32cc5a2 100644
--- a/scripts/setup.mjs
+++ b/scripts/setup.mjs
@@ -11,7 +11,8 @@ for (const [command, args] of [['node', ['--version']], ['ffmpeg', ['-version']]
else console.log((result.stdout || result.stderr).split(/\r?\n/)[0]);
}
if (Number(process.versions.node.split('.')[0]) < 22) { console.error('Node 22+ required.'); missing = true; }
-if (!existsSync(join(root, '.env'))) { copyFileSync(join(root, '.env.example'), join(root, '.env')); console.log('Created local .env. Add your own key only for transcription.'); }
+if (!existsSync(join(root, '.env'))) { copyFileSync(join(root, '.env.example'), join(root, '.env')); console.log('Created local .env. Fill only the keys your chosen services need.'); }
else console.log('Existing .env preserved.');
+console.log('Service setup and alternatives: docs/TOOLS-AND-API-KEYS.md');
console.log('Next: npx hyperframes doctor (Chrome diagnostics), then npm run demo.');
process.exitCode = missing ? 1 : 0;