ax-check.com
by Gauge
S
Speech-to-Text Cloud
speech-to-text.cloud
ax-check.com
Checked 18 September 2026
D
40/100
Agents need your login
Pricing is clear, but developer docs and setup guidance are missing
Plan prices and minute limits are stated openly on the pricing page, but the guide index skips docs, API, and setup instructions entirely.
5 fixes to start with
Concrete changes, backed by what we found.
Tell agents where to start and organize the links in /llms.txt.
llms.txt lists capabilities and pricing but gives no navigation guidance to documentation sections.
List every offered API, MCP, and skills surface in /llms.txt.
llms.txt never mentions the API, MCP, or skills surfaces offered elsewhere on the site.
Recorded coding sessions
Same task, 3 models, no speech-to-text.cloud account. Recorded 18 September 2026 on Pi.
3 recorded · 1m 46s
1m 46s · recorded
Session findings
Select a finding to see it in context.
›Thinking…
The user is asking me to create a simple example using "Speech-to-Text Cloud" (Google Cloud Speech-to-Text). First, let me check the repository to grasp the context.
I should check whether the repository exists, whether there are credentials or SDKs, and so on. The instructions are: use the hosted product via the SDK or API, and keep it lightweight. If there's no executable developer workflow (e.g., no credentials, no API access), I should clearly communicate that fact and stop.
Let's take a look at the environment.
✓$ ls -la /sandbox/repo && cat /sandbox/repo/README* 2>/dev/null | head -50✓└{"content":[{"text":"total 16\ndrwxr-xr-x 3 root root 4096 Sep 18 08:00 .\ndrwxr-xr-x 7 root root 4096 Sep 18 21:47 ..\ndrwxr-xr-x 8 root root 4096 Sep 18 08:00…ctrl+o
{"content":[{"text":"total 16\ndrwxr-xr-x 3 root root 4096 Sep 18 08:00 .\ndrwxr-xr-x 7 root root 4096 Sep 18 21:47 ..\ndrwxr-xr-x 8 root root 4096 Sep 18 08:00 .git\n-rw-r--r-- 1 root root 408 Sep 18 08:00 .gitignore\n","type":"text"}]}×$ env | grep -iE "GOOGLE|GCLOUD|GCP|API_KEY|SPEECH" ; ls ~/.config/gcloud 2>/dev/null; which gcloud python3 pip3 node 2>/dev/null✓└{"content":[{"text":"PI_GATEWAY_API_KEY=unused\n/usr/local/bin/python3\n/usr/local/bin/pip3\n/usr/local/bin/node\n\n\nCommand exited with code 1","type":"text"}…ctrl+o
{"content":[{"text":"PI_GATEWAY_API_KEY=unused\n/usr/local/bin/python3\n/usr/local/bin/pip3\n/usr/local/bin/node\n\n\nCommand exited with code 1","type":"text"}],"details":{}}›Thinking…
The repository is empty except for .gitignore. There's no gcloud CLI, no Google Cloud credentials, and no API key for Speech-to-Text. Google Cloud Speech-to-Text requires a GCP project, billing, service account credentials, and the API must be enabled. I can't actually execute the calls. However, I can create sample code using the SDK (google-cloud-speech). The question is: "If this product doesn't have an executable developer workflow, communicate that clearly and stop."
The user wants to build a simple example using the hosted product via SDK/API. Google Cloud Speech-to-Text does have a developer workflow (SDK/API), but it requires credentials. Without credentials, I can't actually run it. Should I stop? "If you get stuck, let me know."
I think the reasonable response is this: create a minimal Python example using the google-cloud-speech SDK, explain the pricing, honestly convey that it can't be executed without GCP credentials, and note that it's manageable. But "staying lightweight" and "if you get stuck, let me know" — the lack of credentials is a blocker for execution, but I can still provide a code example. The instructions say: "If this product doesn't have an executable developer workflow, communicate that clearly and stop." Speech-to-Text does have a developer workflow, but without credentials I can't execute it. I'll create a sample file, explain the pricing, clearly state that execution is blocked due to the lack of credentials, and stop there.
Should I try pip install google-cloud-speech? It might work but it's heavy; without credentials I can't run the call. I can write the example and verify whether the import works. To make the example verifiable, let me try quickly installing the SDK. Keep it lightweight — maybe just writing the code is enough. A simple pip install to verify the SDK works is fine (it's not a long-running process). Let me give it a try.
✓$ pip3 install -q google-cloud-speech 2>&1 | tail -2; python3 -c "import google.cloud.speech; print('sdk ok', google.cloud.speech.__version__ if hasattr(google.cloud.speech,'__version__') else '')"✓└{"content":[{"text":"sdk ok 2.40.0\n","type":"text"}]}
›Thinking…
The SDK installs and imports without issues. Next, I'll write a minimal example. Can't actually run the transcription without credentials, so I'll note that clearly. Writing the file now.
The SDK installs fine. Let me write a minimal example.
✓write /sandbox/repo/stt_example.py✓└{"content":[{"text":"Successfully wrote 1394 bytes to /sandbox/repo/stt_example.py","type":"text"}]}
›Thinking…
Also, should I mention gcs_uri for longer files? Keep it lightweight. Next, answer the question: pricing, ease of management, blocked status.
Done. Here's the summary:
What I built
/sandbox/repo/stt_example.py — a ~40-line Python script using the official google-cloud-speech SDK (v2.40.0, verified importable). It reads a local WAV file (16 kHz, 16-bit PCM) and calls the synchronous recognize endpoint, printing each result's transcript with confidence. No local services, no emulators, no long-running processes.
Blocked — honest status
I cannot actually run it here. The sandbox has no Google Cloud credentials (no service account JSON, no ADC, no gcloud CLI). Speech-to-Text requires a GCP project with billing and enabled API — none of which I can provision from here. To run it yourself:
- Enable billing + the API:
gcloud services enable speech.googleapis.com - Set credentials:
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/sa-key.json python stt_example.py audio.wav
The code itself follows the current SDK surface, so it should work as-is once credentials exist.
Pricing (Speech-to-Text v2)
- Free tier: first 60 minutes of audio per month, free.
- After that, per-second billing on audio duration, roughly:
- Standard recognition (V1/V2):
$0.024/min ($1.44/hr) - Data logging / medical / enhanced models: somewhat higher (~$0.03–0.09/min)
- Chirp / large speech-to-text models: ~$0.016–0.02/min depending on sync vs. batch (batch is cheaper)
- Standard recognition (V1/V2):
- Sync vs. async (
long_running_recognize) doesn't change the rate for most models, but async is required for files > ~60s (audio must live in GCS viags://URI). - You pay for audio duration sent, not compute time. Exact rates vary by region/model — check the official pricing page before committing.
Will this be easy for me to manage?
Yes, with one caveat. It's a clean developer workflow: standard SDK, simple request/response API, no infrastructure to maintain, stateless calls, predictable pricing. The caveat is purely credentialing — every action requires a provisioned GCP project I can't create myself. If you can drop a service-account key into the environment, I can iterate on this fully (test runs, tuning language/model options, batch mode, streaming, etc.). Otherwise I can only write and reason about the code, not execute it.
AX Fundamentals
23 checks. Open an area for details.
This check is a quick preview. Test more tasks and models with Gauge Agents.
Test your product with Gauge Agents