[BUG] Windows 11: "backend-specific error"

[fmos]

Bug Description

When I try to start transcription by pressing the shortcut, the app shows this overlay error message:

System Information

App Version:

v0.7.10

Operating System:

Edition Windows 11 Pro
Version 24H2
Installiert am ‎13.‎10.‎2025
Betriebssystembuild 26100.7840
Funktionspaket Windows Feature Experience Pack 1000.26100.291.0

CPU:

AMD Ryzen 9 PRO 7940HS

GPU:

Radeon 780M Graphics

Logs

[2026-03-10][08:06:17][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-10][08:06:17][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-10][08:06:17][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-10][08:06:17][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-10][08:06:17][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (Razer BlackShark V2 Pro 2.4)")
Sample rate: 48000
Channels: 2
Format: F32
[2026-03-10][08:06:17][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:17][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:17][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:17][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 60.8276ms
[2026-03-10][08:06:17][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle
[2026-03-10][08:06:30][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-10][08:06:30][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-10][08:06:30][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-10][08:06:30][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-10][08:06:30][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (Razer BlackShark V2 Pro 2.4)")
Sample rate: 48000
Channels: 2
Format: F32
[2026-03-10][08:06:30][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:30][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:30][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-10][08:06:30][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 59.432ms
[2026-03-10][08:06:30][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle
[2026-03-10][08:06:44][tao::platform_impl::platform::event_loop::runner][WARN] NewEvents emitted without explicit RedrawEventsCleared
[2026-03-10][08:06:44][tao::platform_impl::platform::event_loop::runner][WARN] RedrawEventsCleared emitted without explicit MainEventsCleared
[2026-03-10][08:07:02][tao::platform_impl::platform::event_loop::runner][WARN] NewEvents emitted without explicit RedrawEventsCleared
[2026-03-10][08:07:02][tao::platform_impl::platform::event_loop::runner][WARN] RedrawEventsCleared emitted without explicit MainEventsCleared

[cjpais]

@ferologics mind taking a look or any insight? Not sure if this is a new bug or just getting surfaced since we throw some errors now

[StepaevAnton]

Experiencing the same issue, but it occurs on app startup (also on Windows 11, app version 0.7.10).

[2026-03-14][15:57:05][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Intel® Smart Sound Technology for Digital Microphones")
Sample rate: 48000
Channels: 4
Format: F32
[2026-03-14][15:57:05][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: The parameter is incorrect. (0x80070057)

On version 0.7.9 and earlier, the crash occurred during transcribing rather than at startup.

Microphone access is enabled in Windows privacy settings.

[fmos]

Unsurprisingly, but for completeness: The issue remains for today's v0.7.11 with practically identical log messages:

[2026-03-17][15:12:24][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-17][15:12:24][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-17][15:12:24][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (Razer BlackShark V2 Pro 2.4)")
Sample rate: 48000
Channels: 2
Format: F32
[2026-03-17][15:12:24][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-17][15:12:24][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-17][15:12:24][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-17][15:12:24][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 304.697ms
[2026-03-17][15:12:24][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle

[cjpais]

@fmos @StepaevAnton Would you mind trying: #1084 (comment)

It ideally will fix this issue. Can't say for certain. I will try to take a deeper look soon.

[fmos]

@cjpais Don't mind at all. Thank you for your attention to this matter!

I used the handy-pr-1084-x86_64-pc-windows-msvc artifact. Unfortunately, the issue persists (without triggering any of the new warnings in die PR):

[2026-03-18][11:59:27][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-18][11:59:27][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-18][11:59:27][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-18][11:59:27][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-18][11:59:27][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (Razer BlackShark V2 Pro 2.4)")
Sample rate: 48000
Channels: 1
Format: F32
[2026-03-18][11:59:27][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-18][11:59:27][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-18][11:59:27][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-18][11:59:27][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 74.1102ms
[2026-03-18][11:59:27][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle

[cjpais]

@fmos thanks for testing, I will look into it a little further. Do you have other microphones and does it occur with them?

[fmos]

@cjpais Welcome. I tested again with v0.7.12 and each of my three microphones - reproducing the issue each time:

[2026-03-19][07:50:49][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-19][07:50:49][handy_app_lib::managers::transcription][DEBUG] Starting to load model: parakeet-tdt-0.6b-v3
[2026-03-19][07:50:49][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-19][07:50:49][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-19][07:50:49][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-19][07:50:49][transcribe_rs::onnx::session][INFO] Loading int8 model: C:\Users\stanke\AppData\Roaming\com.pais.handy\models\parakeet-tdt-0.6b-v3-int8\encoder-model.int8.onnx
[2026-03-19][07:50:49][transcribe_rs::onnx::session][INFO] Loading int8 model: C:\Users\stanke\AppData\Roaming\com.pais.handy\models\parakeet-tdt-0.6b-v3-int8\decoder_joint-model.int8.onnx
[2026-03-19][07:50:49][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (Razer BlackShark V2 Pro 2.4)")
Sample rate: 48000
Channels: 1
Format: F32
[2026-03-19][07:50:49][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:49][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:49][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:49][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 296.544ms
[2026-03-19][07:50:49][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=audio_signal, type=Tensor { ty: Float32, shape: [-1, 128, -1], dimension_symbols: SymbolicDimensions(["audio_signal_dynamic_axes_1", "", "audio_signal_dynamic_axes_2"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=length, type=Tensor { ty: Int64, shape: [-1], dimension_symbols: SymbolicDimensions(["length_dynamic_axes_1"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=outputs, type=Tensor { ty: Float32, shape: [-1, 1024, -1], dimension_symbols: SymbolicDimensions(["Transposeoutputs_dim_0", "", "Transposeoutputs_dim_2"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=encoded_lengths, type=Tensor { ty: Int64, shape: [-1], dimension_symbols: SymbolicDimensions(["length_dynamic_axes_1"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=encoder_outputs, type=Tensor { ty: Float32, shape: [-1, 1024, -1], dimension_symbols: SymbolicDimensions(["encoder_outputs_dynamic_axes_1", "", "encoder_outputs_dynamic_axes_2"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=targets, type=Tensor { ty: Int32, shape: [-1, -1], dimension_symbols: SymbolicDimensions(["targets_dynamic_axes_1", "targets_dynamic_axes_2"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=target_length, type=Tensor { ty: Int32, shape: [-1], dimension_symbols: SymbolicDimensions(["target_length_dynamic_axes_1"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=input_states_1, type=Tensor { ty: Float32, shape: [2, -1, 640], dimension_symbols: SymbolicDimensions(["", "input_states_1_dynamic_axes_1", ""]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=input_states_2, type=Tensor { ty: Float32, shape: [2, -1, 640], dimension_symbols: SymbolicDimensions(["", "input_states_2_dynamic_axes_1", ""]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=outputs, type=Tensor { ty: Float32, shape: [-1, -1, -1, 8198], dimension_symbols: SymbolicDimensions(["Addoutputs_dim_0", "Addoutputs_dim_1", "Addoutputs_dim_2", ""]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=prednet_lengths, type=Tensor { ty: Int32, shape: [-1], dimension_symbols: SymbolicDimensions(["target_length_dynamic_axes_1"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=output_states_1, type=Tensor { ty: Float32, shape: [2, -1, 640], dimension_symbols: SymbolicDimensions(["", "Concatoutput_states_1_dim_1", ""]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=output_states_2, type=Tensor { ty: Float32, shape: [2, -1, 640], dimension_symbols: SymbolicDimensions(["", "Concatoutput_states_2_dim_1", ""]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=waveforms, type=Tensor { ty: Float32, shape: [-1, -1], dimension_symbols: SymbolicDimensions(["batch_size", "N"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model input: name=waveforms_lens, type=Tensor { ty: Int64, shape: [-1], dimension_symbols: SymbolicDimensions(["batch_size"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=features, type=Tensor { ty: Float32, shape: [-1, 128, -1], dimension_symbols: SymbolicDimensions(["batch_size", "", "T"]) }
[2026-03-19][07:50:51][transcribe_rs::onnx::session][INFO] Model output: name=features_lens, type=Tensor { ty: Int64, shape: [-1], dimension_symbols: SymbolicDimensions(["batch_size"]) }
[2026-03-19][07:50:51][transcribe_rs::decode::tokens][INFO] Loaded 8193 vocab tokens from "C:\\Users\\stanke\\AppData\\Roaming\\com.pais.handy\\models\\parakeet-tdt-0.6b-v3-int8\\vocab.txt"
[2026-03-19][07:50:51][transcribe_rs::onnx::parakeet][INFO] Loaded vocabulary with 8193 tokens, blank_idx=8192
[2026-03-19][07:50:51][handy_app_lib::managers::transcription][DEBUG] Successfully loaded transcription model: parakeet-tdt-0.6b-v3 (took 1920ms)
[2026-03-19][07:50:57][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-19][07:50:57][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-19][07:50:57][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-19][07:50:57][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-19][07:50:57][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Mikrofon (HD Pro Webcam C920)")
Sample rate: 48000
Channels: 2
Format: F32
[2026-03-19][07:50:57][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:57][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:57][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:50:57][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 63.5986ms
[2026-03-19][07:50:57][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle
[2026-03-19][07:51:04][handy_app_lib::actions][DEBUG] TranscribeAction::start called for binding: transcribe
[2026-03-19][07:51:04][enigo::platform::win_impl][DEBUG] �[93mlocation()�[0m
[2026-03-19][07:51:04][handy_app_lib::actions][DEBUG] Microphone mode - always_on: false
[2026-03-19][07:51:04][handy_app_lib::actions][DEBUG] On-demand mode: Starting recording first, then audio feedback
[2026-03-19][07:51:04][handy_app_lib::audio_toolkit::audio::recorder][INFO] Using device: Ok("Microphone Array (AMD Audio Device)")
Sample rate: 48000
Channels: 4
Format: F32
[2026-03-19][07:51:04][handy_app_lib::audio_toolkit::audio::recorder][ERROR] Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:51:04][handy_app_lib::managers::audio][ERROR] Failed to open microphone stream: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:51:04][handy_app_lib::actions][DEBUG] Failed to start recording: Failed to open recorder: Failed to build input stream: A backend-specific error has occurred: Falscher Parameter. (0x80070057)
[2026-03-19][07:51:04][handy_app_lib::actions][DEBUG] TranscribeAction::start completed in 582.8027ms
[2026-03-19][07:51:04][handy_app_lib::transcription_coordinator][DEBUG] Start for 'transcribe' did not begin recording; staying idle

[cjpais]

@fmos thanks, the 0.7.12 has no fixes for this, only the build I specified earlier, that's the one I'm looking for the testing on

[fmos]

@cjpais Yes, the results in #990 (comment) used the special build. I just thought, I'd re-test after todays new release.

[StepaevAnton]

Also tried handy-pr-1084-x86_64-pc-windows-msvc from test build.
Confirming exactly same error :(

[cjpais]

Thanks for all the confirmation

@VirenMohindra @tanshkoul if you have any ideas would be nice

[VirenMohindra]

i think i found the root cause. supported_input_configs() returns exclusive-mode formats on windows WASAPI, but CPAL opens streams in shared mode which needs the mixer's native format. the fix should be to just use default_input_config() on windows directly

the logs show "Intel® Smart Sound Technology" is logging Channels: 4, Format: F32, which is clearly not that device's shared-mode mix format. The "Razer headset" showing Channels: 1 or 2, Format: F32 is similarly suspicious